liqiang/ai-agent-book
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

ai-agent-book 精选快照（<2MB 代码与文档，来自 github.com/bojieli/ai-agent-book）
Build latest book artifacts / build (push) Canceled after 0s
Details
dependency resolution / resolve (3.11) (push) Canceled after 0s
Details
dependency resolution / resolve (3.13) (push) Canceled after 0s
Details
deploy-pages / build (push) Canceled after 0s
Details
deploy-pages / deploy (push) Canceled after 0s
Details
i18n consistency check / check (push) Canceled after 0s
Details
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
Details
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
Details
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
Details
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
Details
web-search-agent tests / test (push) Canceled after 0s
Details
web-search-agent tests / agentbook (push) Canceled after 0s
Details

Browse Source
This commit is contained in:
liqiang
2026-08-20 13:12:50 +00:00
commit b119135836
10275 changed files with 3284984 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
chapter9/gaia-experience/.gitignore
+1
View File
@@ -0,0 +1 @@
output/
chapter9/gaia-experience/AWorld/Dockerfile_gaia
+43
View File
@@ -0,0 +1,43 @@
FROM python:3.11-slim AS base
RUN mkdir -p /app/aworld
WORKDIR /app/aworld
RUN pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
RUN rm -fv /etc/apt/sources.list.d/* && \
echo "deb [trusted=yes] https://mirrors.aliyun.com/debian bookworm main contrib non-free non-free-firmware" > /etc/apt/sources.list && \
echo "deb [trusted=yes] https://mirrors.aliyun.com/debian bookworm-updates main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
echo "deb [trusted=yes] https://mirrors.aliyun.com/debian bookworm-backports main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
echo "deb [trusted=yes] https://mirrors.aliyun.com/debian-security bookworm-security main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
RUN apt-get update && apt-get install -y wget unzip openssh-client procps nodejs npm
# Config Env
RUN npx playwright install chrome
RUN npm install @playwright/mcp @negokaz/excel-mcp-server
# GAIA Docker Image
FROM base AS gaia_env
# Install aworld
COPY ./aworld aworld
COPY ./setup.py setup.py
RUN python setup.py install
# Copy examples
COPY ./examples/ examples
# Create GAIA benchmark directory
RUN mkdir -p gaia-benchmark/fs && \
mkdir -p gaia-benchmark/logs && \
mkdir -p static
ENV PYTHONPATH=/app/aworld${PYTHONPATH:+:${PYTHONPATH}}
ENV GAIA_DATASET_PATH=/app/aworld/examples/gaia/GAIA/2023
ENV LOG_FILE_PATH=/app/aworld/gaia-benchmark/logs
CMD [ "python", "examples/gaia/gaia_agent_server.py" ]
chapter9/gaia-experience/AWorld/LICENSE
+21
View File
@@ -0,0 +1,21 @@
MIT License
Copyright (c) 2025 [Inclusion AI]
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
chapter9/gaia-experience/AWorld/README.md
+736
View File
@@ -0,0 +1,736 @@
<div align="center">
# AWorld: The Agent Runtime for Self-Improvement
</div>
<h4 align="center">
*"Self-awareness: the hardest problem isn't solving within limits, it's discovering one's own limitations"*
[![Twitter Follow][twitter-image]][twitter-url]
[![WeChat QR Code][wechat-image]][wechat-url]
[![Discord][discord-image]][discord-url]
[![License: MIT][license-image]][license-url]
[![DeepWiki][deepwiki-image]][deepwiki-url]
[![arXiv][arxiv-image]][arxiv-url]
[![Tutorial][tutorial-image]][tutorial-url]
<!-- [![arXiv][arxiv-image]][arxiv-url] -->
</h4>
<h4 align="center">
[中文版](./README_zh.md) |
[Quickstart](#️-quickstart) |
[Architecture](#️-architecture-design-principles) |
[Applications](#-applications) |
[Contributing](#contributing) |
[Appendix](#appendix-web-client-usage)
</h4>
---
<!-- **AWorld (Agent World)** is a next-generation framework for agent learning with three key characteristics:
1. **Plug-and-Play:** Box up complex modules with bulletproof protocols and zero-drama state control.
2. **Cloud-Native Velocity:** Train smarter agents that evolve their own brains—prompts, workflows, memory, and tools—on the fly.
3. **Self-Awareness**: Synthesize the agent's own knowledge and experience to achieve ultimate self-improvement. -->
![](./readme_assets/heading_banner.png)
**AWorld (Agent World)** is a next-generation framework engineered for agent self-improvement at scale. Powered by the capabilities above, we enable AI agents to continuously evolve by learning from their own knowledge and experiences across various environments. Using AWorld, you can:
1. **Build workflows**: Design and implement automated task sequences. [Docs](https://inclusionai.github.io/AWorld/Quickstart/workflow_construction/)
2. **Build agents**: Create intelligent AI agents with mcp tools. [Docs](https://inclusionai.github.io/AWorld/Quickstart/agent_construction/)
3. **Build Multi-Agent Systems (MAS)**: Orchestrate collaborative agent ecosystems. [Docs](https://inclusionai.github.io/AWorld/Quickstart/multi-agent_system_construction/)
4. **Train agents efficiently**: Optimize performance within MAS across various environments
---
**Collective Intelligence** 🚀
Demonstrating collective intelligence across diverse domains. Join us in the ongoing projects!
<!--
| **Category** | **Achievement** | **Performance** | **Key Innovation** | **Date** |
|:-------------|:----------------|:----------------|:-------------------|:----------|
| **🤖 Agent** | **GAIA Benchmark Excellence** [![][GAIA]](https://huggingface.co/spaces/gaia-benchmark/leaderboard) | Pass@1: **67.89**, Pass@3: **83.49** (109 tasks) [![][Code]](./examples/gaia/README_GUARD.md) | Multi-agent system stability & orchestration [![][Paper]](https://arxiv.org/abs/2508.09889) | 2025/08/06 |
| **🧠 Reasoning** | **IMO 2025 Problem Solving** [![][IMO]](https://www.imo-official.org/year_info.aspx?year=2025) | 5/6 problems solved in 6 hours [![][Code]](examples/imo/README.md) | Multi-agent collaboration beats solo models | 2025/07/25 |
-->
<table style="width: 100%; border-collapse: collapse; table-layout: fixed;">
<thead>
<tr>
<th style="width: 30%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Category</th>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Achievement</th>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Performance</th>
<th style="width: 25%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Key Innovation</th>
<th style="width: 5%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Date</th>
</tr>
</thead>
<tbody>
<tr>
<td style="padding: 8px; vertical-align: top;">🤖 Agent</td>
<td style="padding: 8px; vertical-align: top;">
<strong>GAIA Benchmark <br>Excellence</strong>
<br>
<a href="https://huggingface.co/spaces/gaia-benchmark/leaderboard" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/GAIA-Leaderboard-blue" alt="GAIA">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">
Pass@1: <strong>67.89</strong> <br>
Pass@3: <strong>83.49</strong>
<br> (109 tasks)
<a href="./examples/gaia/README_GUARD.md" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Code-README-green" alt="Code">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">
Multi-agent system <br>stability & orchestration
<br>
<a href="https://arxiv.org/abs/2508.09889" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Paper-arXiv-red" alt="Paper">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">2025/08/06</td>
</tr>
<tr>
<td style="padding: 8px; vertical-align: top;">🧠 Reasoning</td>
<td style="padding: 8px; vertical-align: top;">
<strong>IMO 2025 <br>Problem Solving</strong>
<br>
<a href="https://www.imo-official.org/year_info.aspx?year=2025" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/IMO-2025-blue" alt="IMO">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">
<strong>5/6</strong> problems <br>solved in 6 hours
<br>
<a href="examples/imo/README.md" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Code-README-green" alt="Code">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">Multi-agent collaboration <br>beats solo models</td>
<td style="padding: 8px; vertical-align: top;">2025/07/25</td>
</tr>
</tbody>
</table>
<details>
<summary style="font-size: 1.2em;font-weight: bold;"> 🌏 View ongoing projects </summary>
<!--
| **Category** | **Achievement** | **Status** | **Expected Impact** |
|:-------------|:----------------|:-----------|:-------------------|
| **🖼️ Multi-Modal** | Advanced OS / Web Interaction | In Progress | Visual reasoning & environment understanding |
| **💻 Code** | Advanced installation, coding, testing, debugging, etc. ability | In Progress | Automated software engineering capabilities |
| **🔧 Tool Use** | Advanced multi-turn function call | Comming soon | Impact the real world |
-->
<table style="width: 100%; border-collapse: collapse; table-layout: fixed;">
<thead>
<tr>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Category</th>
<th style="width: 35%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Achievement</th>
<th style="width: 10%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Status</th>
<th style="width: 35%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Expected Impact</th>
</tr>
</thead>
<tbody>
<tr>
<td style="padding: 8px; vertical-align: top;">🖼️ Multi-Modal</td>
<td style="padding: 8px; vertical-align: top;">Advanced OS / Web Interaction</td>
<td style="padding: 8px; vertical-align: top;">In Progress</td>
<td style="padding: 8px; vertical-align: top;">Visual reasoning <br>environment understanding</td>
</tr>
<tr>
<td style="padding: 8px; vertical-align: top;">💻 Code</td>
<td style="padding: 8px; vertical-align: top;">Advanced installation, coding, <br>testing, debugging, etc. ability</td>
<td style="padding: 8px; vertical-align: top;">In Progress</td>
<td style="padding: 8px; vertical-align: top;">Automated software <br>engineering capabilities</td>
</tr>
<tr>
<td style="padding: 8px; vertical-align: top;">🔧 Tool Use</td>
<td style="padding: 8px; vertical-align: top;">Advanced multi-turn function call</td>
<td style="padding: 8px; vertical-align: top;">Coming soon</td>
<td style="padding: 8px; vertical-align: top;">Impact the real world</td>
</tr>
</tbody>
</table>
</details>
---
**Self-Improvement: Surpassing Collective Intelligence** 🌱
`Agents` can run in various `Env`, collect both positive and negative `Experience`, and learn through `Training`.
<table style="width: 100%; border-collapse: collapse; table-layout: fixed;">
<thead>
<tr>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Agents</th>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Env</th>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Experience</th>
<th style="width: 25%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Training</th>
<th style="width: 15%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">Code</th>
</tr>
</thead>
<tbody>
<tr>
<td style="padding: 8px; vertical-align: top;">GAIA Agent</td>
<td style="padding: 8px; vertical-align: top;">
Terminal, code, search, playwright, and 4 additional tools
</td>
<td style="padding: 8px; vertical-align: top;">
Collected from 165 samples in the GAIA validation dataset <br>
<a href="https://huggingface.co/datasets/gaia-benchmark/GAIA/tree/main/2023/validation" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Dataset-Training-8AB07D" alt="Training Dataset">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">
Rollout, reward calculation, and gradient updates via GRPO
</td>
<td style="padding: 8px; vertical-align: top;">
Three lines of code to run.
<br>
<a href="./train/README.md" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Code-README-green" alt="Code">
</a>
</td>
</tr>
</tbody>
</table>
---
# 🏃‍♀️ Quickstart
## Prerequisites
> [!TIP]
> Python>=3.11
```bash
git clone https://github.com/inclusionAI/AWorld && cd AWorld
pip install .
```
## Hello world examples
We introduce the concepts of `Agent` and `Runners` to help you get started quickly.
For parallel task execution, see the [parallel run examples](examples/parallel_run/README.md).
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
summarizer = Agent(
name="Summary Agent",
system_prompt="You specialize at summarizing.",
)
result = Runners.sync_run(
input="Tell me a succint history about the universe",
agent=summarizer,
)
```
In parallel, we introduce the concepts of `Swarm` to construct a group of agents.
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
from aworld.core.agent.swarm import Swarm
researcher = Agent(
name="Research Agent",
system_prompt="You specialize at researching.",
)
summarizer = Agent(
name="Summary Agent",
system_prompt="You specialize at summarizing.",
)
# Create agent group with collaborative workflow
group = Swarm(topology=[(researcher, summarizer)])
result = Runners.sync_run(
input="Tell me a complete history about the universe",
swarm=group,
)
```
Finally, run your own agents or teams
```bash
# Set LLM credentials
export LLM_MODEL_NAME="gpt-4"
export LLM_API_KEY="your-api-key-here"
export LLM_BASE_URL="https://api.openai.com/v1"
# Run
python /path/to/agents/or/teams
```
<details>
<summary style="font-size: 1.2em;font-weight: bold;"> 🌏 Click to View Advanced Usages </summary>
### Pass AgentConfig Explicitly
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
from aworld.config.conf import AgentConfig
from aworld.core.agent.swarm import Swarm
gpt_conf = AgentConfig(
llm_provider="openai",
llm_model_name="gpt-4o",
llm_api_key="<OPENAI_API_KEY>",
llm_temperature=0.1,
)
openrouter_conf = AgentConfig(
llm_provider="openai",
llm_model_name="google/gemini-2.5-pro",
llm_api_key="<OPENROUTER_API_KEY>",
llm_base_url="https://openrouter.ai/api/v1"
llm_temperature=0.1,
)
researcher = Agent(
name="Research Agent",
conf=gpt_conf,
system_prompt="You specialize at researching.",
)
summarizer = Agent(
name="Summary Agent",
conf=openrouter_conf,
system_prompt="You specialize at summarizing.",
)
# Create agent group with collaborative workflow
group = Swarm(topology=[(researcher, summarizer)])
result = Runners.sync_run(
input="Tell me a complete history about the universe",
swarm=group,
)
```
### Agent Equipped with MCP Tools
```python
import os
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
mcp_config = {
"mcpServers": {
"GorillaFileSystem": {
"type": "stdio",
"command": "python",
"args": ["examples/BFCL/mcp_tools/gorilla_file_system.py"],
},
}
}
file_sys = Agent(
name="file_sys_agent",
system_prompt=(
"You are a helpful agent to use "
"the standard file system to perform file operations."
),
mcp_servers=mcp_config.get("mcpServers", []).keys(),
mcp_config=mcp_config,
)
result = Runners.sync_run(
input=(
"use mcp tools in the GorillaFileSystem server "
"to perform file operations: "
"write the content 'AWorld' into "
"the hello_world.py file with a new line "
"and keep the original content of the file. "
"Make sure the new and old "
"content are all in the file; "
"and display the content of the file"
),
agent=file_sys,
)
```
### Agent Integrated with Memory
It is recommended to use `MemoryFactory` to initialize and access Memory instances.
```python
from aworld.memory.main import MemoryFactory
from aworld.core.memory import MemoryConfig, MemoryLLMConfig
# Simple initialization
memory = MemoryFactory.instance()
# Initialization with LLM configuration
MemoryFactory.init(
config=MemoryConfig(
provider="aworld",
llm_config=MemoryLLMConfig(
provider="openai",
model_name=os.environ["LLM_MODEL_NAME"],
api_key=os.environ["LLM_API_KEY"],
base_url=os.environ["LLM_BASE_URL"]
)
)
)
memory = MemoryFactory.instance()
```
`MemoryConfig` allows you to integrate different embedding models and vector databases.
```python
import os
from aworld.core.memory import MemoryConfig, MemoryLLMConfig, EmbeddingsConfig, VectorDBConfig
MemoryFactory.init(
config=MemoryConfig(
provider="aworld",
llm_config=MemoryLLMConfig(
provider="openai",
model_name=os.environ["LLM_MODEL_NAME"],
api_key=os.environ["LLM_API_KEY"],
base_url=os.environ["LLM_BASE_URL"]
),
embedding_config=EmbeddingsConfig(
provider="ollama", # or huggingface, openai, etc.
base_url="http://localhost:11434",
model_name="nomic-embed-text"
),
vector_store_config=VectorDBConfig(
provider="chroma",
config={
"chroma_data_path": "./chroma_db",
"collection_name": "aworld",
}
)
)
)
```
### Mutil-Agent Systems
We present a classic topology: `Leader-Executor`.
```python
"""
Leader-Executor topology:
┌───── plan ───┐
exec1 exec2
Each agent communicates with a single supervisor agent,
well recognized as Leader-Executor topology,
also referred to as a team topology in Aworld.
We can use this topology to implement paradigms of ReAct and Plan-Execute.
"""
from aworld.agents.llm_agent import Agent
from aworld.core.agent.swarm import Swarm, GraphBuildType
plan = Agent(name="plan", conf=agent_conf)
exec1 = Agent(name="exec1", conf=agent_conf)
exec2 = Agent(name="exec2", conf=agent_conf)
swarm = Swarm(topology=[(plan, exec1), (plan, exec2)], build_type=GraphBuildType.TEAM)
```
</details>
# 🏗️ Architecture Design Principles
<!-- AWorld is a versatile multi-agent framework designed to facilitate collaborative interactions and self-improvement among agents. -->
AWorld provides a comprehensive environment that supports a diverse array of applications, such as `Product Prototype Verification`, `Foundational Model Training`, and the design of `Multi-Agent Systems (MAS)` through meta-learning.
This framework is engineered to be highly adaptable, enabling researchers and developers to explore and innovate across multiple domains, thereby advancing the capabilities and applications of multi-agent systems.
## Concepts & Framework
| Concepts | Description |
| :-------------------------------------- | ------------ |
| [`agent`](./aworld/core/agent/base.py) | Define the foundational classes, descriptions, output parsing, and multi-agent collaboration (swarm) logic for defining, managing, and orchestrating agents in the AWorld system. |
| [`runner`](./aworld/runners) | Contains runner classes that manage the execution loop for agents in environments, handling episode rollouts and parallel training/evaluation workflows. |
| [`task`](./aworld/core/task.py) | Define the base Task class that encapsulates environment objectives, necessary tools, and termination conditions for agent interactions. |
| [`swarm`](./aworld/core/agent/swarm.py) | Implement the SwarmAgent class managing multi-agent coordination and emergent group behaviors through decentralized policies. |
| [`sandbox`](./aworld/sandbox) | Provide a controlled runtime with configurable scenarios for rapid prototyping and validation of agent behaviors. |
| [`tools`](./aworld/tools) | Offer a flexible framework for defining, adapting, and executing tools for agent-environment interaction in the AWorld system. |
| [`context`](./aworld/core/context) | Feature a comprehensive context management system for AWorld agents, enabling complete state tracking, configuration management, prompt optimization, multi-task state handling, and dynamic prompt templating throughout the agent lifecycle. |
| [`memory`](./aworld/memory) | Implement an extensible memory system for agents, supporting short-term and long-term memory, summarization, retrieval, embeddings, and integration.|
| [`trace`](./aworld/trace) | Feature an observable tracing framework for AWorld, enabling distributed tracing, context propagation, span management, and integration with popular frameworks and protocols to monitor and analyze agent, tool, and task execution.|
> 💡 Check the [examples](./examples/) directory to explore diverse AWorld applications.
## Characteristics
<!--
| 1. Agent Construction | 2. Topology Orchestration | 3. Environment |
|:---------------------|:-------------------------|:----------------|
| ✅ Various model providers<br> ✅ Integrated MCP services <br> ✅ Convient customizations | ✅ Encapsulated agent runtime <br> ✅ Flexible MAS patterns | ✅ Runtime state management <br> ✅ Clear state tracing <br> ✅ Distributed & high-concurrency environments for training |
| Agent Construction | Topology Orchestration | Environment |
|:---------------------------|:-----------------------------|:--------------------------------|
| ✅ Multi-model providers | ✅ Encapsulated runtime | ✅ Runtime state management |
| ✅ Integrated MCP services | ✅ Flexible MAS patterns | ✅ Clear state tracing |
| ✅ Customization options | | ✅ Distributed training |
| | | ✅ High-concurrency support |
-->
| Agent Construction | Topology Orchestration | Environment |
|:---------------------------|:----------------------------|:-------------------------------|
| ✅ Integrated MCP services | ✅ Encapsulated runtime | ✅ Runtime state management |
| ✅ Multi-model providers | ✅ Flexible MAS patterns | ✅ High-concurrency support |
| ✅ Customization options | ✅ Clear state tracing | ✅ Distributed training |
## Forward Process Design
![](readme_assets/runtime.jpg)
Here is a forward illustration to collect BFCL forward trajectories: [`tutorial`](./examples/BFCL/README.md).
## Backward Process Design
> During training, an action-state rollout demonstration using **AWorld's distributed environments**.
![](readme_assets/agent_training2.jpg)
Here is an illustration of AWorld-training with various frameworks, like AReal, Verl and Swift. [`tutorial`](./train/README.md).
# 🧩 Technical Report
This section showcases novel research papers developed using AWorld, demonstrating its capacity to incubate cutting-edge multi-agent systems that advance toward Artificial General Intelligence (AGI).
#### Multi-Agent-System (MAS) Meta Learning
1. **Profile-Aware Maneuvering: A Dynamic Multi-Agent System for Robust GAIA Problem Solving by AWorld.** arxiv, 2025. [paper](https://arxiv.org/abs/2508.09889), [code](https://github.com/inclusionAI/AWorld/blob/main/examples/gaia/README_GUARD.md)
*Zhitian Xie, Qintong Wu, Chengyue Yu, Chenyi Zhuang, Jinjie Gu*
#### Model Training
1. **AWorld: Orchestrating the Training Recipe for Agentic AI.** arxiv, 2025. [paper](https://arxiv.org/abs/2508.20404), [code](https://github.com/inclusionAI/AWorld/tree/main/train), [model](https://huggingface.co/inclusionAI/Qwen3-32B-AWorld)
*Chengyue Yu, Siyuan Lu, Chenyi Zhuang, Dong Wang, Qintong Wu, etc.*
2. **FunReason: Enhancing Large Language Models' Function Calling via Self-Refinement Multiscale Loss and Automated Data Refinement.** arxiv, 2025. [paper](https://arxiv.org/abs/2505.20192), [model](https://huggingface.co/Bingguang/FunReason)
*Bingguang Hao, Maolin Wang, Zengzhuang Xu, Cunyin Peng, etc.*
3. **Exploring Superior Function Calls via Reinforcement Learning.** arxiv, 2025. [paper](https://arxiv.org/abs/2508.05118), [code](https://github.com/BingguangHao/RLFC)
*Bingguang Hao, Maolin Wang, Zengzhuang Xu, Yicheng Chen, etc.*
4. **RAG-R1 : Incentivize the Search and Reasoning Capabilities of LLMs through Multi-query Parallelism.** arxiv, 2025. [paper](https://arxiv.org/abs/2507.02962), [code](https://github.com/inclusionAI/AgenticLearning), [model](https://huggingface.co/collections/endertzw/rag-r1-68481d7694b3fca8b809aa29)
*Zhiwen Tan, Jiaming Huang, Qintong Wu, Hongxuan Zhang, Chenyi Zhuang, Jinjie Gu*
5. **V2P: From Background Suppression to Center Peaking for Robust GUI Grounding Task.** arxiv, 2025. [paper](https://arxiv.org/abs/2508.13634), [code](https://github.com/inclusionAI/AgenticLearning/tree/main/V2P)
*Jikai Chen, Long Chen, Dong Wang, Leilei Gan, Chenyi Zhuang, Jinjie Gu*
# Contributing
We warmly welcome developers to join us in building and improving AWorld! Whether you're interested in enhancing the framework, fixing bugs, or adding new features, your contributions are valuable to us.
For academic citations or wish to contact us, please use the following BibTeX entry:
```bibtex
@misc{yu2025aworldorchestratingtrainingrecipe,
title={AWorld: Orchestrating the Training Recipe for Agentic AI},
author={Chengyue Yu and Siyuan Lu and Chenyi Zhuang and Dong Wang and Qintong Wu and Zongyue Li and Runsheng Gan and Chunfeng Wang and Siqi Hou and Gaochi Huang and Wenlong Yan and Lifeng Hong and Aohui Xue and Yanfeng Wang and Jinjie Gu and David Tsai and Tao Lin},
year={2025},
eprint={2508.20404},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2508.20404},
}
```
# Star History
![](https://api.star-history.com/svg?repos=inclusionAI/AWorld&type=Date)
# Appendix: Web Client Usage
![GAIA Agent Runtime Demo](readme_assets/gaia_demo.gif)
Your project structure should look like this:
```text
agent-project-root-dir/
agent_deploy/
my_first_agent/
__init__.py
agent.py
```
Create project folders.
```shell
mkdir my-aworld-project && cd my-aworld-project # project-root-dir
mkdir -p agent_deploy/my_first_agent
```
#### Step 1: Define Your Agent
Create your first agnet in `agent_deploy/my_first_agent`:
`__init__.py`: Create empty `__ini__.py` file.
```shell
cd agent_deploy/my_first_agent
touch __init__.py
```
`agent.py`: Define your agent logic:
```python
import logging
import os
from aworld.cmd.data_model import BaseAWorldAgent, ChatCompletionRequest
from aworld.config.conf import AgentConfig, TaskConfig
from aworld.agents.llm_agent import Agent
from aworld.core.task import Task
from aworld.runner import Runners
logger = logging.getLogger(__name__)
class AWorldAgent(BaseAWorldAgent):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def name(self):
return "My First Agent"
def description(self):
return "A helpful assistant that can answer questions and help with tasks"
async def run(self, prompt: str = None, request: ChatCompletionRequest = None):
# Load LLM configuration from environment variables
agent_config = AgentConfig(
llm_provider=os.getenv("LLM_PROVIDER", "openai"),
llm_model_name=os.getenv("LLM_MODEL_NAME", "gpt-4"),
llm_api_key=os.getenv("LLM_API_KEY"),
llm_base_url=os.getenv("LLM_BASE_URL"),
llm_temperature=float(os.getenv("LLM_TEMPERATURE", "0.7"))
)
# Validate required configuration
if not agent_config.llm_model_name or not agent_config.llm_api_key:
raise ValueError("LLM_MODEL_NAME and LLM_API_KEY must be set!")
# Optional: Configure MCP tools for enhanced capabilities
mcp_config = {
"mcpServers": {
"amap-mcp": {
"type": "sse",
"url": "https://mcp.example.com/sse?key=YOUR_API_KEY", # Replace Your API Key
"timeout": 30,
"sse_read_timeout": 300
}
}
}
# Create the agent instance
agent = Agent(
conf=agent_config,
name="My First Agent",
system_prompt="""You are a helpful AI assistant. Your goal is to:
- Answer questions accurately and helpfully
- Provide clear, step-by-step guidance when needed
- Be friendly and professional in your responses""",
mcp_servers=["amap-mcp"],
mcp_config=mcp_config
)
# Extract user input
user_input = prompt or (request.messages[-1].content if request else "")
# Create and execute task
task = Task(
input=user_input,
agent=agent,
conf=TaskConfig(max_steps=5),
session_id=getattr(request, 'session_id', None)
)
# Stream the agent's response
async for output in Runners.streamed_run_task(task).stream_events():
yield output
```
#### Step 2: Run Agent
Setup environment variables:
```shell
# Navigate back to project root
cd ${agent-project-root-dir}
# Set your LLM credentials
export LLM_MODEL_NAME="gpt-4"
export LLM_API_KEY="your-api-key-here"
export LLM_BASE_URL="https://api.openai.com/v1" # Optional for OpenAI
```
Launch Your Agent:
```shell
# Option 1: Launch with Web UI
aworld web
# Then open http://localhost:8000 in your browser
# Option 2: Launch REST API (For integrations)
aworld api
# Then visit http://localhost:8000/docs for API documentation
```
Success! Your agent is now running and ready to chat!
---
<!-- resource section start -->
<!-- image links -->
[arxiv-image]: https://img.shields.io/badge/Paper-arXiv-B31B1B?style=for-the-badge&logo=arxiv&logoColor=white
[blog-image]: https://img.shields.io/badge/Blog-Coming%20Soon-FF5722?style=for-the-badge&logo=blogger&logoColor=white
[deepwiki-image]: https://img.shields.io/badge/DeepWiki-Explore-blueviolet?style=for-the-badge&logo=wikipedia&logoColor=white
[discord-image]: https://img.shields.io/badge/Discord-Join%20us-blue?style=for-the-badge&logo=discord&logoColor=white
[github-code-image]: https://img.shields.io/badge/Code-GitHub-181717?style=for-the-badge&logo=github&logoColor=white
[huggingface-dataset-image]: https://img.shields.io/badge/Dataset-Coming%20Soon-007ACC?style=for-the-badge&logo=dataset&logoColor=white
[huggingface-model-image]: https://img.shields.io/badge/Model-Hugging%20Face-FF6B6B?style=for-the-badge&logo=huggingface&logoColor=white
[license-image]: https://img.shields.io/badge/License-MIT-yellow?style=for-the-badge
[twitter-image]: https://img.shields.io/badge/Twitter-Follow%20us-1DA1F2?style=for-the-badge&logo=twitter&logoColor=white
[wechat-image]: https://img.shields.io/badge/WeChat-Add%20us-green?style=for-the-badge&logo=wechat&logoColor=white
[tutorial-image]: https://img.shields.io/badge/Tutorial-Get%20Started-FF6B35?style=for-the-badge&logo=book&logoColor=white
<!-- aworld links -->
[deepwiki-url]: https://deepwiki.com/inclusionAI/AWorld
[discord-url]: https://discord.gg/b4Asj2ynMw
[license-url]: https://opensource.org/licenses/MIT
[twitter-url]: https://x.com/InclusionAI666
[wechat-url]: https://raw.githubusercontent.com/inclusionAI/AWorld/main/readme_assets/aworld_wechat.png
[arxiv-url]: https://arxiv.org/abs/2508.
[tutorial-url]: https://inclusionai.github.io/AWorld/
<!-- funreason links -->
[funreason-code-url]: https://github.com/BingguangHao/FunReason
[funreason-model-url]: https://huggingface.co/Bingguang/FunReason
[funreason-paper-url]: https://arxiv.org/pdf/2505.20192
<!-- [funreason-dataset-url]: https://github.com/BingguangHao/FunReason -->
<!-- [funreason-blog-url]: https://github.com/BingguangHao/FunReason -->
<!-- deepsearch links -->
[deepsearch-code-url]: https://github.com/inclusionAI/AgenticLearning
[deepsearch-dataset-url]: https://github.com/inclusionAI/AgenticLearning
[deepsearch-model-url]: https://huggingface.co/collections/endertzw/rag-r1-68481d7694b3fca8b809aa29
[deepsearch-paper-url]: https://arxiv.org/abs/2507.02962
<!-- badge -->
[MAS]: https://img.shields.io/badge/Mutli--Agent-System-EEE1CE
[IMO]: https://img.shields.io/badge/IMO-299D8F
[BFCL]: https://img.shields.io/badge/BFCL-8AB07D
[GAIA]: https://img.shields.io/badge/GAIA-E66F51
[Runtime]: https://img.shields.io/badge/AWorld-Runtime-287271
[Leaderboard]: https://img.shields.io/badge/Leaderboard-FFE6B7
[Benchmark]: https://img.shields.io/badge/Benchmark-FFE6B7
[Cloud-Native]: https://img.shields.io/badge/Cloud--Native-B19CD7
[Forward]: https://img.shields.io/badge/Forward-4A90E2
[Backward]: https://img.shields.io/badge/Backward-7B68EE
[Code]: https://img.shields.io/badge/Code-FF6B6B
[Paper]: https://img.shields.io/badge/Paper-4ECDC4
<!-- resource section end -->
chapter9/gaia-experience/AWorld/README_zh.md
+621
View File
@@ -0,0 +1,621 @@
from examples.for_test import topology<div align="center">
# AWorld: 为智能体自我演进提供多样化的运行环境
</div>
<h4 align="center">
*"自我意识：最难的问题不是解决限制，而是发现自己的局限性"*
[![Twitter Follow][twitter-image]][twitter-url]
[![WeChat QR Code][wechat-image]][wechat-url]
[![Discord][discord-image]][discord-url]
[![License: MIT][license-image]][license-url]
[![DeepWiki][deepwiki-image]][deepwiki-url]
[![arXiv][arxiv-image]][arxiv-url]
[![Tutorial][tutorial-image]][tutorial-url]
<!-- [![arXiv][arxiv-image]][arxiv-url] -->
</h4>
<h4 align="center">
[English](./README.md) |
[快速开始](#快速开始) |
[架构设计](#架构设计原则) |
[应用场景](#应用场景) |
[贡献指南](#贡献指南) |
[附录](#附录)
</h4>
![](./readme_assets/heading_banner.png)
**AWorld (Agent World)** 是一个为大规模智能体自我改进而设计的下一代框架。通过上述功能，我们使AI智能体能够通过从各种环境中的知识和经验中学习来持续进化。使用AWorld，您可以：
1. **构建工作流**：设计和实现自动化任务序列 [文档](https://inclusionai.github.io/AWorld/Quickstart/workflow_construction/)
2. **构建智能体**：创建具有MCP工具的智能AI智能体 [文档](https://inclusionai.github.io/AWorld/Quickstart/agent_construction/)
3. **构建多智能体系统(MAS)**：编排协作智能体生态系统 [文档](https://inclusionai.github.io/AWorld/Quickstart/multi-agent_system_construction/)
4. **高效训练智能体**：让MAS在各种环境中自我演进和优化
---
**群体智能** 🚀
展示群体智能在不同领域的SOTA效果。欢迎加入我们正在进行中的项目！
| **类别** | **成就** | **性能表现** | **核心创新** | **日期** |
|:-------------|:----------------|:----------------|:-------------------|:----------|
| **🤖 智能体** | **GAIA基准测试卓越表现** [![][GAIA]](https://huggingface.co/spaces/gaia-benchmark/leaderboard) | Pass@1: **67.89**, Pass@3: **83.49** (109个任务) [![][Code]](./examples/gaia/README_GUARD.md) | 多智能体系统稳定性与编排 [![][Paper]](https://arxiv.org/abs/2508.09889) | 2025/08/06 |
| **🧠 推理能力** | **IMO 2025问题求解** [![][IMO]](https://www.imo-official.org/year_info.aspx?year=2025) | 6小时内解决5/6道题 [![][Code]](examples/imo/README.md) | 多智能体协作超越单一模型 | 2025/07/25 |
<details>
<summary style="font-size: 1.2em;font-weight: bold;"> 🌏 查看进行中的项目 </summary>
| **类别** | **成就** | **状态** | **预期影响** |
|:-------------|:----------------|:-----------|:-------------------|
| **🖼️ 多模态** | 领先的操作系统/网页交互 | 进行中 | 视觉推理与环境理解 |
| **💻 编程** | 领先的安装、编码、测试、调试等能力 | 进行中 | 自动化软件工程能力 |
| **🔧 工具使用** | 领先的多轮函数调用 | 即将推出 | 影响现实世界 |
</details>
---
**自我改进, 超越群体智能** 🌱
`智能体`可以在各种`环境`中运行，收集正面和负面的`经验`，并通过`训练`进行学习。
<table style="width: 100%; border-collapse: collapse; table-layout: fixed;">
<thead>
<tr>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">智能体</th>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">环境</th>
<th style="width: 20%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">经验</th>
<th style="width: 25%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">训练</th>
<th style="width: 15%; text-align: left; border-bottom: 2px solid #ddd; padding: 8px;">代码</th>
</tr>
</thead>
<tbody>
<tr>
<td style="padding: 8px; vertical-align: top;">GAIA 智能体</td>
<td style="padding: 8px; vertical-align: top;">
终端、代码、搜索、playwright 和 4 个额外工具
</td>
<td style="padding: 8px; vertical-align: top;">
从 GAIA 验证数据集的 165 个样本中收集 <br>
<a href="https://huggingface.co/datasets/gaia-benchmark/GAIA/tree/main/2023/validation" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Dataset-Training-8AB07D" alt="训练数据集">
</a>
</td>
<td style="padding: 8px; vertical-align: top;">
通过 GRPO 进行 rollout、奖励计算和梯度更新
</td>
<td style="padding: 8px; vertical-align: top;">
3行代码即可
<br>
<a href="./train/README_zh.md" target="_blank" style="text-decoration: none;">
<img src="https://img.shields.io/badge/Code-README-green" alt="代码">
</a>
</td>
</tr>
</tbody>
</table>
---
# 快速开始
## 前置要求
> [!TIP]
> Python>=3.11
```bash
git clone https://github.com/inclusionAI/AWorld && cd AWorld
pip install .
```
## Hello world 示例
我们引入 `Agent` 和 `Runners` 的概念来帮助您快速上手。
关于并行任务执行，请参考[并行运行示例](examples/parallel_run/README.md)。
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
summarizer = Agent(
name="Summary Agent",
system_prompt="You specialize at summarizing.",
)
result = Runners.sync_run(
input="Tell me a succint history about the universe",
agent=summarizer,
)
```
同时，我们引入 `Swarm` 的概念来构建智能体团队。
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
from aworld.core.agent.swarm import Swarm
researcher = Agent(
name="Research Agent",
system_prompt="You specialize at researching.",
)
summarizer = Agent(
name="Summary Agent",
system_prompt="You specialize at summarizing.",
)
# 创建具有协作工作流的智能体组 (multi-agent)
group = Swarm(topology=[(researcher, summarizer)])
result = Runners.sync_run(
input="Tell me a complete history about the universe",
swarm=group,
)
```
最后，运行您自己的智能体或团队
```bash
# 设置LLM凭据
export LLM_MODEL_NAME="gpt-4"
export LLM_API_KEY="your-api-key-here"
export LLM_BASE_URL="https://api.openai.com/v1"
# 运行
python /path/to/agents/or/teams
```
<details>
<summary style="font-size: 1.2em;font-weight: bold;"> 🌏 点击查看高级用法 </summary>
### 显式传递AgentConfig
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
from aworld.config.conf import AgentConfig
from aworld.core.agent.swarm import Swarm
gpt_conf = AgentConfig(
llm_provider="openai",
llm_model_name="gpt-4o",
llm_api_key="<OPENAI_API_KEY>",
llm_temperature=0.1,
)
openrouter_conf = AgentConfig(
llm_provider="openai",
llm_model_name="google/gemini-2.5-pro",
llm_api_key="<OPENROUTER_API_KEY>",
llm_base_url="https://openrouter.ai/api/v1"
llm_temperature=0.1,
)
researcher = Agent(
name="Research Agent",
conf=gpt_conf,
system_prompt="You specialize at researching.",
)
summarizer = Agent(
name="Summary Agent",
conf=openrouter_conf,
system_prompt="You specialize at summarizing.",
)
# 创建具有协作工作流的智能体组 (multi-agent)
group = Swarm(topology=[(researcher, summarizer)])
result = Runners.sync_run(
input="Tell me a complete history about the universe",
swarm=group,
)
```
### 配备MCP工具的智能体
```python
from aworld.agents.llm_agent import Agent
from aworld.runner import Runners
mcp_config = {
"mcpServers": {
"GorillaFileSystem": {
"type": "stdio",
"command": "python",
"args": ["examples/BFCL/mcp_tools/gorilla_file_system.py"],
},
}
}
file_sys = Agent(
name="file_sys_agent",
system_prompt=(
"You are a helpful agent to use "
"the standard file system to perform file operations."
),
mcp_servers=mcp_config.get("mcpServers", []).keys(),
mcp_config=mcp_config,
)
result = Runners.sync_run(
input=(
"use mcp tools in the GorillaFileSystem server "
"to perform file operations: "
"write the content 'AWorld' into "
"the hello_world.py file with a new line "
"and keep the original content of the file. "
"Make sure the new and old "
"content are all in the file; "
"and display the content of the file"
),
agent=file_sys,
)
```
### 集成记忆的智能体
建议使用 `MemoryFactory` 来初始化和访问Memory实例。
```python
from aworld.memory.main import MemoryFactory
from aworld.core.memory import MemoryConfig, MemoryLLMConfig
# 简单初始化
memory = MemoryFactory.instance()
# 使用LLM配置进行初始化
MemoryFactory.init(
config=MemoryConfig(
provider="aworld",
llm_config=MemoryLLMConfig(
provider="openai",
model_name=os.environ["LLM_MODEL_NAME"],
api_key=os.environ["LLM_API_KEY"],
base_url=os.environ["LLM_BASE_URL"]
)
)
)
memory = MemoryFactory.instance()
```
`MemoryConfig` 允许您集成不同的嵌入模型和向量数据库。
```python
import os
from aworld.core.memory import MemoryConfig, MemoryLLMConfig, EmbeddingsConfig, VectorDBConfig
MemoryFactory.init(
config=MemoryConfig(
provider="aworld",
llm_config=MemoryLLMConfig(
provider="openai",
model_name=os.environ["LLM_MODEL_NAME"],
api_key=os.environ["LLM_API_KEY"],
base_url=os.environ["LLM_BASE_URL"]
),
embedding_config=EmbeddingsConfig(
provider="ollama", # 或 huggingface, openai 等
base_url="http://localhost:11434",
model_name="nomic-embed-text"
),
vector_store_config=VectorDBConfig(
provider="chroma",
config={
"chroma_data_path": "./chroma_db",
"collection_name": "aworld",
}
)
)
)
```
### 多智能体系统
我们展示一个经典拓扑：`Leader-Executor`。
```python
"""
Leader-Executor 拓扑：
┌───── plan ───┐
exec1 exec2
每个智能体与单个监督智能体通信，
被公认为Leader-Executor拓扑，
在Aworld中也称为团队(Team)拓扑。
我们可以使用该拓扑实现ReAct和Plan-Execute范式。
"""
from aworld.agents.llm_agent import Agent
from aworld.core.agent.swarm import Swarm, GraphBuildType
plan = Agent(name="plan", conf=agent_conf)
exec1 = Agent(name="exec1", conf=agent_conf)
exec2 = Agent(name="exec2", conf=agent_conf)
swarm = Swarm(topology=[(plan, exec1), (plan, exec2)], build_type=GraphBuildType.TEAM)
```
</details>
# 架构设计原则
<!-- AWorld is a versatile multi-agent framework designed to facilitate collaborative interactions and self-improvement among agents. -->
AWorld 提供了一个全面的环境，支持多样化的应用，如 `产品原型验证`、`基础模型训练`，以及通过元学习设计 `多智能体系统 (MAS)`。
该框架被设计为高度适应性，使研究人员和开发人员能够在多个领域探索和创新，从而推进多智能体系统的能力和应用。
## 概念与框架
| 概念 | 描述 |
| :-------------------------------------- | ------------ |
| [`agent`](./aworld/core/agent/base.py) | 定义基础类、描述、输出解析和多智能体协作（群体）逻辑，用于在AWorld系统中定义、管理和编排智能体。 |
| [`runner`](./aworld/runners) | 包含管理环境中智能体执行循环的运行器类，处理情节回放和并行训练/评估工作流。 |
| [`task`](./aworld/core/task.py) | 定义封装环境目标、必要工具和智能体交互终止条件的基础Task类。 |
| [`swarm`](./aworld/core/agent/swarm.py) | 实现管理多智能体协调和通过去中心化策略出现的群体行为的SwarmAgent类。 |
| [`sandbox`](./aworld/sandbox) | 提供具有可配置场景的受控运行时，用于智能体行为的快速原型设计和验证。 |
| [`tools`](./aworld/tools) | 为AWorld系统中智能体-环境交互的工具定义、适配和执行提供灵活框架。 |
| [`context`](./aworld/core/context) | 为AWorld智能体提供全面的上下文管理系统，支持完整的状态跟踪、配置管理、提示优化、多任务状态处理和整个智能体生命周期中的动态提示模板。 |
| [`memory`](./aworld/memory) | 为智能体实现可扩展的记忆系统，支持短期和长期记忆、总结、检索、嵌入和集成。|
| [`trace`](./aworld/trace) | 为AWorld提供可观察的跟踪框架，支持分布式跟踪、上下文传播、跨度管理，以及与流行框架和协议的集成，以监控和分析智能体、工具和任务执行。|
> 💡 查看 [examples](./examples/) 目录以探索多样化的AWorld应用。
## 特性
| 智能体构建 | 拓扑编排 | 环境 |
|:---------------------------|:----------------------------|:-------------------------------|
| ✅ 集成MCP服务 | ✅ 封装的运行时 | ✅ 运行时状态管理 |
| ✅ 多模型提供商 | ✅ 灵活的MAS模式 | ✅ 高并发支持 |
| ✅ 自定义选项 | ✅ 清晰的状态跟踪 | ✅ 分布式训练 |
## 正向过程
![](readme_assets/runtime.jpg)
这里是收集BFCL正向轨迹的正向说明：[`教程`](./examples/BFCL/README.md)。
## 反向过程
> 训练期间，使用 **AWorld的分布式环境** 进行动作-状态回放演示。
![](readme_assets/agent_training2.jpg)
这边有使用AWorld结合各种框架（如 AReal、Verl 和 Swift）进行训练的说明。[`教程`](./train/README.md)。
# 🧩 技术报告
本节展示了使用 AWorld 研发的研究论文，展示了其孵化前沿多智能体系统的能力，这些系统推动着向通用人工智能（AGI）的发展。
#### 多智能体系统（MAS）元学习
1. **Profile-Aware Maneuvering: A Dynamic Multi-Agent System for Robust GAIA Problem Solving by AWorld.** arxiv, 2025. [论文](https://arxiv.org/abs/2508.09889), [代码](https://github.com/inclusionAI/AWorld/blob/main/examples/gaia/README_GUARD.md)
*Zhitian Xie, Qintong Wu, Chengyue Yu, Chenyi Zhuang, Jinjie Gu*
#### 模型训练
1. **AWorld: Orchestrating the Training Recipe for Agentic AI.** arxiv, 2025. [论文](https://arxiv.org/abs/2508.20404), [代码](https://github.com/inclusionAI/AWorld/tree/main/train), [模型](https://huggingface.co/inclusionAI/Qwen3-32B-AWorld)
*Chengyue Yu, Siyuan Lu, Chenyi Zhuang, Dong Wang, Qintong Wu, etc.*
2. **FunReason: Enhancing Large Language Models' Function Calling via Self-Refinement Multiscale Loss and Automated Data Refinement.** arxiv, 2025. [论文](https://arxiv.org/abs/2505.20192), [模型](https://huggingface.co/Bingguang/FunReason)
*Bingguang Hao, Maolin Wang, Zengzhuang Xu, Cunyin Peng, etc.*
3. **Exploring Superior Function Calls via Reinforcement Learning.** arxiv, 2025. [论文](https://arxiv.org/abs/2508.05118), [代码](https://github.com/BingguangHao/RLFC)
*Bingguang Hao, Maolin Wang, Zengzhuang Xu, Yicheng Chen, etc.*
4. **RAG-R1 : Incentivize the Search and Reasoning Capabilities of LLMs through Multi-query Parallelism.** arxiv, 2025. [论文](https://arxiv.org/abs/2507.02962), [代码](https://github.com/inclusionAI/AgenticLearning), [模型](https://huggingface.co/collections/endertzw/rag-r1-68481d7694b3fca8b809aa29)
*Zhiwen Tan, Jiaming Huang, Qintong Wu, Hongxuan Zhang, Chenyi Zhuang, Jinjie Gu*
5. **V2P: From Background Suppression to Center Peaking for Robust GUI Grounding Task.** arxiv, 2025. [论文](https://arxiv.org/abs/2508.13634), [代码](https://github.com/inclusionAI/AgenticLearning/tree/main/V2P)
*Jikai Chen, Long Chen, Dong Wang, Leilei Gan, Chenyi Zhuang, Jinjie Gu*
# 贡献指南
我们热烈欢迎开发者加入我们构建和改进AWorld！无论您对增强框架、修复错误还是添加新功能感兴趣，您的贡献对我们都很宝贵。
对于学术引用或希望联系我们，请使用以下BibTeX条目：
```bibtex
@misc{yu2025aworldorchestratingtrainingrecipe,
title={AWorld: Orchestrating the Training Recipe for Agentic AI},
author={Chengyue Yu and Siyuan Lu and Chenyi Zhuang and Dong Wang and Qintong Wu and Zongyue Li and Runsheng Gan and Chunfeng Wang and Siqi Hou and Gaochi Huang and Wenlong Yan and Lifeng Hong and Aohui Xue and Yanfeng Wang and Jinjie Gu and David Tsai and Tao Lin},
year={2025},
eprint={2508.20404},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2508.20404},
}
```
# Star历史
![](https://api.star-history.com/svg?repos=inclusionAI/AWorld&type=Date)
# 附录
Web客户端使用
![GAIA智能体运行时演示](readme_assets/gaia_demo.gif)
您的项目结构应该如下所示：
```text
agent-project-root-dir/
agent_deploy/
my_first_agent/
__init__.py
agent.py
```
创建项目文件夹。
```shell
mkdir my-aworld-project && cd my-aworld-project # project-root-dir
mkdir -p agent_deploy/my_first_agent
```
#### 步骤1：定义您的智能体
在 `agent_deploy/my_first_agent` 中创建您的第一个智能体：
`__init__.py`：创建空的 `__init__.py` 文件。
```shell
cd agent_deploy/my_first_agent
touch __init__.py
```
`agent.py`：定义您的智能体逻辑：
```python
import logging
import os
from aworld.cmd.data_model import BaseAWorldAgent, ChatCompletionRequest
from aworld.config.conf import AgentConfig, TaskConfig
from aworld.agents.llm_agent import Agent
from aworld.core.task import Task
from aworld.runner import Runners
logger = logging.getLogger(__name__)
class AWorldAgent(BaseAWorldAgent):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def name(self):
return "My First Agent"
def description(self):
return "A helpful assistant that can answer questions and help with tasks"
async def run(self, prompt: str = None, request: ChatCompletionRequest = None):
# Load LLM configuration from environment variables
agent_config = AgentConfig(
llm_provider=os.getenv("LLM_PROVIDER", "openai"),
llm_model_name=os.getenv("LLM_MODEL_NAME", "gpt-4"),
llm_api_key=os.getenv("LLM_API_KEY"),
llm_base_url=os.getenv("LLM_BASE_URL"),
llm_temperature=float(os.getenv("LLM_TEMPERATURE", "0.7"))
)
# Validate required configuration
if not agent_config.llm_model_name or not agent_config.llm_api_key:
raise ValueError("LLM_MODEL_NAME and LLM_API_KEY must be set!")
# Optional: Configure MCP tools for enhanced capabilities
mcp_config = {
"mcpServers": {
"amap-mcp": {
"type": "sse",
"url": "https://mcp.example.com/sse?key=YOUR_API_KEY", # Replace Your API Key
"timeout": 30,
"sse_read_timeout": 300
}
}
}
# Create the agent instance
agent = Agent(
conf=agent_config,
name="My First Agent",
system_prompt="""You are a helpful AI assistant. Your goal is to:
- Answer questions accurately and helpfully
- Provide clear, step-by-step guidance when needed
- Be friendly and professional in your responses""",
mcp_servers=["amap-mcp"],
mcp_config=mcp_config
)
# Extract user input
user_input = prompt or (request.messages[-1].content if request else "")
# Create and execute task
task = Task(
input=user_input,
agent=agent,
conf=TaskConfig(max_steps=5),
session_id=getattr(request, 'session_id', None)
)
# Stream the agent's response
async for output in Runners.streamed_run_task(task).stream_events():
yield output
```
#### 步骤2：运行智能体
设置环境变量：
```shell
# 导航回项目根目录
cd ${agent-project-root-dir}
# 设置您的LLM凭据
export LLM_MODEL_NAME="gpt-4"
export LLM_API_KEY="your-api-key-here"
export LLM_BASE_URL="https://api.openai.com/v1" # Optional for OpenAI
```
启动您的智能体：
```shell
# 选项1：使用Web UI启动
aworld web
# 然后在浏览器中打开 http://localhost:8000
# 选项2：启动REST API（用于集成）
aworld api
# 然后访问 http://localhost:8000/docs 查看API文档
```
成功！您的智能体现在正在运行并准备聊天！
---
<!-- resource section start -->
<!-- image links -->
[arxiv-image]: https://img.shields.io/badge/Paper-arXiv-B31B1B?style=for-the-badge&logo=arxiv&logoColor=white
[blog-image]: https://img.shields.io/badge/Blog-Coming%20Soon-FF5722?style=for-the-badge&logo=blogger&logoColor=white
[deepwiki-image]: https://img.shields.io/badge/DeepWiki-Explore-blueviolet?style=for-the-badge&logo=wikipedia&logoColor=white
[discord-image]: https://img.shields.io/badge/Discord-Join%20us-blue?style=for-the-badge&logo=discord&logoColor=white
[github-code-image]: https://img.shields.io/badge/Code-GitHub-181717?style=for-the-badge&logo=github&logoColor=white
[huggingface-dataset-image]: https://img.shields.io/badge/Dataset-Coming%20Soon-007ACC?style=for-the-badge&logo=dataset&logoColor=white
[huggingface-model-image]: https://img.shields.io/badge/Model-Hugging%20Face-FF6B6B?style=for-the-badge&logo=huggingface&logoColor=white
[license-image]: https://img.shields.io/badge/License-MIT-yellow?style=for-the-badge
[twitter-image]: https://img.shields.io/badge/Twitter-Follow%20us-1DA1F2?style=for-the-badge&logo=twitter&logoColor=white
[wechat-image]: https://img.shields.io/badge/WeChat-Add%20us-green?style=for-the-badge&logo=wechat&logoColor=white
[tutorial-image]: https://img.shields.io/badge/Tutorial-Get%20Started-FF6B35?style=for-the-badge&logo=book&logoColor=white
<!-- aworld links -->
[deepwiki-url]: https://deepwiki.com/inclusionAI/AWorld
[discord-url]: https://discord.gg/b4Asj2ynMw
[license-url]: https://opensource.org/licenses/MIT
[twitter-url]: https://x.com/InclusionAI666
[wechat-url]: https://raw.githubusercontent.com/inclusionAI/AWorld/main/readme_assets/aworld_wechat.png
[arxiv-url]: https://arxiv.org/abs/2508.20404
[tutorial-url]: https://inclusionai.github.io/AWorld/
<!-- funreason links -->
[funreason-code-url]: https://github.com/BingguangHao/FunReason
[funreason-model-url]: https://huggingface.co/Bingguang/FunReason
[funreason-paper-url]: https://arxiv.org/pdf/2505.20192
<!-- [funreason-dataset-url]: https://github.com/BingguangHao/FunReason -->
<!-- [funreason-blog-url]: https://github.com/BingguangHao/FunReason -->
<!-- deepsearch links -->
[deepsearch-code-url]: https://github.com/inclusionAI/AgenticLearning
[deepsearch-dataset-url]: https://github.com/inclusionAI/AgenticLearning
[deepsearch-model-url]: https://huggingface.co/collections/endertzw/rag-r1-68481d7694b3fca8b809aa29
[deepsearch-paper-url]: https://arxiv.org/abs/2507.02962
<!-- badge -->
[MAS]: https://img.shields.io/badge/Mutli--Agent-System-EEE1CE
[IMO]: https://img.shields.io/badge/IMO-299D8F
[BFCL]: https://img.shields.io/badge/BFCL-8AB07D
[GAIA]: https://img.shields.io/badge/GAIA-E66F51
[Runtime]: https://img.shields.io/badge/AWorld-Runtime-287271
[Leaderboard]: https://img.shields.io/badge/Leaderboard-FFE6B7
[Benchmark]: https://img.shields.io/badge/Benchmark-FFE6B7
[Cloud-Native]: https://img.shields.io/badge/Cloud--Native-B19CD7
[Code]: https://img.shields.io/badge/Code-FF6B6B
[Paper]: https://img.shields.io/badge/Paper-4ECDC4
<!-- resource section end -->
chapter9/gaia-experience/AWorld/aworld/__init__.py
+40
View File
@@ -0,0 +1,40 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
import atexit
import os
try:
from aworld.utils.import_package import import_package
import_package("dotenv", install_name="python-dotenv")
from dotenv import load_dotenv
sucess = load_dotenv()
if not sucess:
load_dotenv(os.path.join(os.getcwd(), ".env"))
except Exception as e:
print(e)
def cleanup():
import re
try:
value = os.environ.get("LOCAL_TOOLS_ENV_VAR", '')
if value:
for action_file in value.split(";"):
v = re.split(r"\w{6}__tmp", action_file)[0]
if v == action_file:
continue
tool_file = action_file.replace("_action.py", ".py")
try:
os.remove(action_file)
os.remove(tool_file)
except:
pass
except:
pass
os.environ["LOCAL_TOOLS_ENV_VAR"] = ''
atexit.register(cleanup, )
chapter9/gaia-experience/AWorld/aworld/__main__.py
+7
View File
@@ -0,0 +1,7 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
from aworld.cmd.cli import main
if __name__ == "__main__":
main()
chapter9/gaia-experience/AWorld/aworld/agents/README-multi-agent.md
+276
View File
@@ -0,0 +1,276 @@
# Multi-agent
```python
from aworld.agents.llm_agent import Agent
from aworld.config.conf import AgentConfig
from aworld.core.agent.swarm import Swarm, GraphBuildType
agent_conf = AgentConfig(...)
```
## Builder
Builder represents the way topology is constructed, which is related to runtime execution.
Topology is the definition of structure. For the same topology structure, different builders
will produce execution processes and different results.
```python
"""
Topology:
┌─────A─────┐
B | C
D
"""
A = Agent(name="A", conf=agent_conf)
B = Agent(name="B", conf=agent_conf)
C = Agent(name="C", conf=agent_conf)
D = Agent(name="D", conf=agent_conf)
```
### Workflow
Workflow is a special topological structure that can be executed deterministically, all nodes in the swarm
will be executed. And the starting and ending nodes are **unique** and **indispensable**.
Define:
```python
# default is workflow
Swarm((A, B), (A, C), (A, D))
or
Swarm(A, [B, C, D])
```
The example means A is the start node, and the merge of B, C, and D is the end node.
### Handoff
Handoff using pure AI to drive the flow of the entire topology diagram, one agent's decision hands off
control to another. Agents as tools, depending on the defined pairs of agents.
Define:
```python
Swarm((A, B), (A, C), (A, D), build_type=GraphBuildType.HANDOFF)
or
HandoffSwarm((A, B), (A, C), (A, D))
```
**NOTE**: Handoff supported tuple of paired agents forms only.
### Team
Team requires a leadership agent, and other agents follow its command.
Team is a special case of handoff, which is the leader-follower mode.
Define:
```python
Swarm((A, B), (A, C), (A, D), build_type=GraphBuildType.TEAM)
or
TeamSwarm(A, B, C, D)
or
Swarm(B, C, D, root_agent=A, build_type=GraphBuildType.TEAM)
```
The root_agent or first agent A is the leader; other agents interact with the leader A.
### Debate
TODO
### Hybrid
Hybrid is not a new type of builder of topology. Due to the use of different builders for the same topology,
the execution process varies, so hybrid builder is the fusion of **nested** topologies from different builders.
That is, interaction between multi-agents with multi-agents in different build modes.
For example, in a `WorkflowSwarm`, one node can be a `TeamSwarm`, `HandoffSwarm` or other. Or a node in a
`HandoffSwarm` can also be a `WorkflowSwarm` or other.
Example:
```python
A1 = Agent(name="A1", conf=agent_conf)
B1 = Agent(name="B1", conf=agent_conf)
C1 = Agent(name="C1", conf=agent_conf)
swarm1 = TeamSwarm(A1, B1, C1, build_type=GraphBuildType.TEAM)
Swarm(A, [B, C, swarm1], D)
```
The example shows that workflow swarm. After A completes execution, B, C, and swarm1(TeamSwarm) execute in parallel,
swarm1 will run in plan-execute mode until the end of the swarm1, and finally D is executed.
## Topology
The topology structure of multi-agent is represented by Swarm, Swarm's topology is built based on
various single agents，can use the topology type and build type Swarm to represent different structural types.
### Star
Each agent communicates with a single supervisor agent, also known as star topology,
a special structure of tree topology, also referred to as a team topology in **Aworld**.
A plan agent with other executing agents is a typical example.
```python
"""
Star topology:
┌───── plan ───┐
exec1 exec2
"""
plan = Agent(name="plan", conf=agent_conf)
exec1 = Agent(name="exec1", conf=agent_conf)
exec2 = Agent(name="exec2", conf=agent_conf)
```
We have two ways to construct this topology structure.
```python
swarm = Swarm((plan, exec1), (plan, exec2))
```
or use handoffs mechanism:
```python
plan = Agent(name="plan", conf=agent_conf, agent_names=['exec1', 'exec2'])
swarm = Swarm(plan, register_agents=[exec1, exec2])
```
or use team mechanism:
```python
# The order of the plan agent is the first.
swarm = TeamSwarm(plan, exec1, exec2,
build_type=GraphBuildType.TEAM)
```
Note:
- Whether to execute exec1 or exec2 is decided by LLM.
- If you want to execute all defined nodes with certainty, you need to use the `workflow` pattern.
Like this will execute all the defined nodes:
```python
swarm = Swarm(plan, [exec1, exec2])
```
- If it is necessary to execute exec1, whether to execute exec2 depends on LLM, you can define it as:
```python
plan = Agent(name="plan", conf=agent_conf, agent_names=['exec1', 'exec2'])
swarm = Swarm((plan, exec1), register_agents=[exec2])
```
That means that **GraphBuildType.WORKFLOW** is set, all nodes within the swarm will be executed.
### Tree
This is a generalization of the star topology and allows for more complex control flows.
#### Hierarchical
```python
"""
Hierarchical topology:
┌─────────── root ───────────┐
┌───── parent1 ───┐ ┌─────── parent2 ───────┐
leaf1_1 leaf1_2 leaf1_1 leaf2_2
"""
root = Agent(name="root", conf=agent_conf)
parent1 = Agent(name="parent1", conf=agent_conf)
parent2 = Agent(name="parent2", conf=agent_conf)
leaf1_1 = Agent(name="leaf1_1", conf=agent_conf)
leaf1_2 = Agent(name="leaf1_2", conf=agent_conf)
leaf2_1 = Agent(name="leaf2_1", conf=agent_conf)
leaf2_2 = Agent(name="leaf2_2", conf=agent_conf)
```
```python
swarm = Swarm((root, parent1), (root, parent2),
(parent1, leaf1_1), (parent1, leaf1_2),
(parent2, leaf2_1), (parent2, leaf2_2),
build_type=GraphBuildType.HANDOFF)
```
or use agent handoff:
```python
root = Agent(name="root", conf=agent_conf, agent_names=['parent1', 'parent2'])
parent1 = Agent(name="parent1", conf=agent_conf, agent_names=['leaf1_1', 'leaf1_2'])
parent2 = Agent(name="parent2", conf=agent_conf, agent_names=['leaf2_1', 'leaf2_2'])
swarm = HandoffSwarm((root, parent1), (root, parent2),
register_agents=[leaf1_1, leaf1_2, leaf2_1, leaf2_2])
```
#### Map-reduce
If the topology structure becomes further complex:
```
┌─────────── root ───────────┐
┌───── parent1 ───┐ ┌────── parent2 ──────┐
leaf1_1 leaf1_2 leaf1_1 leaf2_2
└─────result1─────┘ └───────result2───────┘
└───────────final───────────┘
```
We define it as **Map-reduce** topology, equivalent to workflow in terms of execution mode.
Build in this way:
```python
result1 = Agent(name="result1", conf=agent_conf)
result2 = Agent(name="result2", conf=agent_conf)
final = Agent(name="final", conf=agent_conf)
swarm = Swarm(
(root, [parent1, parent2]),
(parent1, [leaf1_1, leaf1_2]),
(parent2, [leaf2_1, leaf2_2]),
([leaf1_1, leaf1_2], result1),
([leaf2_1, leaf2_2], result2),
([result1, result2], final)
)
```
Assuming there is a cycle final -> root in the topology, define it as:
```python
final = LoopableAgent(name="final",
conf=agent_conf,
max_run_times=5,
loop_point=root.name(),
stop_func=...)
```
`stop_func` is a function that determines whether to terminate prematurely.
### Mesh
Divided into a fully meshed topology and a partially meshed topology.
Fully meshed topology means that each agent can communicate with every other agent,
any agent can decide which other agent to call next.
```python
"""
Fully Meshed topology:
┌─────────── A ──────────┐
B ───────────|────────── C
└─────────── D ─────────┘
"""
A = Agent(name="A", conf=agent_conf)
B = Agent(name="B", conf=agent_conf)
C = Agent(name="C", conf=agent_conf)
D = Agent(name="D", conf=agent_conf)
```
Network topology need to use the `handoffs` mechanism:
```python
swarm = HandoffsSwarm((A, B), (B, A),
(A, C), (C, A),
(A, D), (D, A),
(B, C), (C, B),
(B, D), (D, B),
(C, D), (D, C))
```
If a few pairs are removed, it becomes a partially meshed topology.
### Ring
A ring topology structure is a closed loop formed by nodes.
```python
"""
Ring topology:
┌───────────> A >──────────┐
B C
└───────────< D <─────────┘
"""
A = Agent(name="A", conf=agent_conf)
B = Agent(name="B", conf=agent_conf)
C = Agent(name="C", conf=agent_conf)
D = Agent(name="D", conf=agent_conf)
```
```python
swarm = Swarm((A, C), (C, D), (D, B), (B, A))
```
**Note:**
- This defined loop can only be executed once.
- If you want to execute multiple times, need to define it as:
```python
B = LoopableAgent(name="B", max_run_times=5, stop_func=...)
swarm = Swarm((A, C), (C, D), (D, B))
```
### hybrid
A generalization of topology, supporting an arbitrary combination of topologies, internally capable of
loops, parallel, serial dependencies, and groups.
## Execution
chapter9/gaia-experience/AWorld/aworld/agents/README.md
+107
View File
@@ -0,0 +1,107 @@
# AI Agents
Intelligent agents that control devices or tools in env using AI models or policy.
![Agent Architecture](../../readme_assets/framework_agent.png)
Most of the time, we directly use existing tools to build different types of agents that use LLM,
using frameworks makes it easy to write various agents.
Detailed steps for building an agent:
1. Define your `Agent`
2. Write prompt used to the agent, also choose not to set it.
3. Run it.
We provide a complete and simple example for writing an agent and multi-agent:
```python
from aworld.config.conf import AgentConfig
from aworld.agents.llm_agent import Agent
prompt = """
Please act as a search agent, constructing appropriate keywords and searach terms, using search toolkit to collect relevant information, including urls, webpage snapshots, etc.
Here are some tips that help you perform web search:
- Never add too many keywords in your search query! Some detailed results need to perform browser interaction to get, not using search toolkit.
- If the question is complex, search results typically do not provide precise answers. It is not likely to find the answer directly using search toolkit only, the search query should be concise and focuses on finding official sources rather than direct answers.
For example, as for the question "What is the maximum length in meters of #9 in the first National Geographic short on YouTube that was ever released according to the Monterey Bay Aquarium website?", your first search term must be coarse-grained like "National Geographic YouTube" to find the youtube website first, and then try other fine-grained search terms step-by-step to find more urls.
- The results you return do not have to directly answer the original question, you only need to collect relevant information.
Here are the question: {task}
Please perform web search and return the listed search result, including urls and necessary webpage snapshots, introductions, etc.
Your output should be like the followings (at most 3 relevant pages from coa):
[
{{
"url": [URL],
"information": [INFORMATION OR CONTENT]
}},
...
]
"""
# Step1
agent_config = AgentConfig(
llm_provider="openai",
llm_model_name="gpt-4o",
llm_temperature=1,
# need to set llm_api_key for use LLM
llm_api_key=""
)
search = Agent(
conf=agent_config,
name="search_agent",
system_prompt="You are a helpful search agent.",
# used to opt the result, also choose not to set it
agent_prompt=prompt,
tool_names=["search_api"]
)
```
It can also quickly develop multi-agent based on the framework.
On the basis of the above agent(SearchAgent), we provide a multi-agent example:
```python
from aworld.agents.llm_agent import Agent
summary_prompt = """
Summarize the following text in one clear and concise paragraph, capturing the key ideas without missing critical points.
Ensure the summary is easy to understand and avoids excessive detail.
Here are the content:
{task}
"""
summary = Agent(
conf=agent_config,
name="summary_agent",
system_prompt="You are a helpful general summary agent.",
# used to opt the result, also choose not to set it
agent_prompt=summary_prompt
)
```
You can run single-agent or multi-agent through Swarm.
NOTE: Need to set some environment variables first! Effective GOOGLE_API_KEY, GOOGLE_ENGINE_ID, OPENAI_API_KEY and OPENAI_ENDPOINT.
```python
from aworld.core.agent.swarm import Swarm
from aworld.runner import Runners
if __name__ == '__main__':
task = "search 1+1=?"
# build topology graph, the correct order is necessary
swarm = Swarm(search, summary, max_steps=1)
prefix = ""
# can special search google, wiki, duck go, or baidu. such as:
# prefix = "search wiki: "
res = Runners.sync_run(
input=prefix + """What is an agent.""",
swarm=swarm
)
```
You can view search example [code](../../examples/multi_agents/workflow/search).
chapter9/gaia-experience/AWorld/aworld/agents/__init__.py
+2
View File
@@ -0,0 +1,2 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
chapter9/gaia-experience/AWorld/aworld/agents/llm_agent.py
+935
View File
@@ -0,0 +1,935 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
import copy
import json
import time
import traceback
import uuid
from collections import OrderedDict
from datetime import datetime
from typing import Dict, Any, List, Callable, Optional
import aworld.trace as trace
from aworld.core.agent.agent_desc import get_agent_desc
from aworld.core.agent.base import BaseAgent, AgentResult, is_agent_by_name, is_agent
from aworld.core.common import ActionResult, Observation, ActionModel, Config, TaskItem
from aworld.core.context.base import Context
from aworld.core.context.processor.prompt_processor import PromptProcessor
from aworld.core.context.prompts import BasePromptTemplate
from aworld.core.context.prompts.string_prompt_template import StringPromptTemplate
from aworld.core.event import eventbus
from aworld.core.event.base import Message, ToolMessage, Constants, AgentMessage, GroupMessage, TopicType
from aworld.core.model_output_parser import ModelOutputParser
from aworld.core.tool.tool_desc import get_tool_desc
from aworld.events.util import send_message
from aworld.logs.util import logger, color_log, Color
from aworld.mcp_client.utils import mcp_tool_desc_transform
from aworld.memory.main import MemoryFactory
from aworld.memory.models import MessageMetadata, MemoryAIMessage, MemoryToolMessage, MemoryHumanMessage, \
MemorySystemMessage, MemoryMessage
from aworld.models.llm import get_llm_model, acall_llm_model, acall_llm_model_stream
from aworld.models.model_response import ModelResponse, ToolCall, LLMResponseError
from aworld.models.utils import tool_desc_transform, agent_desc_transform
from aworld.output import Outputs
from aworld.output.base import MessageOutput, Output
from aworld.runners.hook.hooks import HookPoint
from aworld.sandbox.base import Sandbox
from aworld.trace.constants import SPAN_NAME_PREFIX_AGENT
from aworld.trace.instrumentation import semconv
from aworld.utils.common import sync_exec, nest_dict_counter
from aworld.utils.serialized_util import to_serializable
class LlmOutputParser(ModelOutputParser[ModelResponse, AgentResult]):
async def parse(self, resp: ModelResponse, **kwargs) -> AgentResult:
"""Standard parse based Openai API."""
if not resp:
logger.warning("no valid content to parse!")
return AgentResult(actions=[], current_state=None)
agent_id = kwargs.get("agent_id")
if not agent_id:
logger.warning("need agent_id param.")
raise RuntimeError("no `agent_id` param.")
results = []
is_call_tool = False
content = '' if resp.content is None else resp.content
if kwargs.get("use_tools_in_prompt"):
tool_calls = []
for tool in self.use_tool_list(content):
tool_calls.append(ToolCall.from_dict({
"id": tool.get("id"),
"function": {
"name": tool.get("tool"),
"arguments": tool.get("arguments")
}
}))
if tool_calls:
resp.tool_calls = tool_calls
if resp.tool_calls:
is_call_tool = True
for tool_call in resp.tool_calls:
full_name: str = tool_call.function.name
if not full_name:
logger.warning("tool call response no tool name.")
continue
try:
params = json.loads(tool_call.function.arguments)
except:
logger.warning(f"{tool_call.function.arguments} parse to json fail.")
params = {}
# format in framework
names = full_name.split("__")
tool_name = names[0]
if is_agent_by_name(full_name):
param_info = params.get('content', "") + ' ' + params.get('info', '')
results.append(ActionModel(tool_name=full_name,
tool_call_id=tool_call.id,
agent_name=agent_id,
params=params,
policy_info=content + param_info))
else:
action_name = '__'.join(names[1:]) if len(names) > 1 else ''
results.append(ActionModel(tool_name=tool_name,
tool_call_id=tool_call.id,
action_name=action_name,
agent_name=agent_id,
params=params,
policy_info=content))
else:
content = content.replace("```json", "").replace("```", "")
results.append(ActionModel(agent_name=agent_id, policy_info=content))
return AgentResult(actions=results, current_state=None, is_call_tool=is_call_tool)
def use_tool_list(self, content: str) -> List[Dict[str, Any]]:
tool_list = []
try:
content = content.replace('\n', '').replace('\r', '')
response_json = json.loads(content)
use_tool_list = response_json.get("use_tool_list", [])
for use_tool in use_tool_list:
tool_name = use_tool.get("tool", None)
if tool_name:
tool_list.append(use_tool)
except Exception:
logger.debug(f"tool_parse error, content: {content}, \n{traceback.format_exc()}")
return tool_list
class Agent(BaseAgent[Observation, List[ActionModel]]):
"""Basic agent for unified protocol within the framework."""
def __init__(self,
name: str,
conf: Config | None = None,
desc: str = None,
agent_id: str = None,
*,
task: Any = None,
tool_names: List[str] = None,
agent_names: List[str] = None,
mcp_servers: List[str] = None,
mcp_config: Dict[str, Any] = None,
feedback_tool_result: bool = True,
wait_tool_result: bool = False,
sandbox: Sandbox = None,
system_prompt: str = None,
system_prompt_template: BasePromptTemplate = None,
agent_prompt: str = None,
need_reset: bool = True,
step_reset: bool = True,
use_tools_in_prompt: bool = False,
black_tool_actions: Dict[str, List[str]] = None,
model_output_parser: ModelOutputParser[..., AgentResult] = LlmOutputParser(),
tool_aggregate_func: Callable[..., Any] = None,
event_handler_name: str = None,
event_driven: bool = True,
**kwargs):
"""A api class implementation of agent, using the `Observation` and `List[ActionModel]` protocols.
Args:
system_prompt: Instruction of the agent.
agent_prompt: Optimized prompt of the agent.
need_reset: Whether need to reset the status in start.
step_reset: Reset the status at each step
use_tools_in_prompt: Whether the tool description in prompt.
black_tool_actions: Black list of actions of the tool.
model_output_parser: Llm response parse function for the agent standard output, transform llm response.
tool_aggregate_func: Aggregation strategy for multiple tool results.
event_handler_name: Custom handlers for certain types of events.
"""
super(Agent, self).__init__(name, conf, desc, agent_id,
task=task,
tool_names=tool_names,
agent_names=agent_names,
mcp_servers=mcp_servers,
mcp_config=mcp_config,
black_tool_actions=black_tool_actions,
feedback_tool_result=feedback_tool_result,
wait_tool_result=wait_tool_result,
sandbox=sandbox,
**kwargs)
conf = self.conf
self.model_name = conf.llm_config.llm_model_name
self._llm = None
self.memory = MemoryFactory.instance()
self.memory_config = conf.memory_config
self.system_prompt: str = system_prompt if system_prompt else conf.system_prompt
self.system_prompt_template: str = system_prompt_template if (
system_prompt_template) else conf.system_prompt_template
# for backward compatibility
if not self.system_prompt_template:
self.system_prompt_template = StringPromptTemplate.from_template(self.system_prompt)
if isinstance(self.system_prompt_template, str):
self.system_prompt_template = StringPromptTemplate.from_template(self.system_prompt_template)
if not self.system_prompt:
self.system_prompt = self.system_prompt_template.template
self.agent_prompt: str = agent_prompt if agent_prompt else conf.agent_prompt
self.event_driven = event_driven
self.need_reset = need_reset if need_reset else conf.need_reset
# whether to keep contextual information, False means keep, True means reset in every step by the agent call
self.step_reset = step_reset
# tool_name: [tool_action1, tool_action2, ...]
# self.black_tool_actions: Dict[str, List[str]] = black_tool_actions if black_tool_actions \
# else conf.get('black_tool_actions', {})
self.model_output_parser = model_output_parser
self.use_tools_in_prompt = use_tools_in_prompt if use_tools_in_prompt else conf.use_tools_in_prompt
self.tools_aggregate_func = tool_aggregate_func if tool_aggregate_func else self._tools_aggregate_func
self.event_handler_name = event_handler_name
@property
def llm(self):
# lazy
if self._llm is None:
llm_config = self.conf.llm_config or None
conf = llm_config if llm_config and (
llm_config.llm_provider or llm_config.llm_base_url or llm_config.llm_api_key or llm_config.llm_model_name) else self.conf
self._llm = get_llm_model(conf)
return self._llm
def desc_transform(self, context: Context) -> None:
"""Transform of descriptions of supported tools, agents, and MCP servers in the framework to support function calls of LLM."""
sync_exec(self.async_desc_transform, context)
async def async_desc_transform(self, context: Context) -> None:
"""Transform of descriptions of supported tools, agents, and MCP servers in the framework to support function calls of LLM."""
# Stateless tool
self.tools = tool_desc_transform(get_tool_desc(),
tools=self.tool_names if self.tool_names else [],
black_tool_actions=self.black_tool_actions)
# Agents as tool
self.tools.extend(agent_desc_transform(get_agent_desc(),
agents=self.handoffs if self.handoffs else []))
# MCP servers are tools
if self.sandbox:
mcp_tools = await self.sandbox.mcpservers.list_tools(context)
self.tools.extend(mcp_tools)
else:
self.tools.extend(await mcp_tool_desc_transform(self.mcp_servers, self.mcp_config))
def messages_transform(self,
content: str,
image_urls: List[str] = None,
observation: Observation = None,
message: Message = None,
**kwargs) -> List[Dict[str, Any]]:
return sync_exec(self.async_messages_transform, image_urls=image_urls, observation=observation,
message=message, **kwargs)
async def async_messages_transform(self,
image_urls: List[str] = None,
observation: Observation = None,
message: Message = None,
**kwargs) -> List[Dict[str, Any]]:
"""Transform the original content to LLM messages of native format.
Args:
observation: Observation by env.
image_urls: List of images encoded using base64.
message: Event received by the Agent.
Returns:
Message list for LLM.
"""
agent_prompt = self.agent_prompt
messages = []
# append sys_prompt to memory
await self._add_system_message_to_memory(context=message.context, content=observation.content)
session_id = message.context.get_task().session_id
task_id = message.context.get_task().id
histories = self.memory.get_all(filters={
"agent_id": self.id(),
"session_id": session_id,
"task_id": task_id,
"memory_type": "message"
})
last_history = histories[-1] if histories and len(histories) > 0 else None
# append observation to memory
if observation.is_tool_result:
for action_item in observation.action_result:
tool_call_id = action_item.tool_call_id
await self._add_tool_result_to_memory(tool_call_id, tool_result=action_item, context=message.context)
elif last_history and last_history.metadata and "tool_calls" in last_history.metadata and \
last_history.metadata[
'tool_calls']:
for tool_call in last_history.metadata['tool_calls']:
tool_call_id = tool_call['id']
tool_name = tool_call['function']['name']
if tool_name and tool_name == message.sender:
await self._add_tool_result_to_memory(tool_call_id, tool_result=observation.content,
context=message.context)
break
else:
content = observation.content
logger.debug(f"agent_prompt: {agent_prompt}")
if agent_prompt:
content = agent_prompt.format(task=content, current_date=datetime.now().strftime("%Y-%m-%d"))
if image_urls:
urls = [{'type': 'text', 'text': content}]
for image_url in image_urls:
urls.append(
{'type': 'image_url', 'image_url': {"url": image_url}})
content = urls
await self._add_human_input_to_memory(content, message.context, memory_type="message")
# from memory get last n messages
histories = self.memory.get_last_n(self.memory_config.history_rounds, filters={
"agent_id": self.id(),
"session_id": session_id,
"task_id": task_id
}, agent_memory_config=self.memory_config)
if histories:
# default use the first tool call
for history in histories:
if isinstance(history, MemoryMessage):
messages.append(history.to_openai_message())
else:
if not self.use_tools_in_prompt and "tool_calls" in history.metadata and history.metadata[
'tool_calls']:
messages.append({'role': history.metadata['role'], 'content': history.content,
'tool_calls': [history.metadata["tool_calls"][0]]})
else:
messages.append({'role': history.metadata['role'], 'content': history.content,
"tool_call_id": history.metadata.get("tool_call_id")})
return messages
async def init_observation(self, observation: Observation) -> Observation:
# supported string only
# if self.task and isinstance(self.task, str) and self.task != observation.content:
# observation.content = f"base task is: {self.task}\n{observation.content}"
# # `task` only needs to be processed once and reflected in the context
# self.task = None
# default use origin observation
return observation
def _log_messages(self, messages: List[Dict[str, Any]], **kwargs) -> None:
"""Log the sequence of messages for debugging purposes"""
logger.info(f"[agent] Invoking LLM with {len(messages)} messages:")
logger.debug(f"[agent] use tools: {self.tools}")
for i, msg in enumerate(messages):
prefix = msg.get('role')
logger.info(
f"[agent] Message {i + 1}: {prefix} ===================================")
if isinstance(msg['content'], list):
try:
for item in msg['content']:
if item.get('type') == 'text':
logger.info(
f"[agent] Text content: {item.get('text')}")
elif item.get('type') == 'image_url':
image_url = item.get('image_url', {}).get('url', '')
if image_url.startswith('data:image'):
logger.info(f"[agent] Image: [Base64 image data]")
else:
logger.info(
f"[agent] Image URL: {image_url[:30]}...")
except Exception as e:
logger.error(f"[agent] Error parsing msg['content']: {msg}. Error: {e}")
content = str(msg['content'])
chunk_size = 500
for j in range(0, len(content), chunk_size):
chunk = content[j:j + chunk_size]
if j == 0:
logger.info(f"[agent] Content: {chunk}")
else:
logger.info(f"[agent] Content (continued): {chunk}")
else:
content = str(msg['content'])
chunk_size = 500
for j in range(0, len(content), chunk_size):
chunk = content[j:j + chunk_size]
if j == 0:
logger.info(f"[agent] Content: {chunk}")
else:
logger.info(f"[agent] Content (continued): {chunk}")
if 'tool_calls' in msg and msg['tool_calls']:
for tool_call in msg.get('tool_calls'):
if isinstance(tool_call, dict):
logger.info(
f"[agent] Tool call: {tool_call.get('function', {}).get('name', {})} - ID: {tool_call.get('id')}")
args = str(tool_call.get('function', {}).get(
'arguments', {}))[:1000]
logger.info(f"[agent] Tool args: {args}...")
elif isinstance(tool_call, ToolCall):
logger.info(
f"[agent] Tool call: {tool_call.function.name} - ID: {tool_call.id}")
args = str(tool_call.function.arguments)[:1000]
logger.info(f"[agent] Tool args: {args}...")
def _agent_result(self, actions: List[ActionModel], caller: str, input_message: Message):
if not actions:
raise Exception(f'{self.id()} no action decision has been made.')
if self.event_handler_name:
return Message(payload=actions,
caller=caller,
sender=self.id(),
receiver=actions[0].tool_name,
category=self.event_handler_name,
session_id=input_message.context.session_id if input_message.context else "",
headers=self._update_headers(input_message))
tools = OrderedDict()
agents = []
for action in actions:
if is_agent(action):
agents.append(action)
else:
if action.tool_name not in tools:
tools[action.tool_name] = []
tools[action.tool_name].append(action)
_group_name = None
# agents and tools exist simultaneously, more than one agent/tool name
if (agents and tools) or len(agents) > 1 or len(tools) > 1:
_group_name = f"{self.id()}_{uuid.uuid1().hex}"
# complex processing
if _group_name:
return GroupMessage(payload=actions,
caller=caller,
sender=self.id(),
receiver=actions[0].tool_name,
session_id=input_message.context.session_id if input_message.context else "",
group_id=_group_name,
topic=TopicType.GROUP_ACTIONS,
headers=self._update_headers(input_message))
elif agents:
return AgentMessage(payload=actions,
caller=caller,
sender=self.id(),
receiver=actions[0].tool_name,
session_id=input_message.context.session_id if input_message.context else "",
headers=self._update_headers(input_message))
else:
return ToolMessage(payload=actions,
caller=caller,
sender=self.id(),
receiver=actions[0].tool_name,
session_id=input_message.context.session_id if input_message.context else "",
headers=self._update_headers(input_message))
def post_run(self, policy_result: List[ActionModel], policy_input: Observation, message: Message = None) -> Message:
return self._agent_result(
policy_result,
policy_input.from_agent_name if policy_input.from_agent_name else policy_input.observer,
message
)
async def async_post_run(self, policy_result: List[ActionModel], policy_input: Observation,
message: Message = None) -> Message:
return self._agent_result(
policy_result,
policy_input.from_agent_name if policy_input.from_agent_name else policy_input.observer,
message
)
def policy(self, observation: Observation, info: Dict[str, Any] = {}, message: Message = None, **kwargs) -> List[
ActionModel]:
"""The strategy of an agent can be to decide which tools to use in the environment, or to delegate tasks to other agents.
Args:
observation: The state observed from tools in the environment.
info: Extended information is used to assist the agent to decide a policy.
Returns:
ActionModel sequence from agent policy
"""
return sync_exec(self.async_policy, observation, info, message, **kwargs)
async def async_policy(self, observation: Observation, info: Dict[str, Any] = {}, message: Message = None,
**kwargs) -> List[ActionModel]:
"""The strategy of an agent can be to decide which tools to use in the environment, or to delegate tasks to other agents.
Args:
observation: The state observed from tools in the environment.
info: Extended information is used to assist the agent to decide a policy.
Returns:
ActionModel sequence from agent policy
"""
logger.info(f"Agent{type(self)}#{self.id()}: async_policy start")
# Get current step information for trace recording
source_span = trace.get_current_span()
self._finished = False
if hasattr(observation, 'context') and observation.context:
self.task_histories = observation.context
try:
events = []
async for event in self.run_hooks(message.context, HookPoint.PRE_LLM_CALL):
events.append(event)
except Exception:
logger.debug(traceback.format_exc())
messages = await self.build_llm_input(observation, info, message=message, **kwargs)
serializable_messages = to_serializable(messages)
llm_response = None
if source_span:
source_span.set_attribute("messages", json.dumps(serializable_messages, ensure_ascii=False))
try:
llm_response = await self.invoke_model(messages, message=message, **kwargs)
except Exception as e:
logger.warn(traceback.format_exc())
raise e
finally:
if llm_response:
if llm_response.error:
logger.info(f"llm result error: {llm_response.error}")
if eventbus is not None:
output_message = Message(
category=Constants.OUTPUT,
payload=Output(
data=f"llm result error: {llm_response.error}"
),
sender=self.id(),
session_id=message.context.session_id if message.context else "",
headers={"context": message.context}
)
await send_message(output_message)
else:
await self._add_llm_response_to_memory(llm_response, message.context, history_messages=messages)
else:
logger.error(f"{self.id()} failed to get LLM response")
raise RuntimeError(f"{self.id()} failed to get LLM response")
try:
events = []
async for event in self.run_hooks(message.context, HookPoint.POST_LLM_CALL):
events.append(event)
except Exception as e:
logger.debug(traceback.format_exc())
agent_result = await self.model_output_parser.parse(llm_response,
agent_id=self.id(),
use_tools_in_prompt=self.use_tools_in_prompt)
logger.info(f"agent_result: {agent_result}")
policy_result: Optional[List[ActionModel]] = None
if self.is_agent_finished(llm_response, agent_result):
policy_result = agent_result.actions
else:
if not self.wait_tool_result:
policy_result = agent_result.actions
else:
policy_result = await self.execution_tools(agent_result.actions, message)
await self.send_llm_response_output(llm_response, agent_result, message.context, kwargs.get("outputs"))
return policy_result
async def execution_tools(self, actions: List[ActionModel], message: Message = None, **kwargs) -> List[ActionModel]:
"""Tool execution operations.
Returns:
ActionModel sequence. Tool execution result.
"""
from aworld.utils.run_util import exec_tool
tool_results = []
for act in actions:
if is_agent(act):
continue
act_result = await exec_tool(tool_name=act.tool_name,
action_name=act.action_name,
params=act.params,
agent_name=self.id(),
context=message.context.deep_copy(),
sub_task=True,
outputs=message.context.outputs,
task_group_id=message.context.get_task().group_id or uuid.uuid4().hex)
if not act_result.success:
color_log(f"Agent {self.id()} _execute_tool failed with exception: {act_result.msg}",
color=Color.red)
continue
tool_results.append(
ActionResult(tool_call_id=act.tool_call_id, tool_name=act.tool_name, content=act_result.answer))
await self._add_tool_result_to_memory(act.tool_call_id, act_result.answer,
context=message.context)
result = sync_exec(self.tools_aggregate_func, tool_results)
return result
async def _tools_aggregate_func(self, tool_results: List[ActionResult]) -> List[ActionModel]:
"""Aggregate tool results
Args:
tool_results: Tool results
Returns:
ActionModel sequence
"""
content = ""
for res in tool_results:
content += f"{res.content}\n"
return [ActionModel(agent_name=self.id(), policy_info=content)]
async def build_llm_input(self,
observation: Observation,
info: Dict[str, Any] = {},
message: Message = None,
**kwargs):
"""Build LLM input.
Args:
observation: The state observed from the environment
info: Extended information to assist the agent in decision-making
"""
await self.async_desc_transform(message.context)
# observation secondary processing
observation = await self.init_observation(observation)
images = observation.images if self.conf.use_vision else None
if self.conf.use_vision and not images and observation.image:
images = [observation.image]
messages = await self.async_messages_transform(image_urls=images, observation=observation, message=message)
# truncate and other process
try:
messages = self._process_messages(messages=messages, context=message.context)
except Exception as e:
logger.warning(f"Failed to process messages in messages_transform: {e}")
logger.debug(f"Process messages error details: {traceback.format_exc()}")
self._log_messages(messages, context=message.context)
return messages
def _process_messages(self, messages: List[Dict[str, Any]],
context: Context = None) -> Optional[List[Dict[str, Any]]]:
origin_messages = messages
st = time.time()
with trace.span(f"{SPAN_NAME_PREFIX_AGENT}llm_context_process", attributes={
"start_time": st,
semconv.AGENT_ID: self.id()
}) as compress_span:
if self.conf.context_rule is None:
logger.debug('debug|skip process_messages context_rule is None')
return messages
origin_len = compressed_len = len(str(messages))
origin_messages_count = truncated_messages_count = len(messages)
try:
prompt_processor = PromptProcessor(self.conf.context_rule, self.conf.llm_config)
result = prompt_processor.process_messages(messages, context)
messages = result.processed_messages
compressed_len = len(str(messages))
truncated_messages_count = len(messages)
logger.debug(
f'debug|llm_context_process|{origin_len}|{compressed_len}|{origin_messages_count}|{truncated_messages_count}|\n|{origin_messages}\n|{messages}')
return messages
finally:
compress_span.set_attributes({
"end_time": time.time(),
"duration": time.time() - st,
# messages length
"origin_messages_count": origin_messages_count,
"truncated_messages_count": truncated_messages_count,
"truncated_ratio": round(truncated_messages_count / origin_messages_count,
2) if origin_messages_count > 0 else 0,
# token length
"origin_len": origin_len,
"compressed_len": compressed_len,
"compress_ratio": round(compressed_len / origin_len, 2)
})
async def invoke_model(self,
messages: List[Dict[str, str]] = [],
message: Message = None,
**kwargs) -> ModelResponse:
"""Perform LLM call.
Args:
messages: LLM model input messages.
message: Event message.
**kwargs: Other parameters
Returns:
LLM response
"""
llm_response = None
source_span = trace.get_current_span()
serializable_messages = to_serializable(messages)
message.context.context_info["llm_input"] = serializable_messages
if source_span:
source_span.set_attribute("messages", json.dumps(
serializable_messages, ensure_ascii=False))
try:
stream_mode = kwargs.get("stream", False)
float_temperature = float(self.conf.llm_config.llm_temperature)
if stream_mode:
llm_response = ModelResponse(
id="", model="", content="", tool_calls=[])
resp_stream = acall_llm_model_stream(
self.llm,
messages=messages,
model=self.model_name,
temperature=float_temperature,
tools=self.tools if not self.use_tools_in_prompt and self.tools else None,
stream=True
)
async def async_call_llm(resp_stream, json_parse=False):
llm_resp = ModelResponse(
id="", model="", content="", tool_calls=[])
# Async streaming with acall_llm_model
async def async_generator():
async for chunk in resp_stream:
if chunk.content:
llm_resp.content += chunk.content
yield chunk.content
if chunk.tool_calls:
llm_resp.tool_calls.extend(chunk.tool_calls)
if chunk.error:
llm_resp.error = chunk.error
llm_resp.id = chunk.id
llm_resp.model = chunk.model
llm_resp.usage = nest_dict_counter(
llm_resp.usage, chunk.usage)
return MessageOutput(source=async_generator(), json_parse=json_parse), llm_resp
output, response = await async_call_llm(resp_stream)
llm_response = response
else:
llm_response = await acall_llm_model(
self.llm,
messages=messages,
model=self.model_name,
temperature=float_temperature,
tools=self.tools if not self.use_tools_in_prompt and self.tools else None,
stream=kwargs.get("stream", False)
)
logger.info(f"Execute response: {json.dumps(llm_response.to_dict(), ensure_ascii=False)}")
except Exception as e:
logger.warn(traceback.format_exc())
await send_message(Message(
category=Constants.OUTPUT,
payload=Output(
data=f"Failed to call llm model: {e}"
),
sender=self.id(),
session_id=message.context.session_id if message.context else "",
headers={"context": message.context}
))
if "Please reduce the length of the messages" in str(e):
# Meaning context too long, will return directly. You can develop a Processor to truncate or compress it.
await send_message(Message(
category=Constants.TASK,
topic=TopicType.CANCEL,
payload=TaskItem(data=messages, msg=str(e)),
sender=self.id(),
priority=-1,
session_id=message.context.session_id if message.context else "",
headers={"context": message.context}
))
return ModelResponse(id=uuid.uuid4().hex, model=self.model_name, content=to_serializable(messages))
raise e
finally:
message.context.context_info["llm_output"] = llm_response
return llm_response
def _init_context(self, context: Context):
super()._init_context(context)
logger.debug(f'init_context llm_agent {self.name()} {self.conf} {self.conf.context_rule}')
async def run_hooks(self, context: Context, hook_point: str):
"""Execute hooks asynchronously"""
from aworld.runners.hook.hook_factory import HookFactory
from aworld.core.event.base import Message
# Get all hooks for the specified hook point
all_hooks = HookFactory.hooks(hook_point)
hooks = all_hooks.get(hook_point, [])
for hook in hooks:
try:
# Create a temporary Message object to pass to the hook
message = Message(
category="agent_hook",
payload=None,
sender=self.id(),
session_id=context.session_id if hasattr(
context, 'session_id') else None,
headers={"context": message.context}
)
# Execute hook
msg = await hook.exec(message, context)
if msg:
logger.debug(f"Hook {hook.point()} executed successfully")
yield msg
except Exception as e:
logger.warning(f"Hook {hook.point()} execution failed: {traceback.format_exc()}")
async def _add_system_message_to_memory(self, context: Context, content: str):
if not self.system_prompt:
return
session_id = context.get_task().session_id
task_id = context.get_task().id
user_id = context.get_task().user_id
histories = self.memory.get_last_n(0, filters={
"agent_id": self.id(),
"session_id": session_id,
"task_id": task_id
}, agent_memory_config=self.memory_config)
if histories:
logger.debug(f"🧠 [MEMORY:short-term] histories is not empty, do not need add system input to agent memory")
return
content = await self.custom_system_prompt(context=context, content=content, tool_list=self.tools)
await self.memory.add(MemorySystemMessage(
content=content,
metadata=MessageMetadata(
session_id=session_id,
user_id=user_id,
task_id=task_id,
agent_id=self.id(),
agent_name=self.name(),
)
), agent_memory_config=self.memory_config)
async def custom_system_prompt(self, context: Context, content: str, tool_list: List[str] = None):
logger.info(f"llm_agent custom_system_prompt .. agent#{type(self)}#{self.id()}")
return self.system_prompt_template.format(context=context, task=content, tool_list=tool_list)
async def _add_human_input_to_memory(self, content: Any, context: Context, memory_type="init"):
"""Add user input to memory"""
session_id = context.get_task().session_id
user_id = context.get_task().user_id
task_id = context.get_task().id
await self.memory.add(MemoryHumanMessage(
content=content,
metadata=MessageMetadata(
session_id=session_id,
user_id=user_id,
task_id=task_id,
agent_id=self.id(),
agent_name=self.name(),
),
memory_type=memory_type
), agent_memory_config=self.memory_config)
async def _add_llm_response_to_memory(self, llm_response, context: Context, history_messages: list, **kwargs):
"""Add LLM response to memory"""
ai_message = MemoryAIMessage(
content=llm_response.content,
tool_calls=llm_response.tool_calls,
metadata=MessageMetadata(
session_id=context.get_task().session_id,
user_id=context.get_task().user_id,
task_id=context.get_task().id,
agent_id=self.id(),
agent_name=self.name()
)
)
await self.memory.add(ai_message, agent_memory_config=self.memory_config)
async def _add_tool_result_to_memory(self, tool_call_id: str, tool_result: ActionResult, context: Context):
"""Add tool result to memory"""
if hasattr(tool_result, 'content') and isinstance(tool_result.content, str) and tool_result.content.startswith(
"data:image"):
image_content = tool_result.content
tool_result.content = "this picture is below "
await self._do_add_tool_result_to_memory(tool_call_id, tool_result, context)
image_content = [
{
"type": "text",
"text": f"this is file of tool_call_id:{tool_result.tool_call_id}"
},
{
"type": "image_url",
"image_url": {
"url": image_content
}
}
]
await self._add_human_input_to_memory(image_content, context, "message")
else:
await self._do_add_tool_result_to_memory(tool_call_id, tool_result, context)
async def _do_add_tool_result_to_memory(self, tool_call_id: str, tool_result: ActionResult, context: Context):
"""Add tool result to memory"""
tool_use_summary = None
if isinstance(tool_result, ActionResult):
tool_use_summary = tool_result.metadata.get("tool_use_summary")
await self.memory.add(MemoryToolMessage(
content=tool_result.content if hasattr(tool_result, 'content') else tool_result,
tool_call_id=tool_call_id,
status="success",
metadata=MessageMetadata(
session_id=context.get_task().session_id,
user_id=context.get_task().user_id,
task_id=context.get_task().id,
agent_id=self.id(),
agent_name=self.name(),
summary_content=tool_use_summary
)
), agent_memory_config=self.memory_config)
async def send_llm_response_output(self, llm_response: ModelResponse, agent_result: AgentResult, context: Context,
outputs: Outputs = None):
"""Send LLM response to output"""
if not llm_response or llm_response.error:
return
if eventbus is None:
logger.warn("=============== eventbus is none ============")
llm_resp_output = MessageOutput(
source=llm_response,
metadata={"agent_id": self.id(), "agent_name": self.name(), "is_finished": self.finished}
)
if eventbus is not None and llm_response:
await send_message(Message(
category=Constants.OUTPUT,
payload=llm_resp_output,
sender=self.id(),
session_id=context.session_id if context else "",
headers={"context": context}
))
elif not self.event_driven and outputs:
await outputs.add_output(llm_resp_output)
def is_agent_finished(self, llm_response: ModelResponse, agent_result: AgentResult) -> bool:
if not agent_result.is_call_tool:
self._finished = True
return self.finished
def _update_headers(self, input_message: Message) -> Dict[str, Any]:
headers = input_message.headers.copy()
headers['context'] = input_message.context
headers['level'] = headers.get('level', 0) + 1
if input_message.group_id:
headers['parent_group_id'] = input_message.group_id
return headers
chapter9/gaia-experience/AWorld/aworld/agents/loop_llm_agent.py
+46
View File
@@ -0,0 +1,46 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
from typing import Any, Callable
from aworld.agents.llm_agent import Agent
class LoopableAgent(Agent):
"""Support for loop agents in the swarm.
The parameters of the extension function are the agent itself, which can obtain internal information of the agent.
`stop_func` function example:
>>> def stop(agent: LoopableAgent):
>>> ...
`loop_point_finder` function example:
>>> def find(agent: LoopableAgent):
>>> ...
"""
max_run_times: int = 1
cur_run_times: int = 0
# The loop agent special the loop point (agent name)
loop_point: str = None
# Used to determine the loop point for multiple loops
loop_point_finder: Callable[..., Any] = None
# def stop(agent: LoopableAgent): ...
stop_func: Callable[..., Any] = None
@property
def goto(self):
"""The next loop point is what the loop agent wants to reach."""
if self.loop_point_finder:
return self.loop_point_finder(self)
if self.loop_point:
return self.loop_point
return self.id()
@property
def finished(self) -> bool:
"""Loop agent termination state detection, achieved loop count or termination condition."""
if self.cur_run_times >= self.max_run_times or (self.stop_func and self.stop_func(self)):
self._finished = True
return True
self._finished = False
return False
chapter9/gaia-experience/AWorld/aworld/agents/parallel_llm_agent.py
+67
View File
@@ -0,0 +1,67 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
import asyncio
from typing import List, Dict, Any, Callable
from aworld.agents.llm_agent import Agent
from aworld.core.common import Observation, ActionModel
from aworld.core.event.base import Message
from aworld.utils.run_util import exec_agent
class ParallelizableAgent(Agent):
"""Support for parallel agents in the swarm.
The parameters of the extension function are the agent itself, which can obtain internal information of the agent.
`aggregate_func` function example:
>>> def agg(agent: ParallelizableAgent, res: Dict[str, Any]) -> ActionModel:
>>> ...
"""
def __init__(self,
agents: List[Agent] = None,
aggregate_func: Callable[['ParallelizableAgent', Dict[str, Any]], ActionModel] = None,
**kwargs):
super().__init__(**kwargs)
self.agents = agents if agents else []
# The function of aggregating the results of the parallel execution of agents.
self.aggregate_func = aggregate_func
async def async_policy(self, observation: Observation, info: Dict[str, Any] = {}, **kwargs) -> List[ActionModel]:
tasks = []
if self.agents:
for agent in self.agents:
tasks.append(asyncio.create_task(exec_agent(observation.content, agent, self.context, sub_task=True)))
results = await asyncio.gather(*tasks)
res = []
for idx, result in enumerate(results):
if result.success:
con = result.answer
else:
con = result.msg
res.append(ActionModel(agent_name=self.agents[idx].id(), policy_info=con))
if self.aggregate_func:
res = [self.aggregate_func(self, {action.agent_name: action.policy_info for action in res})]
return res
async def _agent_result(self, actions: List[ActionModel], caller: str, input_message: Message):
if self.aggregate_func:
return super()._agent_result(actions, caller, input_message)
if not actions:
raise Exception(f'{self.id()} no action decision has been made.')
action = ActionModel(agent_name=self.id(),
policy_info={action.agent_name: action.policy_info for action in actions})
return Message(payload=[action],
caller=caller,
sender=self.id(),
receiver=actions[0].tool_name,
category=self.event_handler_name,
session_id=input_message.context.session_id if input_message.context else "",
headers=self._update_headers(input_message))
def finished(self) -> bool:
return all([agent.finished for agent in self.agents])
chapter9/gaia-experience/AWorld/aworld/agents/serial_llm_agent.py
+64
View File
@@ -0,0 +1,64 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
from typing import List, Dict, Any, Callable
from aworld.core.event.base import Message
from aworld.utils.run_util import exec_agent
from aworld.agents.llm_agent import Agent
from aworld.core.common import Observation, ActionModel, Config
from aworld.logs.util import logger
class SerialableAgent(Agent):
"""Support for serial execution of agents based on dependency relationships in the swarm.
The parameters of the extension function are the agent itself, which can obtain internal information of the agent.
`aggregate_func` function example:
>>> def agg(agent: SerialableAgent, res: Dict[str, Any]) -> ActionModel:
>>> ...
>>> return ActionModel(agent_name=agent.id(), policy_info='...')
"""
def __init__(self,
agents: List[Agent] = None,
aggregate_func: Callable[['SerialableAgent', Dict[str, Any]], ActionModel] = None,
**kwargs):
super().__init__(**kwargs)
self.agents = agents if agents else []
self.aggregate_func = aggregate_func
async def async_policy(self, observation: Observation, info: Dict[str, Any] = {}, **kwargs) -> List[ActionModel]:
self.results = None
results = {}
action = ActionModel(agent_name=self.id(), policy_info=observation.content)
if self.agents:
for agent in self.agents:
result = await exec_agent(observation.content, agent, self.context, sub_task=True)
if result:
if result.success:
con = result.answer
else:
con = result.msg
action = ActionModel(agent_name=agent.id(), policy_info=con)
observation = self._action_to_observation(action, agent.id())
results[agent.id()] = con
else:
raise Exception(f"{agent.id()} execute fail.")
if self.aggregate_func:
return [self.aggregate_func(self, results)]
return [action]
def _action_to_observation(self, policy: ActionModel, agent_name: str):
if not policy:
logger.warning("no agent policy, will use default error info.")
return Observation(content=f"{agent_name} no policy")
logger.debug(f"{policy.policy_info}")
return Observation(content=policy.policy_info, observer=agent_name)
def finished(self) -> bool:
return all([agent.finished for agent in self.agents])
chapter9/gaia-experience/AWorld/aworld/agents/task_llm_agent.py
+47
View File
@@ -0,0 +1,47 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
from typing import List, Dict, Any
from aworld.core.exceptions import AWorldRuntimeException
from aworld.core.agent.swarm import Swarm
from aworld.core.task import Task, TaskResponse
from aworld.utils.run_util import exec_tasks
from aworld.agents.llm_agent import Agent
from aworld.core.common import Observation, ActionModel
class TaskAgent(Agent):
"""Support for swarm execution of in the hybrid nested swarm."""
def __init__(self,
swarm: Swarm,
**kwargs):
super().__init__(**kwargs)
self.swarm = swarm
if not self.swarm:
raise AWorldRuntimeException("no swarm in task agent.")
def reset(self, options: Dict[str, Any] = None):
super().reset(options)
if not options:
self.swarm.reset()
else:
self.swarm.reset(options.get("task"), options.get("context"), options.get("tools"))
async def async_policy(self, observation: Observation, info: Dict[str, Any] = {}, **kwargs) -> List[ActionModel]:
self._finished = False
task = Task(input=observation.content, swarm=self.swarm)
results = await exec_tasks([task])
res = []
for key, result in results.items():
# result is TaskResponse
if result.success:
info = result.answer
else:
info = result.msg
res.append(ActionModel(agent_name=self.id(), policy_info=info))
self._finished = True
return res
chapter9/gaia-experience/AWorld/aworld/checkpoint/README.md
+98
View File
@@ -0,0 +1,98 @@
# Checkpoint Module
## Overview
The Checkpoint module provides a robust and extensible framework for managing state snapshots (checkpoints) in Python applications. It is designed for scenarios where you need to persist, restore, and version the state of a process, session, or task.
```mermaid
sequenceDiagram
participant Application
participant CheckpointRepository
participant BackendStorage
Note over Application,BackendStorage: Create and store a checkpoint
%% Create and store a checkpoint
Application->>CheckpointRepository: create checkpoint
CheckpointRepository->>BackendStorage: put(checkpoint)
BackendStorage-->>CheckpointRepository: success
CheckpointRepository-->>Application: ack
Note over Application,BackendStorage: Retrieve the latest checkpoint by session
%% Retrieve the latest checkpoint by session
Application->>CheckpointRepository: get checkpoint by session_id
CheckpointRepository->>BackendStorage: get_by_session(session_id)
BackendStorage-->>CheckpointRepository: Checkpoint
CheckpointRepository-->>Application: Checkpoint
```
## Key Features
- **Structured Data Model**: Uses Pydantic's `BaseModel` for strong typing and validation of checkpoint data and metadata.
- **Versioning Support**: Built-in version management utilities for checkpoint evolution and comparison.
- **Extensible Repository Pattern**: Abstract base class (`BaseCheckpointRepository`) defines a standard interface for checkpoint storage, supporting both synchronous and asynchronous operations.
- **In-Memory Implementation**: Includes a simple, ready-to-use in-memory repository for development and testing.
- **Utility Functions**: Helper methods for creating, copying, and managing checkpoints.
## Data Structures
```mermaid
classDiagram
class Application {
+CheckpointRepository repo
+create_checkpoint()
+get_checkpoint_by_session()
}
class CheckpointRepository {
+put(checkpoint)
+get_by_session(session_id)
+delete_by_session(session_id)
-BackendStorage backend
}
class BackendStorage {
+put(checkpoint)
+get_by_session(session_id)
+delete_by_session(session_id)
}
Application --> CheckpointRepository : uses
CheckpointRepository --> BackendStorage : delegates
class Checkpoint {
+id: str
+ts: str
+metadata: CheckpointMetadata
+values: dict
+version: int
+parent_id: str
+namespace: str
}
class CheckpointMetadata {
+session_id: str
+task_id: str
}
Checkpoint o-- CheckpointMetadata
CheckpointRepository o-- Checkpoint
BackendStorage o-- Checkpoint
```
## Usage Example
```python
from aworld.checkpoint import (
Checkpoint, CheckpointMetadata, empty_checkpoint, create_checkpoint, InMemoryCheckpointRepository
)
# Create a new checkpoint
metadata = CheckpointMetadata(session_id="session-123", task_id="task-456")
values = {"step": 1, "score": 100}
checkpoint = create_checkpoint(values=values, metadata=metadata)
# Store and retrieve using the in-memory repository
repo = InMemoryCheckpointRepository()
repo.put(checkpoint)
restored = repo.get(checkpoint.id)
```
## Extensibility
- Implement custom repositories by inheriting from `BaseCheckpointRepository` (e.g., for database, file, or cloud storage).
- Extend versioning logic via the `VersionUtils` class.
chapter9/gaia-experience/AWorld/aworld/checkpoint/__init__.py
+245
View File
@@ -0,0 +1,245 @@
from typing import Any, Dict, Optional, List
import copy
import uuid
from datetime import datetime, timezone
from abc import ABC, abstractmethod
import asyncio
from pydantic import BaseModel, Field, ConfigDict
class CheckpointMetadata(BaseModel):
"""
Metadata for a checkpoint, including session and task identifiers.
Attributes:
session_id (str): The session identifier (required).
task_id (Optional[str]): The task identifier (optional).
artifact_id (Optional[str]): The artifact identifier (optional).
"""
session_id: str = Field(..., description="The session identifier.")
task_id: Optional[str] = Field(None, description="The task identifier.")
artifact_id: Optional[str] = Field(None, description="The artifact identifier.")
model_config = ConfigDict(extra="allow")
class Checkpoint(BaseModel):
"""
Core structure for a state checkpoint.
Attributes:
id (str): Unique identifier for the checkpoint.
ts (str): Timestamp of the checkpoint.
metadata (CheckpointMetadata): Metadata associated with the checkpoint.
values (dict[str, Any]): State values stored in the checkpoint.
version (str): Version of the checkpoint format.
parent_id (Optional[str]): Parent checkpoint identifier, if any.
namespace (str): Namespace for the checkpoint, default is 'aworld'.
"""
id: str = Field(..., description="Unique identifier for the checkpoint.")
ts: str = Field(..., description="Timestamp of the checkpoint.")
metadata: CheckpointMetadata = Field(..., description="Metadata associated with the checkpoint.")
values: Dict[str, Any] = Field(..., description="State values stored in the checkpoint.")
version: int = Field(..., description="Version of the checkpoint format.")
parent_id: Optional[str] = Field(default=None, description="Parent checkpoint identifier, if any.")
namespace: str = Field(default="aworld", description="Namespace for the checkpoint, default is 'aworld'.")
def empty_checkpoint() -> Checkpoint:
"""
Create an empty checkpoint with default values.
Returns:
Checkpoint: An empty checkpoint structure.
"""
return Checkpoint(
id=str(uuid.uuid4()),
ts=datetime.now(timezone.utc).isoformat(),
metadata=CheckpointMetadata(session_id="", task_id=None),
values={},
version=1,
parent_id=None,
namespace="aworld",
)
def copy_checkpoint(checkpoint: Checkpoint) -> Checkpoint:
"""
Create a deep copy of a checkpoint.
Args:
checkpoint (Checkpoint): The checkpoint to copy.
Returns:
Checkpoint: A deep copy of the provided checkpoint.
"""
return copy.deepcopy(checkpoint)
def create_checkpoint(
values: Dict[str, Any],
metadata: CheckpointMetadata,
parent_id: Optional[str] = None,
version: int = 1,
namespace: str = 'aworld',
) -> Checkpoint:
"""
Create a new checkpoint from provided state values and metadata.
Args:
values (dict[str, Any]): State values to store in the checkpoint.
metadata (CheckpointMetadata): Metadata for the checkpoint.
parent_id (Optional[str]): Parent checkpoint identifier, if any.
version (str): Version of the checkpoint format.
namespace (str): Namespace for the checkpoint.
Returns:
Checkpoint: The newly created checkpoint.
"""
return Checkpoint(
id=str(uuid.uuid4()),
ts=datetime.now(timezone.utc).isoformat(),
metadata=metadata,
values=values,
version=version,
parent_id=parent_id,
namespace=namespace,
)
class BaseCheckpointRepository(ABC):
"""
Abstract base class for a checkpoint repository.
Provides synchronous and asynchronous methods for checkpoint management.
"""
@abstractmethod
def get(self, checkpoint_id: str) -> Optional[Checkpoint]:
"""
Retrieve a checkpoint by its unique identifier.
Args:
checkpoint_id (str): The unique identifier of the checkpoint.
Returns:
Optional[Checkpoint]: The checkpoint if found, otherwise None.
"""
pass
@abstractmethod
def list(self, params: Dict[str, Any]) -> List[Checkpoint]:
"""
List checkpoints matching the given parameters.
Args:
params (dict): Parameters to filter checkpoints.
Returns:
List[Checkpoint]: List of matching checkpoints.
"""
pass
@abstractmethod
def put(self, checkpoint: Checkpoint) -> None:
"""
Store a checkpoint.
Args:
checkpoint (Checkpoint): The checkpoint to store.
"""
pass
@abstractmethod
def get_by_session(self, session_id: str) -> Optional[Checkpoint]:
"""
Get the latest checkpoint for a session.
Args:
session_id (str): The session identifier.
Returns:
Optional[Checkpoint]: The latest checkpoint if found, otherwise None.
"""
pass
@abstractmethod
def delete_by_session(self, session_id: str) -> None:
"""
Delete all checkpoints related to a session.
Args:
session_id (str): The session identifier.
"""
pass
# Async methods
async def aget(self, checkpoint_id: str) -> Optional[Checkpoint]:
"""
Asynchronously retrieve a checkpoint by its unique identifier.
Args:
checkpoint_id (str): The unique identifier of the checkpoint.
Returns:
Optional[Checkpoint]: The checkpoint if found, otherwise None.
"""
return await asyncio.to_thread(self.get, checkpoint_id)
async def alist(self, params: Dict[str, Any]) -> List[Checkpoint]:
"""
Asynchronously list checkpoints matching the given parameters.
Args:
params (dict): Parameters to filter checkpoints.
Returns:
List[Checkpoint]: List of matching checkpoints.
"""
return await asyncio.to_thread(self.list, params)
async def aput(self, checkpoint: Checkpoint) -> None:
"""
Asynchronously store a checkpoint.
Args:
checkpoint (Checkpoint): The checkpoint to store.
"""
await asyncio.to_thread(self.put, checkpoint)
async def aget_by_session(self, session_id: str) -> Optional[Checkpoint]:
"""
Asynchronously get the latest checkpoint for a session.
Args:
session_id (str): The session identifier.
Returns:
Optional[Checkpoint]: The latest checkpoint if found, otherwise None.
"""
return await asyncio.to_thread(self.get_by_session, session_id)
async def adelete_by_session(self, session_id: str) -> None:
"""
Asynchronously delete all checkpoints related to a session.
Args:
session_id (str): The session identifier.
"""
await asyncio.to_thread(self.delete_by_session, session_id)
class VersionUtils:
@staticmethod
def get_next_version(version: int) -> int:
"""
Get the next version of the checkpoint.
"""
return version + 1
@staticmethod
def get_previous_version(version: int) -> int:
"""
Get the previous version of the checkpoint.
"""
return version - 1
@staticmethod
def is_version_greater(checkpoint: Checkpoint, version: int) -> bool:
"""
Check if the checkpoint version is greater than the given version.
"""
return checkpoint.version > version
@staticmethod
def is_version_less(checkpoint: Checkpoint, version: int) -> bool:
"""
Check if the checkpoint version is less than the given version.
"""
return checkpoint.version < version
chapter9/gaia-experience/AWorld/aworld/checkpoint/inmemory.py
+116
View File
@@ -0,0 +1,116 @@
from typing import Any, Dict, List, Optional
from . import Checkpoint, BaseCheckpointRepository, VersionUtils
class InMemoryCheckpointRepository(BaseCheckpointRepository):
"""
In-memory implementation of BaseCheckpointRepository.
Stores checkpoints in a simple in-memory dictionary.
Thread safety is not guaranteed.
"""
def __init__(self) -> None:
"""
Initialize the in-memory checkpoint repository.
"""
self._checkpoints: Dict[str, Checkpoint] = {}
self._session_index: Dict[str, List[str]] = {}
def get(self, checkpoint_id: str) -> Optional[Checkpoint]:
"""
Retrieve a checkpoint by its unique identifier.
Args:
checkpoint_id (str): The unique identifier of the checkpoint.
Returns:
Optional[Checkpoint]: The checkpoint if found, otherwise None.
"""
return self._checkpoints.get(checkpoint_id)
def list(self, params: Dict[str, Any]) -> List[Checkpoint]:
"""
List checkpoints matching the given parameters.
Args:
params (dict): Parameters to filter checkpoints.
Returns:
List[Checkpoint]: List of matching checkpoints.
"""
result = []
for cp in self._checkpoints.values():
match = True
for k, v in params.items():
if k == 'session_id':
if cp.metadata.session_id != v:
match = False
break
elif k == 'task_id':
if cp.metadata.task_id != v:
match = False
break
elif cp.get(k) != v:
match = False
break
if match:
result.append(cp)
return result
def put(self, checkpoint: Checkpoint) -> None:
"""
Store a checkpoint.
Args:
checkpoint (Checkpoint): The checkpoint to store.
"""
# Find last version checkpoint by session_id
last_checkpoint = self.get_by_session(checkpoint.metadata.session_id)
if last_checkpoint:
# Compare versions to ensure optimistic locking
if VersionUtils.is_version_less(checkpoint, last_checkpoint.version):
raise ValueError(f"New checkpoint version {checkpoint.version} must be greater than last version {last_checkpoint.version}")
# Store the new checkpoint
self._checkpoints[checkpoint.id] = checkpoint
# Update session index
session_id = checkpoint.metadata.session_id
if session_id:
if session_id not in self._session_index:
self._session_index[session_id] = []
self._session_index[session_id].append(checkpoint.id)
def get_by_session(self, session_id: str) -> Optional[Checkpoint]:
"""
Get the latest checkpoint for a session.
Args:
session_id (str): The session identifier.
Returns:
Optional[Checkpoint]: The latest checkpoint if found, otherwise None.
"""
ids = self._session_index.get(session_id, [])
if not ids:
return None
# Assume the last one is the latest
last_id = ids[-1]
return self._checkpoints.get(last_id)
def delete_by_session(self, session_id: str) -> None:
"""
Delete all checkpoints related to a session.
Args:
session_id (str): The session identifier.
"""
ids = self._session_index.pop(session_id, [])
for cid in ids:
self._checkpoints.pop(cid, None)
async def alist(self, params: Dict[str, Any]) -> List[Checkpoint]:
return self.list(params)
async def aget(self, checkpoint_id: str) -> Optional[Checkpoint]:
return self.get(checkpoint_id)
async def aput(self, checkpoint: Checkpoint) -> None:
self.put(checkpoint)
async def aget_by_session(self, session_id: str) -> Optional[Checkpoint]:
return self.get_by_session(session_id)
async def adelete_by_session(self, session_id: str) -> None:
self.delete_by_session(session_id)
chapter9/gaia-experience/AWorld/aworld/cmd/__init__.py
+2
View File
@@ -0,0 +1,2 @@
# coding: utf-8
# Copyright (c) 2025 inclusionAI.
chapter9/gaia-experience/AWorld/aworld/cmd/cli.py
+38
View File
@@ -0,0 +1,38 @@
import click
@click.group()
def main(*args, **kwargs):
print(
"""\
AWorld CLI Help:
aworld web: run aworld web ui server
aworld api: run aworld api server
aworld help: show help"""
)
@main.command("web")
@click.option(
"--port", type=int, default=8000, help="Port to run the AWorld api server"
)
@click.argument("args", nargs=-1)
def main_web(port, args=None, **kwargs):
from .web import web_server
web_server.run_server(port, args, **kwargs)
@main.command("api")
@click.option(
"--port", type=int, default=8000, help="Port to run the AWorld api server"
)
@click.argument("args", nargs=-1)
def main_api(port, args=None, **kwargs):
from .web import api_server
api_server.run_server(port, args, **kwargs)
if __name__ == "__main__":
main()
chapter9/gaia-experience/AWorld/aworld/cmd/data_model.py
+78
View File
@@ -0,0 +1,78 @@
import datetime
import uuid
from abc import abstractmethod
from typing import Any, AsyncGenerator, List, Optional
from pydantic import BaseModel, Field
from aworld.output.base import Output
class ChatCompletionMessage(BaseModel):
role: str = Field(..., description="The role of the message")
content: str = Field(..., description="The content of the message")
trace_id: Optional[str] = Field(None, description="The trace id")
class ChatCompletionRequest(BaseModel):
user_id: Optional[str] = Field(None, description="The user id")
session_id: str = Field(
None,
description="The session id, if not provided, a new session will be created",
)
query_id: Optional[str] = Field(None, description="The query id")
trace_id: Optional[str] = Field(None, description="The trace id")
model: str = Field(..., description="The model to use")
messages: List[ChatCompletionMessage] = Field(
..., description="The messages to send to the agent"
)
class ChatCompletionChoice(BaseModel):
index: int = 0
delta: ChatCompletionMessage = Field(
..., description="The delta message from the agent"
)
class ChatCompletionResponse(BaseModel):
object: str = "chat.completion.chunk"
id: str = uuid.uuid4().hex
choices: List[ChatCompletionChoice] = Field(
..., description="The choices from the agent"
)
class AgentModel(BaseModel):
id: str = Field(..., description="The agent id")
name: Optional[str] = Field(None, description="The agent name")
description: Optional[str] = Field(None, description="The agent description")
path: str = Field(..., description="The agent path")
instance: Any = Field(..., description="The agent module instance", exclude=True)
class BaseAWorldAgent:
@abstractmethod
def name(self) -> str:
pass
@abstractmethod
def description(self) -> str:
pass
@abstractmethod
async def run(
self, prompt: str = None, request: ChatCompletionRequest = None
) -> AsyncGenerator[Output, None]:
pass
class SessionModel(BaseModel):
user_id: str = Field(..., description="The user id")
session_id: str = Field(..., description="The session id")
name: str = Field(None, description="The session name")
description: str = Field(None, description="The session description")
created_at: datetime.datetime = Field(None, description="The session created at")
updated_at: datetime.datetime = Field(None, description="The session updated at")
messages: List[ChatCompletionMessage] = Field(
None, description="The messages in the session"
)
chapter9/gaia-experience/AWorld/aworld/cmd/utils/__init__.py
View File
chapter9/gaia-experience/AWorld/aworld/cmd/utils/agent_executor.py
+87
View File
@@ -0,0 +1,87 @@
from typing import AsyncGenerator
from aworld.cmd.utils.agent_server import AgentServer
from aworld.output.ui.base import AworldUI
from aworld.output.workspace import WorkSpace
from aworld.cmd.data_model import (
BaseAWorldAgent,
ChatCompletionChoice,
ChatCompletionMessage,
ChatCompletionRequest,
ChatCompletionResponse,
)
from .agent_ui_parser import AWorldWebAgentUI
import logging
import os
import uuid
from dotenv import load_dotenv
import traceback
logger = logging.getLogger(__name__)
async def stream_run(request: ChatCompletionRequest, agent_server: AgentServer):
if not request.session_id:
request.session_id = str(uuid.uuid4())
if not request.query_id:
request.query_id = str(uuid.uuid4())
if request.messages and request.messages[-1].trace_id is None:
request.messages[-1].trace_id = request.trace_id
logger.info(f"Stream run agent: request={request.model_dump_json()}")
agent = agent_server.get_agent(request.model)
instance: BaseAWorldAgent = agent.instance
env_file = os.path.join(agent.path, ".env")
if os.path.exists(env_file):
logger.info(f"Loading environment variables from {env_file}")
load_dotenv(env_file, override=True, verbose=True)
final_response: str = ""
def build_response(delta_content: str):
nonlocal final_response
final_response += delta_content
logger.info(f"Agent {agent.name} response: {delta_content}")
return ChatCompletionResponse(
choices=[
ChatCompletionChoice(
index=0,
delta=ChatCompletionMessage(
role="assistant",
content=delta_content,
trace_id=request.trace_id,
),
)
]
)
rich_ui = AWorldWebAgentUI(
session_id=request.session_id,
workspace=WorkSpace.from_local_storages(
workspace_id=request.session_id,
),
)
await agent_server.on_chat_completion_request(request)
try:
async for output in instance.run(request=request):
try:
logger.info(f"Agent {agent.name} output: {output}")
if isinstance(output, str):
yield build_response(output)
else:
res = await AworldUI.parse_output(output, rich_ui)
for item in res if isinstance(res, list) else [res]:
if isinstance(item, AsyncGenerator):
async for sub_item in item:
yield build_response(sub_item)
else:
yield build_response(item)
except:
logger.error(
f"Agent {agent.name} output error! output={output}, error={traceback.format_exc()}"
)
except:
logger.error(f"Agent {agent.name} error: {traceback.format_exc()}")
finally:
await agent_server.on_chat_completion_end(request, final_response)
chapter9/gaia-experience/AWorld/aworld/cmd/utils/agent_loader.py
+112
View File
@@ -0,0 +1,112 @@
import os
import importlib
import subprocess
import sys
import traceback
import logging
from typing import List, Dict
from aworld.cmd.data_model import AgentModel
logger = logging.getLogger(__name__)
_agent_cache: Dict[str, AgentModel] = {}
def list_agents(server_dir: str) -> Dict[str, AgentModel]:
"""
List all cached agents
Returns:
Dict[str, AgentModel]: The map of agent models
"""
if len(_agent_cache) == 0:
for m in _list_agents(server_dir):
_agent_cache[m.id] = m
return _agent_cache
def _list_agents(server_dir: str) -> List[AgentModel]:
agents_dir = os.path.join(server_dir, "agent_deploy")
if not os.path.exists(agents_dir):
logger.warning(f"Agents directory {agents_dir} does not exist")
return []
if agents_dir not in sys.path:
sys.path.append(agents_dir)
agents = []
for agent_id in os.listdir(agents_dir):
if agent_id.startswith("_"):
continue
try:
agent_path = os.path.join(agents_dir, agent_id)
if os.path.isdir(agent_path):
requirements_file = os.path.join(agent_path, "requirements.txt")
if os.path.exists(requirements_file):
p = subprocess.Popen(
["pip", "install", "-U", "-r", requirements_file],
cwd=agent_path,
)
p.wait()
if p.returncode != 0:
logger.error(
f"Error installing requirements for agent {agent_id}, path {agent_path}"
)
continue
agent_file = os.path.join(agent_path, "agent.py")
if os.path.exists(agent_file):
try:
instance = _get_agent_instance(agent_id)
if hasattr(instance, "name"):
name = instance.name()
else:
name = agent_id
if hasattr(instance, "description"):
description = instance.description()
else:
description = ""
agent_model = AgentModel(
id=agent_id,
name=name,
description=description,
path=agent_path,
instance=instance,
)
agents.append(agent_model)
logger.info(
f"Loaded agent {agent_id} successfully, path {agent_path}"
)
except Exception as e:
logger.error(
f"Error loading agent {agent_id}: {traceback.format_exc()}"
)
continue
else:
logger.warning(f"Agent {agent_id} does not have agent.py file")
except Exception as e:
logger.error(
f"Error loading agent {agent_id}, path {agent_path} : {traceback.format_exc()}"
)
continue
return agents
def _get_agent_instance(agent_name):
try:
agent_module = importlib.import_module(
name=f"{agent_name}.agent",
)
except Exception as e:
msg = f"Error loading agent {agent_name}, cwd:{os.getcwd()}, sys.path:{sys.path}: {traceback.format_exc()}"
logger.error(msg)
raise Exception(msg)
if hasattr(agent_module, "AWorldAgent"):
agent = agent_module.AWorldAgent()
return agent
else:
raise Exception(f"Agent {agent_name} does not have AWorldAgent class")
chapter9/gaia-experience/AWorld/aworld/cmd/utils/agent_server.py
+116
View File
@@ -0,0 +1,116 @@
from abc import abstractmethod
import asyncio
from pathlib import Path
from typing import Dict, List
import os
from dotenv import load_dotenv
from aworld import trace
from aworld.cmd.data_model import (
AgentModel,
ChatCompletionMessage,
ChatCompletionRequest,
)
from aworld.session.base_session_service import BaseSessionService
from aworld.session.simple_session_service import SimpleSessionService
from . import agent_loader
from aworld.trace.config import ObservabilityConfig
from aworld.trace.opentelemetry.memory_storage import InMemoryWithPersistStorage
# bugfix for tracer exception
trace.configure(ObservabilityConfig(trace_storage=(InMemoryWithPersistStorage())))
class ChatCallBack:
@abstractmethod
async def on_chat_completion_request(
self, server: "AgentServer", request: ChatCompletionRequest
):
pass
@abstractmethod
async def on_chat_completion_end(
self, server: "AgentServer", request: ChatCompletionRequest, final_response: str
):
pass
class SessionChatCallBack(ChatCallBack):
async def on_chat_completion_request(
self, server: "AgentServer", request: ChatCompletionRequest
):
await server.get_session_service().append_messages(
request.user_id,
request.session_id,
request.messages[-1:],
)
async def on_chat_completion_end(
self, server: "AgentServer", request: ChatCompletionRequest, final_response: str
):
await server.get_session_service().append_messages(
request.user_id,
request.session_id,
[
ChatCompletionMessage(
role="assistant",
content=final_response,
trace_id=request.trace_id,
),
],
)
class AgentServer:
server_id: str
server_name: str
server_dir: str
session_service: BaseSessionService = None
agent_instances: Dict[str, AgentModel] = {}
chat_call_backs: List[ChatCallBack] = [SessionChatCallBack()]
def __init__(
self,
server_id: str,
server_name: str,
server_dir: str = os.getcwd(),
session_service: BaseSessionService = SimpleSessionService(),
):
"""
Initialize AgentServer
"""
self.server_id = server_id
self.server_name = server_name
self.server_dir = server_dir
self.session_service = session_service
# Load server global env
load_dotenv(Path(self.server_dir) / ".env", override=True, verbose=True)
# Load agent instances
self.agent_instances = agent_loader.list_agents(self.server_dir)
def list_agents(self) -> Dict[str, AgentModel]:
return self.agent_instances
def get_agent(self, agent_id: str) -> AgentModel:
return self.agent_instances.get(agent_id)
def get_session_service(self) -> BaseSessionService:
return self.session_service
async def on_chat_completion_request(self, request: ChatCompletionRequest):
tasks = []
for chat_call_back in self.chat_call_backs:
tasks.append(chat_call_back.on_chat_completion_request(self, request))
await asyncio.gather(*tasks)
async def on_chat_completion_end(
self, request: ChatCompletionRequest, final_response: str
):
tasks = []
for chat_call_back in self.chat_call_backs:
tasks.append(
chat_call_back.on_chat_completion_end(self, request, final_response)
)
await asyncio.gather(*tasks)
chapter9/gaia-experience/AWorld/aworld/cmd/utils/agent_ui_parser.py
+253
View File
@@ -0,0 +1,253 @@
import json
from dataclasses import dataclass
import uuid
from pydantic import Field, BaseModel, ConfigDict
from aworld.output import (
MessageOutput,
AworldUI,
Output,
WorkSpace,
)
from aworld.output.artifact import Artifact, ArtifactType
from aworld.output.base import StepOutput, ToolResultOutput
from aworld.output.utils import consume_content
from abc import ABC, abstractmethod
from typing_extensions import override
class ToolCard(BaseModel):
model_config = ConfigDict(extra="forbid")
tool_type: str = Field(None, description="tool type")
tool_name: str = Field(None, description="tool name")
function_name: str = Field(None, description="function name")
tool_call_id: str = Field(None, description="tool call id")
arguments: str = Field(None, description="arguments")
results: str = Field(None, description="results")
card_type: str = Field(None, description="card type")
card_data: dict = Field(None, description="card data")
artifacts: list = Field(default_factory=list, description="artifacts")
@staticmethod
def from_tool_result(output: ToolResultOutput) -> "ToolCard":
return ToolCard(
tool_type=output.tool_type,
tool_name=output.tool_name,
function_name=output.origin_tool_call.function.name,
tool_call_id=output.origin_tool_call.id,
arguments=output.origin_tool_call.function.arguments,
results=output.data,
artifacts=[],
)
class BaseToolResultParser(ABC):
def __init__(self, tool_name: str = None):
self.tool_name = tool_name or self.__class__.__name__
@abstractmethod
async def parse(self, output: ToolResultOutput, workspace: WorkSpace):
pass
class DefaultToolResultParser(BaseToolResultParser):
@override
async def parse(self, output: ToolResultOutput, workspace: WorkSpace):
tool_card = ToolCard.from_tool_result(output)
tool_card.card_type = "tool_call_card_default"
# screenshots
if (
output.metadata.get("screenshots")
and isinstance(output.metadata.get("screenshots"), list)
and len(output.metadata.get("screenshots")) > 0
):
for _, screenshot in enumerate(output.metadata.get("screenshots")):
image_artifact = Artifact(
artifact_id=str(uuid.uuid4()),
artifact_type=ArtifactType.IMAGE,
content=screenshot.get("ossPath"),
)
await workspace.add_artifact(image_artifact)
tool_card.artifacts.append(
{
"artifact_type": image_artifact.artifact_type.value,
"artifact_id": image_artifact.artifact_id,
}
)
return f"""\
\n\n**🔧 Tool: {tool_card.tool_name}#{tool_card.function_name}**\n\n
```tool_card
{json.dumps(tool_card.model_dump(), ensure_ascii=False, indent=2)}
```\n
"""
class SearchToolResultParser(BaseToolResultParser):
@override
async def parse(self, output: ToolResultOutput, workspace: WorkSpace):
tool_card = ToolCard.from_tool_result(output)
query = ""
try:
args = json.loads(tool_card.arguments)
query = args.get("query")
# aworld search server
if not query:
query = args.get("query_list")
except Exception:
pass
result_items = []
try:
result_items = json.loads(tool_card.results)
# aworld search server return url, not link
if result_items and isinstance(result_items, list):
for item in result_items:
if not item.get("link", None) and item.get("url", None):
item["link"] = item.get("url")
except Exception:
pass
if len(result_items) > 0:
tool_card.results = ""
tool_card.card_type = "tool_call_card_link_list"
tool_card.card_data = {
"title": "🔎 Google Search",
"query": query,
"search_items": result_items,
}
artifact_id = str(uuid.uuid4())
await workspace.create_artifact(
artifact_type=ArtifactType.WEB_PAGES,
artifact_id=artifact_id,
content=result_items,
metadata={
"query": query,
},
)
tool_card.artifacts.append(
{
"artifact_type": ArtifactType.WEB_PAGES.value,
"artifact_id": artifact_id,
}
)
return f"""\
\n\n**🔎 Search Results**\n\n
```tool_card
{json.dumps(tool_card.model_dump(), ensure_ascii=False, indent=2)}
```\n
"""
class ToolResultParserFactory:
def get_parser(self, tool_type: str, tool_name: str):
if "search" in tool_name and ("search" in tool_name or tool_name == None):
return SearchToolResultParser()
else:
return DefaultToolResultParser()
@dataclass
class AWorldWebAgentUI(AworldUI):
session_id: str = Field(default="", description="session id")
workspace: WorkSpace = Field(default=None, description="workspace")
tool_result_parser_factory: ToolResultParserFactory = Field(
default=ToolResultParserFactory, description="tool result parser factory"
)
def __init__(
self,
session_id: str = None,
workspace: WorkSpace = None,
tool_result_parser_factory: ToolResultParserFactory = None,
**kwargs,
):
"""
Initialize MarkdownAworldUI
Args:"""
super().__init__(**kwargs)
self.session_id = session_id
self.workspace = workspace
self.tool_result_parser_factory = (
tool_result_parser_factory or ToolResultParserFactory()
)
@override
async def message_output(self, __output__: MessageOutput):
"""
Returns an async generator that yields each message item.
"""
# Sentinel object for queue completion
_SENTINEL = object()
async def async_generator():
async def __log_item(item):
await queue.put(item)
from asyncio import Queue
queue = Queue()
async def consume_all():
# Consume all relevant generators
if __output__.reason_generator or __output__.response_generator:
if __output__.reason_generator:
await consume_content(__output__.reason_generator, __log_item)
if __output__.response_generator:
await consume_content(__output__.response_generator, __log_item)
else:
await consume_content(__output__.reasoning, __log_item)
await consume_content(__output__.response, __log_item)
# Only after all are done, put the sentinel
await queue.put(_SENTINEL)
# Start the consumer in the background
import asyncio
consumer_task = asyncio.create_task(consume_all())
while True:
item = await queue.get()
if item is _SENTINEL:
break
yield item
await consumer_task # Ensure background task is finished
return async_generator()
@override
async def tool_result(self, output: ToolResultOutput):
"""
tool_result
"""
parser = self.tool_result_parser_factory.get_parser(
output.tool_type, output.tool_name
)
return await parser.parse(output, workspace=self.workspace)
@override
async def step(self, output: StepOutput):
emptyLine = "\n\n"
if output.status == "START":
return f"\n\n # {output.show_name} \n\n"
elif output.status == "FINISHED":
return f"{emptyLine}"
elif output.status == "FAILED":
return f"\n\n{output.name} 💥FAILED: reason is {output.data} {emptyLine}"
else:
return f"\n\n{output.name} ❓❓❓UNKNOWN#{output.status} {emptyLine}"
@override
async def custom_output(self, output: Output):
return output.data
chapter9/gaia-experience/AWorld/aworld/cmd/utils/trace_summarize.py
+175
View File
@@ -0,0 +1,175 @@
import os
import logging
import traceback
import asyncio
import re
import json
import pickle
from asyncio.tasks import Task
from aworld.config.conf import AgentConfig
from aworld.agents.llm_agent import Agent
from typing import Dict, Union
from aworld.core.context.base import Context
from aworld.utils.run_util import exec_agent
logger = logging.getLogger(__name__)
class SimpleSummaryCache:
def __init__(self) -> None:
self._cache_file = os.path.join(os.curdir, "data", "trace_summary_cache.pkl")
self._cache: Dict[str, str] = {}
self._load_cache()
def _load_cache(self):
if os.path.exists(self._cache_file):
try:
with open(self._cache_file, "rb") as f:
self._cache = pickle.load(f)
except (pickle.PickleError, EOFError):
logger.warning("Cache file is corrupted, creating new cache")
if self._cache_file.exists():
self._cache_file.unlink()
def _save_cache(self):
serializable_cache = {
k: v for k, v in self._cache.items() if not isinstance(v, Task)
}
try:
with open(self._cache_file, "wb") as f:
pickle.dump(serializable_cache, f)
except pickle.PickleError:
logger.error("Failed to save cache")
def add_to_cache(self, trace_id: str, value: Union[str, Task]):
self._cache[trace_id] = value
if not isinstance(value, Task):
self._save_cache()
def get_value(self, trace_id: str) -> Union[str, Task]:
return self._cache.get(trace_id)
def trace_exists(self, trace_id: str) -> bool:
return trace_id in self._cache
# _trace_summary_cache: Dict[str, Union[str, Task]] = {}
_trace_summary_cache = SimpleSummaryCache()
trace_sys_prompt = "You are a helpful tracking summary agent."
trace_prompt = """
you can use tracking tools to obtain tracking data and then summarize the main tasks completed by each agent and their token usage.
You can identify which spans are agents, which spans are tool calls, and which spans are large model calls based on the following criteria:
1 Agent span: the prefix for 'name' is 'event.agent.'
2 LLM span: The prefix for 'name' is 'llm.'
3 Tool span: The prefix for 'name' is 'event.tool.'
requirement:
1. Please summarize and output separately for agents with different event.id.
2. Agent Span with the same name but different event.id are also considered as different agents.
3. There may be a parent-child relationship between agents. Please select the LLM span and Tool span from the nearest child span to the current agent for summarizing.
4. Ensure that all agent spans have their own independent summaries, and the number of summaries is exactly the same as the number of agent spans. For example: {{"name":"event.agent.a","attributes":{{"event.id":"111"}},"children":[{{"name":"llm.gpt-4o"}},{{"name":"event.tool.1","children":[{{"name":"event.agent.a","attributes":{{"event.id":"222"}},"children":[{{"name":"llm.gpt-4o"}}]}}]}}]}}, both of the above two agent names are event.agent.a, but event.id is different and needs to be summarized separately for 111 and 222. So you need to identify all agent spans without any omissions, which is very important.
5. Please output in the following standard JSON format without any additional explanatory text:
[{{"agent":"947cc4c1b7ed406ab7fbf38b9d2b1f5a",,"summary":"xxx"}},{{}}]
6. Pay attention to controlling the length of the summary, so that the overall output does not exceed your output length limit.
Here are the trace_id: {task}
"""
agent_config = None
async def _do_summarize_trace(trace_id: str):
logger.info(f"_do_summarize_trace trace_id: {trace_id}")
global agent_config
trace_agent = Agent(
conf=agent_config,
name="trace_agent",
system_prompt=trace_sys_prompt,
agent_prompt=trace_prompt,
tool_names=["trace"],
feedback_tool_result=True,
)
if trace_agent.conf.llm_config.llm_api_key is None:
logger.warning(
"LLM_API_KEY_TRACE is not set, trace summarize will not be executed."
)
return ""
try:
res = await exec_agent(trace_id, trace_agent, Context())
summary = _fetch_json_from_result(res.answer)
_trace_summary_cache.add_to_cache(trace_id, summary)
return summary
except Exception as e:
logger.error(traceback.format_exc())
def summarize_trace(trace_id: str):
global agent_config
if agent_config is None:
llm_provider = os.getenv("LLM_PROVIDER_TRACE", "openai")
llm_model_name = os.getenv("LLM_MODEL_NAME_TRACE", None)
llm_base_url = os.getenv("LLM_BASE_URL_TRACE", None)
llm_api_key = os.getenv("LLM_API_KEY_TRACE", None)
if (
not llm_provider
or not llm_model_name
or not llm_base_url
or not llm_api_key
):
logger.warning(
"LLM_MODEL_NAME_TRACE, LLM_BASE_URL_TRACE, LLM_API_KEY_TRACE is not set, trace summarize will not be executed."
)
return
agent_config = AgentConfig(
llm_provider=os.getenv("LLM_PROVIDER_TRACE", "openai"),
llm_model_name=os.getenv("LLM_MODEL_NAME_TRACE", None),
llm_base_url=os.getenv("LLM_BASE_URL_TRACE", None),
llm_api_key=os.getenv("LLM_API_KEY_TRACE", None),
)
llm_config = agent_config.llm_config
if not _trace_summary_cache.trace_exists(trace_id):
if (
llm_config.llm_api_key is None
or not llm_config.llm_base_url
or not llm_config.llm_model_name
):
logger.warning(
"LLM_MODEL_NAME_TRACE, LLM_BASE_URL_TRACE, LLM_API_KEY_TRACE is not set, trace summarize will not be executed."
)
return
task = asyncio.create_task(_do_summarize_trace(trace_id))
_trace_summary_cache.add_to_cache(trace_id, task)
async def get_summarize_trace(trace_id: str):
if not _trace_summary_cache.trace_exists(trace_id):
return None
cached_value = _trace_summary_cache.get_value(trace_id)
if isinstance(cached_value, Task):
# try:
# result = await cached_value
# if isinstance(result, Task):
# result = await result
# _trace_summary_cache[trace_id] = _fetch_json_from_result(result)
# except Exception as e:
# logger.error(traceback.format_exc())
return None
return cached_value
def _fetch_json_from_result(input_str):
json_match = re.search(r"\[.*\]", input_str, re.DOTALL)
if json_match:
json_str = json_match.group(0)
try:
json.loads(json_str)
return json_str
except json.JSONDecodeError as e:
logger.warning(f"_fetch_json_from_result json_str: {json_str} error: {e}")
return ""
chapter9/gaia-experience/AWorld/aworld/cmd/utils/webui_builder.py
+26
View File
@@ -0,0 +1,26 @@
import subprocess
import logging
from pathlib import Path
import sys
logger = logging.getLogger(__name__)
def build_webui(force_rebuild: bool = False) -> str:
webui_path = Path(__file__).parent.parent / "web" / "webui"
static_path = webui_path / "dist"
if (not static_path.exists()) or force_rebuild:
logger.warning(f"Build WebUI at {webui_path}")
try:
subprocess.check_call(
["sh", "-c", "npm install && npm run build"],
cwd=webui_path,
)
logger.info("WebUI build successfully")
except:
logger.error(f"Failed to build WebUI at {webui_path}")
sys.exit(1)
return static_path
chapter9/gaia-experience/AWorld/aworld/cmd/web/__init__.py
View File
chapter9/gaia-experience/AWorld/aworld/cmd/web/api_server.py
+32
View File
@@ -0,0 +1,32 @@
import logging
from fastapi import FastAPI
import uvicorn
from aworld.cmd.utils.agent_server import AgentServer
logger = logging.getLogger(__name__)
app = FastAPI()
agent_server = AgentServer(
server_id="default_server",
server_name="default_server",
)
app.state.agent_server = agent_server
from .routers import chats, workspaces, sessions
app.include_router(chats.router, prefix=chats.prefix)
app.include_router(workspaces.router, prefix=workspaces.prefix)
app.include_router(sessions.router, prefix=sessions.prefix)
def run_server(port, args=None, **kwargs):
logger.info(f"Running API server on port {port}")
uvicorn.run(
app,
host="0.0.0.0",
port=port,
)
chapter9/gaia-experience/AWorld/aworld/cmd/web/package-lock.json
Generated
+6
View File
@@ -0,0 +1,6 @@
{
"name": "web",
"lockfileVersion": 3,
"requires": true,
"packages": {}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/routers/__init__.py
View File
chapter9/gaia-experience/AWorld/aworld/cmd/web/routers/chats.py
+52
View File
@@ -0,0 +1,52 @@
import logging
import json
from typing import Dict
from fastapi import APIRouter, Depends, Request
from fastapi.responses import StreamingResponse
from aworld.cmd.data_model import AgentModel, ChatCompletionRequest
from aworld.cmd.utils import agent_executor
from aworld.cmd.utils.trace_summarize import summarize_trace
from aworld.cmd.web.utils.users import get_user_id_from_jwt
import aworld.trace as trace
logger = logging.getLogger(__name__)
router = APIRouter()
prefix = "/api/agent"
@router.get("/list")
@router.get("/models")
async def list_agents(request: Request) -> Dict[str, AgentModel]:
return request.app.state.agent_server.list_agents()
@router.post("/chat/completions")
async def chat_completion(
form_data: ChatCompletionRequest,
request: Request,
user_id: str = Depends(get_user_id_from_jwt),
) -> StreamingResponse:
# Set user_id from JWT to form_data
form_data.user_id = user_id
async def generate_stream():
async with trace.span(
"/chat/chat_completion", attributes={"model": form_data.model}
) as span:
form_data.trace_id = span.get_trace_id()
async for chunk in agent_executor.stream_run(
form_data, request.app.state.agent_server
):
yield f"data: {json.dumps(chunk.model_dump(), ensure_ascii=False)}\n\n"
summarize_trace(form_data.trace_id)
return StreamingResponse(
generate_stream(),
media_type="text/event-stream",
headers={
"Cache-Control": "no-cache",
"Connection": "keep-alive",
},
)
chapter9/gaia-experience/AWorld/aworld/cmd/web/routers/sessions.py
+52
View File
@@ -0,0 +1,52 @@
import logging
from typing import List
from fastapi import APIRouter, Depends, Request
from pydantic import BaseModel, Field
from aworld.cmd.data_model import SessionModel
from aworld.cmd.web.utils.users import get_user_id_from_jwt
logger = logging.getLogger(__name__)
router = APIRouter()
prefix = "/api/session"
@router.get("/list")
async def list_sessions(
request: Request,
user_id: str = Depends(get_user_id_from_jwt),
) -> List[SessionModel]:
return await request.app.state.agent_server.get_session_service().list_sessions(
user_id
)
class CommonResponse(BaseModel):
code: int = Field(..., description="The code")
message: str = Field(..., description="The message")
@staticmethod
def success(message: str = "success"):
return CommonResponse(code=0, message=message)
@staticmethod
def error(message: str):
return CommonResponse(code=1, message=message)
class DeleteSessionRequest(BaseModel):
session_id: str = Field(..., description="The session id")
@router.post("/delete")
async def delete_session(
request: DeleteSessionRequest, user_id: str = Depends(get_user_id_from_jwt)
) -> CommonResponse:
try:
await request.app.state.agent_server.get_session_service().delete_session(
user_id, request.session_id
)
return CommonResponse.success()
except Exception as e:
return CommonResponse.error(str(e))
chapter9/gaia-experience/AWorld/aworld/cmd/web/routers/traces.py
+50
View File
@@ -0,0 +1,50 @@
import json
import logging
from fastapi import APIRouter
from aworld.trace.server import get_trace_server
from aworld.trace.server.util import build_trace_tree, get_agent_flow
from aworld.cmd.utils.trace_summarize import get_summarize_trace
logger = logging.getLogger(__name__)
router = APIRouter()
prefix = "/api/trace"
@router.get("/list")
async def list_traces():
storage = get_trace_server().get_storage()
trace_data = []
for trace_id in storage.get_all_traces():
spans = storage.get_all_spans(trace_id)
spans_sorted = sorted(spans, key=lambda x: x.start_time)
trace_tree = build_trace_tree(spans_sorted)
trace_data.append({
'trace_id': trace_id,
'root_span': trace_tree,
})
return {
"data": trace_data
}
@router.get("/agent")
async def get_agent_trace(trace_id: str):
data = get_agent_flow(trace_id)
await _add_trace_summary(trace_id, data.get('nodes'))
return data
async def _add_trace_summary(trace_id, spans):
summary = await get_summarize_trace(trace_id)
json_summary_dict = {}
if summary:
json_summary = json.loads(summary)
json_summary_dict = {item['agent']: json.dumps(
item) for item in json_summary}
for span in spans:
if summary and "event_id" in span:
span['summary'] = json_summary_dict.get(span['event_id'])
span['attributes'] = None
chapter9/gaia-experience/AWorld/aworld/cmd/web/routers/workspaces.py
+74
View File
@@ -0,0 +1,74 @@
import logging
import os
from typing import List, Optional
from pydantic import BaseModel
from fastapi import APIRouter, HTTPException, status, Query, Body
from aworld.output import WorkSpace, ArtifactType
from aworld.output.utils import load_workspace
router = APIRouter()
prefix = "/api/workspaces"
@router.get("/{workspace_id}/tree")
async def get_workspace_tree(workspace_id: str):
logging.info(f"get_workspace_tree: {workspace_id}")
workspace = await get_workspace(workspace_id)
return workspace.generate_tree_data()
class ArtifactRequest(BaseModel):
artifact_ids: Optional[List[str]] = None
artifact_types: Optional[List[str]] = None
@router.post("/{workspace_id}/artifacts")
async def get_workspace_artifacts(workspace_id: str, request: ArtifactRequest):
"""
Get artifacts by workspace id and filter by a list of artifact types.
Args:
workspace_id: Workspace ID
request: Request body containing optional artifact_types list
Returns:
Dict with filtered artifacts
"""
artifact_types = request.artifact_types
if artifact_types:
# Validate all types
invalid_types = [t for t in artifact_types if t not in ArtifactType.__members__]
if invalid_types:
logging.error(f"Invalid artifact_types: {invalid_types}")
raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST,
detail=f"Invalid artifact types: {invalid_types}")
logging.info(f"Fetching artifacts of types: {artifact_types}")
else:
logging.info(f"Fetching all artifacts (no type filter)")
workspace = await get_workspace(workspace_id)
all_artifacts = workspace.list_artifacts()
filtered_artifacts = all_artifacts
if request.artifact_ids:
filtered_artifacts = [a for a in filtered_artifacts if a.artifact_id in request.artifact_ids]
if artifact_types:
filtered_artifacts = [a for a in filtered_artifacts if a.artifact_type.name in artifact_types]
return {
"data": filtered_artifacts
}
@router.get("/{workspace_id}/file/{artifact_id}/content")
async def get_workspace_file_content(workspace_id: str, artifact_id: str):
logging.info(f"get_workspace_file_content: {workspace_id}, {artifact_id}")
workspace = await get_workspace(workspace_id)
return {
"data": workspace.get_file_content_by_artifact_id(artifact_id)
}
async def get_workspace(workspace_id: str) -> WorkSpace:
workspace_type = os.environ.get("WORKSPACE_TYPE", "local")
workspace_path = os.environ.get("WORKSPACE_PATH", "./data/workspaces")
return await load_workspace(workspace_id, workspace_type, workspace_path)
chapter9/gaia-experience/AWorld/aworld/cmd/web/utils/__init__.py
+1
View File
@@ -0,0 +1 @@
chapter9/gaia-experience/AWorld/aworld/cmd/web/utils/users.py
+4
View File
@@ -0,0 +1,4 @@
from fastapi import Request
def get_user_id_from_jwt(request: Request) -> str:
return f"default_user_001"
chapter9/gaia-experience/AWorld/aworld/cmd/web/web_server.py
+62
View File
@@ -0,0 +1,62 @@
import asyncio
import logging
from fastapi import FastAPI, Request, Response
from fastapi.responses import RedirectResponse
import uvicorn
import os
from fastapi.staticfiles import StaticFiles
from starlette.middleware.base import BaseHTTPMiddleware
from aworld.cmd.utils.agent_server import AgentServer
from aworld.cmd.utils.webui_builder import build_webui
logger = logging.getLogger(__name__)
app = FastAPI()
@app.get("/")
async def root():
return RedirectResponse("/index.html")
agent_server = AgentServer(
server_id="default_server",
server_name="default_server",
)
app.state.agent_server = agent_server
from .routers import chats, workspaces, sessions, traces # noqa
app.include_router(chats.router, prefix=chats.prefix)
app.include_router(workspaces.router, prefix=workspaces.prefix)
app.include_router(sessions.router, prefix=sessions.prefix)
app.include_router(traces.router, prefix=traces.prefix)
static_path = build_webui(force_rebuild=os.getenv("AWORLD_WEB_UI_FORCE_REBUILD", False))
logger.info(f"Mounting static files from {static_path}")
app.mount("/", StaticFiles(directory=static_path, html=True), name="static")
class TimeoutMiddleware(BaseHTTPMiddleware):
def __init__(self, app, timeout: int = 300):
super().__init__(app)
self.timeout = timeout
async def dispatch(self, request: Request, call_next):
try:
return await asyncio.wait_for(call_next(request), timeout=self.timeout)
except asyncio.TimeoutError:
return Response("Request timeout", status_code=408)
app.add_middleware(TimeoutMiddleware, timeout=300)
def run_server(port, args=None, **kwargs):
logger.info(f"Running Web server on port {port}")
uvicorn.run(
app,
host="0.0.0.0",
port=port,
)
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/README.md
+1
View File
@@ -0,0 +1 @@
Front End Code Here
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/App-Dp0UqpP7.css
Vendored
+1
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/App-NnVSuaee.js
Vendored
+261
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/aworld_logo-DnLSZeN8.png
Vendored
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 36 KiB

chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/index-C7nkBYbk.js
Vendored
+68
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/index-CcSzV6ZK.js
Vendored
+281
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/index-DVzAU7Zt.js
Vendored
+85
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/index-I4DyY5Gh.css
Vendored
+1
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/index-TZrNw7dA.css
Vendored
+1
View File
@@ -0,0 +1 @@
body{margin:0}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/assets/index-n9x2b4FL.css
Vendored
+1
View File
File diff suppressed because one or more lines are too long
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/aworld_logo.png
Vendored
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 36 KiB

chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/index.html
Vendored
+14
View File
@@ -0,0 +1,14 @@
<!doctype html>
<html lang="en">
<head>
<meta charset="UTF-8" />
<link rel="icon" type="image/svg+xml" href="/aworld_logo.png" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>Aworld</title>
<script type="module" crossorigin src="/assets/index-C7nkBYbk.js"></script>
<link rel="stylesheet" crossorigin href="/assets/index-TZrNw7dA.css">
</head>
<body>
<div id="root"></div>
</body>
</html>
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/dist/trace_ui.html
Vendored
+526
View File
@@ -0,0 +1,526 @@
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Trace Viewer V2</title>
<link rel="stylesheet" href="https://unpkg.com/element-plus/dist/index.css">
<script src="https://unpkg.com/vue@3/dist/vue.global.js"></script>
<script src="https://unpkg.com/element-plus"></script>
<script src="https://unpkg.com/@element-plus/icons-vue"></script>
<script src="https://d3js.org/d3.v7.min.js"></script>
<style>
.trace-container {
display: flex;
flex-direction: column;
height: 100vh;
font-family: 'Helvetica Neue', Arial, sans-serif;
}
.trace-content {
display: flex;
flex: 1;
overflow: hidden;
}
.trace-list {
width: 30%;
overflow-y: auto;
border-right: 1px solid #e6e6e6;
}
.trace-detail {
width: 70%;
padding: 20px;
overflow-y: auto;
}
.timeline {
height: 120px;
min-width: 100%;
background: #f5f5f5;
padding: 10px;
border-bottom: 1px solid #e6e6e6;
}
.span-node {
cursor: pointer;
padding: 5px 0;
}
.span-node:hover {
background-color: #f0f7ff;
}
.span-duration {
color: #666;
font-size: 12px;
}
.timeline-bg {
fill: #f8f8f8;
}
.axis--x path {
stroke: #333;
stroke-width: 1px;
}
.axis--x line {
stroke: #ddd;
}
.axis--x text {
font-size: 12px;
fill: #333;
}
.timeline-visualization {
flex: 1;
padding: 20px;
background: #f8f8f8;
border-left: 1px solid #e6e6e6;
overflow-y: auto;
position: relative;
height: 100%;
}
.span-visualization-container {
position: relative;
height: 100%;
margin-top: 40px;
}
.span-visualization {
height: 20px;
background: #409EFF;
position: absolute;
margin-top: 2px;
border-radius: 2px;
}
.span-label {
font-size: 8px;
color: white;
padding: 0 5px;
white-space: nowrap;
overflow: hidden;
text-overflow: ellipsis;
}
.trace-timeline {
background: #f5f5f5;
padding: 10px;
border-radius: 4px;
}
.trace-timeline svg {
display: block;
}
.trace-timeline .axis path {
stroke: #333;
stroke-width: 1px;
}
.trace-timeline .axis line {
stroke: #ddd;
}
.trace-timeline .axis text {
font-size: 12px;
fill: #333;
}
</style>
</head>
<body>
<div id="app" class="trace-container">
<!-- Top timeline -->
<div class="timeline">
<div id="timeline-chart"></div>
</div>
<div class="trace-content">
<div class="trace-list">
<div style="padding: 10px; border-bottom: 1px solid #e6e6e6;">
<el-input v-model="searchTraceId" placeholder="输入Trace ID搜索" style="width: 100%;"
@keyup.enter="searchByTraceId">
<template #append>
<el-button @click="searchByTraceId">
<el-icon>
<search />
</el-icon>
</el-button>
</template>
</el-input>
</div>
<el-tree :data="traceTree" node-key="span_id" :props="treeProps" :expand-on-click-node="false"
@node-click="handleNodeClick" :default-expanded-keys="expandedNodes">
<template #default="{ node, data }">
<span class="span-node">
{{ data.name }}
<span class="span-duration">({{ data.duration_ms.toFixed(2) }}ms)</span>
</span>
</template>
</el-tree>
</div>
<div class="timeline-visualization" v-if="selectedSpan" v-html="renderTimelineVisualization()">
</div>
</div>
<!-- Span detail -->
<el-dialog v-model="dialogVisible" title="Span Details" width="70%">
<el-descriptions :column="2" border>
<el-descriptions-item label="Trace ID">{{ selectedSpan.trace_id }}</el-descriptions-item>
<el-descriptions-item label="Span ID">{{ selectedSpan.span_id }}</el-descriptions-item>
<el-descriptions-item label="Parent Span ID">{{ selectedSpan.parent_id || 'None'
}}</el-descriptions-item>
<el-descriptions-item label="Name">{{ selectedSpan.name }}</el-descriptions-item>
<el-descriptions-item label="Status">
<div style="display: flex; justify-content: space-between; align-items: center;">
<span :style="{color: selectedSpan.status.code === 'StatusCode.ERROR' ? '#F56C6C' : ''}">
{{ selectedSpan.status.code }}
</span>
<el-button v-if="selectedSpan.status.code === 'StatusCode.ERROR'" type="text" size="small"
@click="showStacktrace = true" icon="View" style="color: #F56C6C">
View Stack
</el-button>
</div>
</el-descriptions-item>
<el-descriptions-item label="Start Time">{{ selectedSpan.start_time}}</el-descriptions-item>
<el-descriptions-item label="End Time">{{ selectedSpan.end_time }}</el-descriptions-item>
<el-descriptions-item label="Duration">{{ selectedSpan.duration_ms.toFixed(2) }}
ms</el-descriptions-item>
</el-descriptions>
<el-card style="margin-top: 20px;">
<template #header>
<h4>Attributes</h4>
</template>
<pre style="
max-height: 400px;
overflow: auto;
white-space: pre-wrap;
word-break: break-all;
background: #f8f8f8;
padding: 10px;
border-radius: 4px;
">{{ formatAttributes(selectedSpan.attributes) }}</pre>
</el-card>
</el-dialog>
<el-dialog v-model="showStacktrace" title="Stacktrace Details" width="70%">
<pre>{{ formatStacktrace(selectedSpan.attributes?.['exception.stacktrace'] || "No stacktrace available") }}</pre>
</el-dialog>
</div>
<script>
const { createApp, ref, onMounted, nextTick } = Vue;
const { Search } = ElementPlusIconsVue;
createApp({
setup() {
const traces = ref([]);
const traceTree = ref([]);
const selectedSpan = ref(null);
const expandedNodes = ref([]);
const searchTraceId = ref('');
const showStacktrace = ref(false);
const treeProps = {
label: 'name',
children: 'children'
};
const dialogVisible = ref(false);
function searchByTraceId() {
if (!searchTraceId.value) {
buildTraceTree();
return;
}
const filtered = traces.value.filter(trace =>
trace.trace_id.includes(searchTraceId.value)
);
const tree = [];
filtered.forEach(trace => {
if (trace.root_span && trace.root_span.length > 0) {
const root = buildSpanTree(trace.root_span[0]);
tree.push(root);
}
});
traceTree.value = tree;
}
function initTimeline() {
const timelineContainer = document.getElementById('timeline-chart');
const width = timelineContainer.clientWidth;
const height = 100;
const margin = { top: 20, right: 20, bottom: 30, left: 20 };
const svg = d3.select(timelineContainer)
.append('svg')
.attr('width', width)
.attr('height', height);
const now = new Date();
const oneDayAgo = new Date(now.getTime() - 24 * 60 * 60 * 1000);
const x = d3.scaleTime()
.domain([oneDayAgo, now])
.range([margin.left, width - margin.right]);
svg.append('g')
.attr('transform', `translate(0,${height - margin.bottom})`)
.call(d3.axisBottom(x)
.ticks(d3.timeHour.every(2))
.tickFormat(d3.timeFormat("%H:%M")));
svg.append('g')
.attr('class', 'grid')
.attr('transform', `translate(0,${height - margin.bottom})`)
.call(d3.axisBottom(x)
.ticks(d3.timeMinute.every(10))
.tickSize(-5)
.tickFormat(''));
if (traces.value && traces.value.length > 0) {
const colorScale = d3.scaleOrdinal()
.domain(traces.value.map((_, i) => i))
.range(d3.schemeCategory10);
traces.value.forEach((trace, index) => {
if (trace.root_span && trace.root_span.length > 0) {
const span = trace.root_span[0];
const startTime = new Date(span.start_time);
const endTime = new Date(span.end_time);
const duration = endTime - startTime;
if (startTime >= oneDayAgo && startTime <= now) {
svg.append('rect')
.attr('x', x(startTime))
.attr('y', margin.top + 30)
.attr('width', Math.max(3, x(endTime) - x(startTime)))
.attr('height', 20)
.attr('fill', colorScale(index))
.attr('rx', 2)
.attr('opacity', 0.7)
.on('mouseover', function () {
d3.select(this).attr('opacity', 1);
})
.on('mouseout', function () {
d3.select(this).attr('opacity', 0.7);
});
}
}
});
}
}
function renderTimelineVisualization() {
if (!selectedSpan.value) return '';
const currentTrace = traceTree.value.find(t => t.trace_id === selectedSpan.value.trace_id);
if (!currentTrace) return '';
const rootSpan = currentTrace.root_span?.[0] || currentTrace;
let minTime = new Date(rootSpan.start_time).getTime();
let maxTime = new Date(rootSpan.end_time).getTime();
const timelineContainer = document.createElement('div');
timelineContainer.className = 'trace-timeline';
timelineContainer.style.height = '60px';
timelineContainer.style.marginBottom = '20px';
timelineContainer.style.width = '100%';
const svg = d3.select(timelineContainer)
.append('svg')
.attr('width', '100%')
.attr('height', '100%')
.attr('viewBox', '0 0 1000 60');
const margin = { top: 10, right: 0, bottom: 30, left: 0 };
const width = 1000 - margin.left - margin.right;
const height = 60 - margin.top - margin.bottom;
const g = svg.append('g')
.attr('transform', `translate(${margin.left},${margin.top})`);
const x = d3.scaleTime()
.domain([new Date(minTime), new Date(maxTime)])
.range([0, width]);
g.append('g')
.attr('class', 'axis axis--x')
.attr('transform', `translate(0,${height})`)
.call(d3.axisBottom(x)
.ticks(5)
.tickFormat(d3.timeFormat("%H:%M:%S.%L")));
g.selectAll(".grid-line")
.data(x.ticks(5))
.enter().append("line")
.attr("class", "grid-line")
.attr("x1", d => x(d))
.attr("x2", d => x(d))
.attr("y1", 0)
.attr("y2", height)
.attr("stroke", "#eee")
.attr("stroke-width", 1);
const timelineHtml = timelineContainer.outerHTML;
function renderSpans(span, depth = 0, rowIndex = 0) {
const spanStart = new Date(span.start_time).getTime();
const spanEnd = new Date(span.end_time).getTime();
const position = Math.min(13, Math.max(5, ((spanStart - minTime) / (maxTime - minTime)) * 10 * 0.9 + 5));
const width = Math.min(92, Math.max(2, ((spanEnd - spanStart) / (maxTime - minTime)) * 100 * 0.9 + 2));
//const position = ((spanStart - minTime) / (maxTime - minTime)) * 10 * 0.9 + 5;
//const width = ((spanEnd - spanStart) / (maxTime - minTime)) * 100 * 0.9 + 2;
const minWidth = 0.5;
const adjustedWidth = Math.max(width, minWidth);
const row = rowIndex * 24;
let childrenHtml = '';
let nextRowIndex = rowIndex + 1;
if (span.children && span.children.length > 0) {
childrenHtml = span.children.map(child => {
const childHtml = renderSpans(child, depth + 1, nextRowIndex);
nextRowIndex += countSpans(child);
return childHtml;
}).join('');
}
return `
<div class="span-visualization"
style="top: ${row}px;
left: ${position}%;
width: ${adjustedWidth}%;
background: ${span.status.code === 'StatusCode.ERROR' ? '#F56C6C' : '#409EFF'};
opacity: ${span.span_id === selectedSpan.value.span_id ? 1 : 0.6}"
onclick="window.handleSpanClick.call(this, ${JSON.stringify(span).replace(/"/g, '&quot;')})">
<span class="span-label">${span.duration_ms} ${span.name}</span>
</div>
${childrenHtml}
`;
}
return `
<h3>Timeline Visualization</h3>
${timelineHtml}
<div class="span-visualization-container" style="height: ${traceTree.value.length * 24 + 100}px">
${renderSpans(rootSpan)}
</div>
`;
}
function countSpans(span) {
let count = 1;
if (span.children && span.children.length > 0) {
span.children.forEach(child => {
count += countSpans(child);
});
}
return count;
}
async function fetchTraces() {
try {
const response = await fetch('/api/trace/list');
const data = await response.json();
traces.value = data.data;
buildTraceTree();
initTimeline();
} catch (error) {
console.error('Error loading traces:', error);
}
}
function buildTraceTree() {
const tree = [];
traces.value.forEach(trace => {
if (trace.root_span && trace.root_span.length > 0) {
const root = buildSpanTree(trace.root_span[0]);
tree.push(root);
}
});
traceTree.value = tree;
}
function buildSpanTree(span) {
const node = {
...span,
children: []
};
if (span.children && span.children.length > 0) {
span.children.forEach(child => {
node.children.push(buildSpanTree(child));
});
}
return node;
}
function handleNodeClick(data) {
selectedSpan.value = data;
nextTick(() => {
renderTimelineVisualization();
});
}
function handleSpanClick(data) {
selectedSpan.value = data;
dialogVisible.value = true;
if (!expandedNodes.value.includes(data.span_id)) {
expandedNodes.value.push(data.span_id);
}
}
function formatTime(timestamp) {
return timestamp.split('.')[0];
}
function formatAttributes(attrs) {
return JSON.stringify(attrs, null, 2);
}
function formatStacktrace(stacktrace) {
if (!stacktrace) return 'No stacktrace available';
try {
return JSON.stringify(JSON.parse(stacktrace), null, 2);
} catch {
return stacktrace;
}
}
onMounted(() => {
fetchTraces();
//setInterval(fetchTraces, 5000);
window.handleSpanClick = handleSpanClick;
});
return {
traces,
traceTree,
selectedSpan,
expandedNodes,
treeProps,
handleNodeClick,
handleSpanClick,
formatTime,
formatAttributes,
dialogVisible,
renderTimelineVisualization,
searchTraceId,
searchByTraceId,
showStacktrace,
formatStacktrace
};
}
}).use(ElementPlus).component('search', Search).mount('#app');
</script>
</body>
</html>
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/eslint.config.js
+28
View File
@@ -0,0 +1,28 @@
import js from '@eslint/js'
import globals from 'globals'
import reactHooks from 'eslint-plugin-react-hooks'
import reactRefresh from 'eslint-plugin-react-refresh'
import tseslint from 'typescript-eslint'
export default tseslint.config(
{ ignores: ['dist'] },
{
extends: [js.configs.recommended, ...tseslint.configs.recommended],
files: ['**/*.{ts,tsx}'],
languageOptions: {
ecmaVersion: 2020,
globals: globals.browser,
},
plugins: {
'react-hooks': reactHooks,
'react-refresh': reactRefresh,
},
rules: {
...reactHooks.configs.recommended.rules,
'react-refresh/only-export-components': [
'warn',
{ allowConstantExport: true },
],
},
},
)
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/index.html
+13
View File
@@ -0,0 +1,13 @@
<!doctype html>
<html lang="en">
<head>
<meta charset="UTF-8" />
<link rel="icon" type="image/svg+xml" href="/aworld_logo.png" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>Aworld</title>
</head>
<body>
<div id="root"></div>
<script type="module" src="/src/main.tsx"></script>
</body>
</html>
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/package.json
+42
View File
@@ -0,0 +1,42 @@
{
"name": "Aworld-UI",
"private": true,
"version": "0.0.0",
"type": "module",
"scripts": {
"dev": "vite",
"build": "tsc -b && vite build",
"lint": "eslint .",
"preview": "vite preview"
},
"dependencies": {
"@ant-design/x": "^1.4.0",
"@xyflow/react": "^12.8.1",
"antd": "^5.26.0",
"antd-style": "^3.7.1",
"dagre": "^0.8.5",
"mermaid": "^11.7.0",
"react": "^18.2.0",
"react-dom": "^18.2.0",
"react-markdown": "^10.1.0",
"react-router-dom": "^6.30.1",
"uuid": "^11.1.0"
},
"devDependencies": {
"@eslint/js": "^9.25.0",
"@types/dagre": "^0.7.53",
"@types/node": "^24.0.4",
"@types/react": "^18.2.0",
"@types/react-dom": "^18.2.0",
"@vitejs/plugin-react": "^4.4.1",
"eslint": "^9.25.0",
"eslint-plugin-react-hooks": "^5.2.0",
"eslint-plugin-react-refresh": "^0.4.19",
"globals": "^16.0.0",
"less": "^4.3.0",
"typescript": "~5.8.3",
"typescript-eslint": "^8.30.1",
"vite": "^6.3.5"
},
"repository": "git@github.com:inclusionAI/AWorld.git"
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/public/aworld_logo.png
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 36 KiB

chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/public/trace_ui.html
+526
View File
@@ -0,0 +1,526 @@
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Trace Viewer V2</title>
<link rel="stylesheet" href="https://unpkg.com/element-plus/dist/index.css">
<script src="https://unpkg.com/vue@3/dist/vue.global.js"></script>
<script src="https://unpkg.com/element-plus"></script>
<script src="https://unpkg.com/@element-plus/icons-vue"></script>
<script src="https://d3js.org/d3.v7.min.js"></script>
<style>
.trace-container {
display: flex;
flex-direction: column;
height: 100vh;
font-family: 'Helvetica Neue', Arial, sans-serif;
}
.trace-content {
display: flex;
flex: 1;
overflow: hidden;
}
.trace-list {
width: 30%;
overflow-y: auto;
border-right: 1px solid #e6e6e6;
}
.trace-detail {
width: 70%;
padding: 20px;
overflow-y: auto;
}
.timeline {
height: 120px;
min-width: 100%;
background: #f5f5f5;
padding: 10px;
border-bottom: 1px solid #e6e6e6;
}
.span-node {
cursor: pointer;
padding: 5px 0;
}
.span-node:hover {
background-color: #f0f7ff;
}
.span-duration {
color: #666;
font-size: 12px;
}
.timeline-bg {
fill: #f8f8f8;
}
.axis--x path {
stroke: #333;
stroke-width: 1px;
}
.axis--x line {
stroke: #ddd;
}
.axis--x text {
font-size: 12px;
fill: #333;
}
.timeline-visualization {
flex: 1;
padding: 20px;
background: #f8f8f8;
border-left: 1px solid #e6e6e6;
overflow-y: auto;
position: relative;
height: 100%;
}
.span-visualization-container {
position: relative;
height: 100%;
margin-top: 40px;
}
.span-visualization {
height: 20px;
background: #409EFF;
position: absolute;
margin-top: 2px;
border-radius: 2px;
}
.span-label {
font-size: 8px;
color: white;
padding: 0 5px;
white-space: nowrap;
overflow: hidden;
text-overflow: ellipsis;
}
.trace-timeline {
background: #f5f5f5;
padding: 10px;
border-radius: 4px;
}
.trace-timeline svg {
display: block;
}
.trace-timeline .axis path {
stroke: #333;
stroke-width: 1px;
}
.trace-timeline .axis line {
stroke: #ddd;
}
.trace-timeline .axis text {
font-size: 12px;
fill: #333;
}
</style>
</head>
<body>
<div id="app" class="trace-container">
<!-- Top timeline -->
<div class="timeline">
<div id="timeline-chart"></div>
</div>
<div class="trace-content">
<div class="trace-list">
<div style="padding: 10px; border-bottom: 1px solid #e6e6e6;">
<el-input v-model="searchTraceId" placeholder="输入Trace ID搜索" style="width: 100%;"
@keyup.enter="searchByTraceId">
<template #append>
<el-button @click="searchByTraceId">
<el-icon>
<search />
</el-icon>
</el-button>
</template>
</el-input>
</div>
<el-tree :data="traceTree" node-key="span_id" :props="treeProps" :expand-on-click-node="false"
@node-click="handleNodeClick" :default-expanded-keys="expandedNodes">
<template #default="{ node, data }">
<span class="span-node">
{{ data.name }}
<span class="span-duration">({{ data.duration_ms.toFixed(2) }}ms)</span>
</span>
</template>
</el-tree>
</div>
<div class="timeline-visualization" v-if="selectedSpan" v-html="renderTimelineVisualization()">
</div>
</div>
<!-- Span detail -->
<el-dialog v-model="dialogVisible" title="Span Details" width="70%">
<el-descriptions :column="2" border>
<el-descriptions-item label="Trace ID">{{ selectedSpan.trace_id }}</el-descriptions-item>
<el-descriptions-item label="Span ID">{{ selectedSpan.span_id }}</el-descriptions-item>
<el-descriptions-item label="Parent Span ID">{{ selectedSpan.parent_id || 'None'
}}</el-descriptions-item>
<el-descriptions-item label="Name">{{ selectedSpan.name }}</el-descriptions-item>
<el-descriptions-item label="Status">
<div style="display: flex; justify-content: space-between; align-items: center;">
<span :style="{color: selectedSpan.status.code === 'StatusCode.ERROR' ? '#F56C6C' : ''}">
{{ selectedSpan.status.code }}
</span>
<el-button v-if="selectedSpan.status.code === 'StatusCode.ERROR'" type="text" size="small"
@click="showStacktrace = true" icon="View" style="color: #F56C6C">
View Stack
</el-button>
</div>
</el-descriptions-item>
<el-descriptions-item label="Start Time">{{ selectedSpan.start_time}}</el-descriptions-item>
<el-descriptions-item label="End Time">{{ selectedSpan.end_time }}</el-descriptions-item>
<el-descriptions-item label="Duration">{{ selectedSpan.duration_ms.toFixed(2) }}
ms</el-descriptions-item>
</el-descriptions>
<el-card style="margin-top: 20px;">
<template #header>
<h4>Attributes</h4>
</template>
<pre style="
max-height: 400px;
overflow: auto;
white-space: pre-wrap;
word-break: break-all;
background: #f8f8f8;
padding: 10px;
border-radius: 4px;
">{{ formatAttributes(selectedSpan.attributes) }}</pre>
</el-card>
</el-dialog>
<el-dialog v-model="showStacktrace" title="Stacktrace Details" width="70%">
<pre>{{ formatStacktrace(selectedSpan.attributes?.['exception.stacktrace'] || "No stacktrace available") }}</pre>
</el-dialog>
</div>
<script>
const { createApp, ref, onMounted, nextTick } = Vue;
const { Search } = ElementPlusIconsVue;
createApp({
setup() {
const traces = ref([]);
const traceTree = ref([]);
const selectedSpan = ref(null);
const expandedNodes = ref([]);
const searchTraceId = ref('');
const showStacktrace = ref(false);
const treeProps = {
label: 'name',
children: 'children'
};
const dialogVisible = ref(false);
function searchByTraceId() {
if (!searchTraceId.value) {
buildTraceTree();
return;
}
const filtered = traces.value.filter(trace =>
trace.trace_id.includes(searchTraceId.value)
);
const tree = [];
filtered.forEach(trace => {
if (trace.root_span && trace.root_span.length > 0) {
const root = buildSpanTree(trace.root_span[0]);
tree.push(root);
}
});
traceTree.value = tree;
}
function initTimeline() {
const timelineContainer = document.getElementById('timeline-chart');
const width = timelineContainer.clientWidth;
const height = 100;
const margin = { top: 20, right: 20, bottom: 30, left: 20 };
const svg = d3.select(timelineContainer)
.append('svg')
.attr('width', width)
.attr('height', height);
const now = new Date();
const oneDayAgo = new Date(now.getTime() - 24 * 60 * 60 * 1000);
const x = d3.scaleTime()
.domain([oneDayAgo, now])
.range([margin.left, width - margin.right]);
svg.append('g')
.attr('transform', `translate(0,${height - margin.bottom})`)
.call(d3.axisBottom(x)
.ticks(d3.timeHour.every(2))
.tickFormat(d3.timeFormat("%H:%M")));
svg.append('g')
.attr('class', 'grid')
.attr('transform', `translate(0,${height - margin.bottom})`)
.call(d3.axisBottom(x)
.ticks(d3.timeMinute.every(10))
.tickSize(-5)
.tickFormat(''));
if (traces.value && traces.value.length > 0) {
const colorScale = d3.scaleOrdinal()
.domain(traces.value.map((_, i) => i))
.range(d3.schemeCategory10);
traces.value.forEach((trace, index) => {
if (trace.root_span && trace.root_span.length > 0) {
const span = trace.root_span[0];
const startTime = new Date(span.start_time);
const endTime = new Date(span.end_time);
const duration = endTime - startTime;
if (startTime >= oneDayAgo && startTime <= now) {
svg.append('rect')
.attr('x', x(startTime))
.attr('y', margin.top + 30)
.attr('width', Math.max(3, x(endTime) - x(startTime)))
.attr('height', 20)
.attr('fill', colorScale(index))
.attr('rx', 2)
.attr('opacity', 0.7)
.on('mouseover', function () {
d3.select(this).attr('opacity', 1);
})
.on('mouseout', function () {
d3.select(this).attr('opacity', 0.7);
});
}
}
});
}
}
function renderTimelineVisualization() {
if (!selectedSpan.value) return '';
const currentTrace = traceTree.value.find(t => t.trace_id === selectedSpan.value.trace_id);
if (!currentTrace) return '';
const rootSpan = currentTrace.root_span?.[0] || currentTrace;
let minTime = new Date(rootSpan.start_time).getTime();
let maxTime = new Date(rootSpan.end_time).getTime();
const timelineContainer = document.createElement('div');
timelineContainer.className = 'trace-timeline';
timelineContainer.style.height = '60px';
timelineContainer.style.marginBottom = '20px';
timelineContainer.style.width = '100%';
const svg = d3.select(timelineContainer)
.append('svg')
.attr('width', '100%')
.attr('height', '100%')
.attr('viewBox', '0 0 1000 60');
const margin = { top: 10, right: 0, bottom: 30, left: 0 };
const width = 1000 - margin.left - margin.right;
const height = 60 - margin.top - margin.bottom;
const g = svg.append('g')
.attr('transform', `translate(${margin.left},${margin.top})`);
const x = d3.scaleTime()
.domain([new Date(minTime), new Date(maxTime)])
.range([0, width]);
g.append('g')
.attr('class', 'axis axis--x')
.attr('transform', `translate(0,${height})`)
.call(d3.axisBottom(x)
.ticks(5)
.tickFormat(d3.timeFormat("%H:%M:%S.%L")));
g.selectAll(".grid-line")
.data(x.ticks(5))
.enter().append("line")
.attr("class", "grid-line")
.attr("x1", d => x(d))
.attr("x2", d => x(d))
.attr("y1", 0)
.attr("y2", height)
.attr("stroke", "#eee")
.attr("stroke-width", 1);
const timelineHtml = timelineContainer.outerHTML;
function renderSpans(span, depth = 0, rowIndex = 0) {
const spanStart = new Date(span.start_time).getTime();
const spanEnd = new Date(span.end_time).getTime();
const position = Math.min(13, Math.max(5, ((spanStart - minTime) / (maxTime - minTime)) * 10 * 0.9 + 5));
const width = Math.min(92, Math.max(2, ((spanEnd - spanStart) / (maxTime - minTime)) * 100 * 0.9 + 2));
//const position = ((spanStart - minTime) / (maxTime - minTime)) * 10 * 0.9 + 5;
//const width = ((spanEnd - spanStart) / (maxTime - minTime)) * 100 * 0.9 + 2;
const minWidth = 0.5;
const adjustedWidth = Math.max(width, minWidth);
const row = rowIndex * 24;
let childrenHtml = '';
let nextRowIndex = rowIndex + 1;
if (span.children && span.children.length > 0) {
childrenHtml = span.children.map(child => {
const childHtml = renderSpans(child, depth + 1, nextRowIndex);
nextRowIndex += countSpans(child);
return childHtml;
}).join('');
}
return `
<div class="span-visualization"
style="top: ${row}px;
left: ${position}%;
width: ${adjustedWidth}%;
background: ${span.status.code === 'StatusCode.ERROR' ? '#F56C6C' : '#409EFF'};
opacity: ${span.span_id === selectedSpan.value.span_id ? 1 : 0.6}"
onclick="window.handleSpanClick.call(this, ${JSON.stringify(span).replace(/"/g, '&quot;')})">
<span class="span-label">${span.duration_ms} ${span.name}</span>
</div>
${childrenHtml}
`;
}
return `
<h3>Timeline Visualization</h3>
${timelineHtml}
<div class="span-visualization-container" style="height: ${traceTree.value.length * 24 + 100}px">
${renderSpans(rootSpan)}
</div>
`;
}
function countSpans(span) {
let count = 1;
if (span.children && span.children.length > 0) {
span.children.forEach(child => {
count += countSpans(child);
});
}
return count;
}
async function fetchTraces() {
try {
const response = await fetch('/api/trace/list');
const data = await response.json();
traces.value = data.data;
buildTraceTree();
initTimeline();
} catch (error) {
console.error('Error loading traces:', error);
}
}
function buildTraceTree() {
const tree = [];
traces.value.forEach(trace => {
if (trace.root_span && trace.root_span.length > 0) {
const root = buildSpanTree(trace.root_span[0]);
tree.push(root);
}
});
traceTree.value = tree;
}
function buildSpanTree(span) {
const node = {
...span,
children: []
};
if (span.children && span.children.length > 0) {
span.children.forEach(child => {
node.children.push(buildSpanTree(child));
});
}
return node;
}
function handleNodeClick(data) {
selectedSpan.value = data;
nextTick(() => {
renderTimelineVisualization();
});
}
function handleSpanClick(data) {
selectedSpan.value = data;
dialogVisible.value = true;
if (!expandedNodes.value.includes(data.span_id)) {
expandedNodes.value.push(data.span_id);
}
}
function formatTime(timestamp) {
return timestamp.split('.')[0];
}
function formatAttributes(attrs) {
return JSON.stringify(attrs, null, 2);
}
function formatStacktrace(stacktrace) {
if (!stacktrace) return 'No stacktrace available';
try {
return JSON.stringify(JSON.parse(stacktrace), null, 2);
} catch {
return stacktrace;
}
}
onMounted(() => {
fetchTraces();
//setInterval(fetchTraces, 5000);
window.handleSpanClick = handleSpanClick;
});
return {
traces,
traceTree,
selectedSpan,
expandedNodes,
treeProps,
handleNodeClick,
handleSpanClick,
formatTime,
formatAttributes,
dialogVisible,
renderTimelineVisualization,
searchTraceId,
searchByTraceId,
showStacktrace,
formatStacktrace
};
}
}).use(ElementPlus).component('search', Search).mount('#app');
</script>
</body>
</html>
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/api/trace.ts
+5
View File
@@ -0,0 +1,5 @@
import { request } from '@/utils/http';
export const fetchTraceData = (traceId: string) => {
return request(`/api/trace/agent?trace_id=${traceId}`);
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/api/workspace.ts
+66
View File
@@ -0,0 +1,66 @@
import { request } from '../utils/http';
/**
* 工作空间树节点数据结构
*/
export interface WorkspaceTreeResponse {
id: string; // 节点ID
name: string; // 节点名称
type: string; // 节点类型 (dir/file)
parentId: string | null; // 父节点ID
depth: number; // 节点深度
expanded: boolean; // 是否展开
children: WorkspaceTreeResponse[]; // 子节点列表
}
/**
* Get Artifact的请求参数
*/
export interface ArtifactQueryRequest {
artifact_types: string[]; // Artifact类型
artifact_ids: string[]; // Artifact ID
}
/**
* 创建Artifact的请求参数
*/
export interface ArtifactCreateRequest {
name: string; // Artifact名称
type: string; // Artifact类型
content: any; // Artifact内容
}
/**
* 创建Artifact的响应数据
*/
export interface ArtifactCreateResponse {
id: string; // 创建的Artifact ID
status: 'success' | 'failed'; // 操作状态
message?: string; // 可选的状态信息
}
/**
* 获取工作空间树
*/
export const getWorkspaceTree = (sessionId: string) =>
request(`api/workspaces/${sessionId}/tree`);
/**
* 获取工作空间Artifacts
*/
export const getWorkspaceArtifacts = (sessionId: string, body: ArtifactQueryRequest) =>
request(`api/workspaces/${sessionId}/artifacts`, {
method: 'POST',
body
});
/**
* 创建工作空间Artifact
*/
export const createArtifact = (workspaceId: string, body: ArtifactCreateRequest) =>
request(`api/workspaces/${workspaceId}/artifacts`, {
method: 'POST',
body
});
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/assets/aworld_logo.png
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 36 KiB

chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/constants/index.ts
+1
View File
@@ -0,0 +1 @@
export const DEFAULT_NAME = '';
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/global.d.ts
Vendored
+4
View File
@@ -0,0 +1,4 @@
declare module '*.less' {
const classes: { [key: string]: string };
export default classes;
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/global.less
+3
View File
@@ -0,0 +1,3 @@
body{
margin: 0;
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/hooks/useAgentId.ts
+44
View File
@@ -0,0 +1,44 @@
import { useEffect, useState } from 'react';
export const useAgentId = () => {
const [agentId, setAgentId] = useState<string>('');
// 从URL参数中获取agent ID
const getAgentIdFromURL = (): string => {
const urlParams = new URLSearchParams(window.location.search);
return urlParams.get('agentid') || '';
};
// 更新URL参数中的agent ID
const updateURLAgentId = (id: string) => {
const url = new URL(window.location.href);
if (id) {
url.searchParams.set('agentid', id);
} else {
url.searchParams.delete('agentid');
}
window.history.replaceState({}, '', url.toString());
};
// 设置新的agent ID并更新URL
const setAgentIdAndUpdateURL = (id: string) => {
setAgentId(id);
updateURLAgentId(id);
};
useEffect(() => {
// 初始化时检查URL中是否有agent ID
const urlAgentId = getAgentIdFromURL();
if (urlAgentId) {
// 如果URL中有agent ID，使用它
setAgentId(urlAgentId);
}
}, []);
return {
agentId,
setAgentIdAndUpdateURL,
updateURLAgentId,
};
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/hooks/useSessionId.ts
+48
View File
@@ -0,0 +1,48 @@
import { useEffect, useState } from 'react';
import { v4 as uuidv4 } from 'uuid';
export const useSessionId = () => {
const [sessionId, setSessionId] = useState<string>('');
// 从URL参数中获取session ID
const getSessionIdFromURL = (): string => {
const urlParams = new URLSearchParams(window.location.search);
return urlParams.get('session_id') || '';
};
// 更新URL参数中的session ID
const updateURLSessionId = (id: string) => {
const url = new URL(window.location.href);
url.searchParams.set('session_id', id);
window.history.replaceState({}, '', url.toString());
};
// 生成新的session ID并更新URL
const generateNewSessionId = (): string => {
const newId = uuidv4();
setSessionId(newId);
updateURLSessionId(newId);
console.log('generateNewSessionId', newId);
return newId;
};
useEffect(() => {
// 初始化时检查URL中是否有session ID
const urlSessionId = getSessionIdFromURL();
if (urlSessionId) {
// 如果URL中有session ID，使用它
setSessionId(urlSessionId);
} else {
// 如果URL中没有session ID，生成一个新的
generateNewSessionId();
}
}, []);
return {
sessionId,
setSessionId,
generateNewSessionId,
updateURLSessionId,
};
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/main.tsx
+12
View File
@@ -0,0 +1,12 @@
import { StrictMode } from 'react'
import { createRoot } from 'react-dom/client'
import { HashRouter } from 'react-router-dom'
import './global.less'
import Router from './router'
createRoot(document.getElementById('root')!).render(
<StrictMode>
<HashRouter>
<Router />
</HashRouter>
</StrictMode>,
)
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/App.tsx
+1044
View File
File diff suppressed because it is too large Load Diff
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/app.less
+3
View File
@@ -0,0 +1,3 @@
.ant-bubble-content .ant-bubble-content-filled{
background-color: red;
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Bubble/index.less
View File
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Bubble/index.tsx
View File
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/cardDefault/index.less
+30
View File
@@ -0,0 +1,30 @@
.defaultbox{
position: relative;
.btn-workspace{
position: absolute;
top: -40px;
right: 0;
}
.pre-wrap{
white-space: pre-wrap;
}
.action-btn {
color: #1890ff;
cursor: pointer;
transition: color 0.3s;
padding: 0 4px;
&:hover {
color: #40a9ff;
}
&:active {
color: #096dd9;
}
}
.ant-collapse{
// width: 668px;
width: 100%;
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/cardDefault/index.tsx
+84
View File
@@ -0,0 +1,84 @@
import { MenuUnfoldOutlined } from '@ant-design/icons';
import { Button, Collapse, Space, message } from 'antd';
import React, { useCallback, useState } from 'react';
import type { ToolCardData } from '../utils';
import './index.less';
interface Props {
sessionId: string;
data: ToolCardData;
onOpenWorkspace: (data: ToolCardData) => void;
}
const CardDefault: React.FC<Props> = ({ sessionId, data, onOpenWorkspace }) => {
// 当前展开的面板keys
const [activeKeys, setActiveKeys] = useState<string[]>([]);
// 处理复制
const handleCopy = useCallback(
async (panelKey: string) => {
try {
const content = panelKey === '1' ? data.arguments : data.results;
await navigator.clipboard.writeText(content);
message.success('Copy Successful');
} catch (error) {
message.error('Copy Failed');
}
},
[data]
);
// 打开workspace
const handleOpenWorkspace = useCallback(() => {
if (onOpenWorkspace) {
onOpenWorkspace(data);
}
}, [onOpenWorkspace, sessionId, data]);
//操作按钮
const renderExtra = useCallback(
(panelKey: string) => (
<Space size="small" onClick={(e) => e.stopPropagation()}>
<span className="action-btn" onClick={() => handleCopy(panelKey)}>
Copy
</span>
</Space>
),
[handleCopy]
);
const items = [
{
key: '1',
label: 'tool_call_arguments',
extra: renderExtra('1'),
children: (
<pre className="pre-wrap">
<code>{data.arguments}</code>
</pre>
)
},
{
key: '2',
label: 'tool_call_result',
extra: renderExtra('2'),
children: (
<pre className="pre-wrap">
<code>{data.results}</code>
</pre>
)
}
];
return (
<div className="defaultbox">
{data?.artifacts?.length > 0 && (
<Button type="link" className="btn-workspace" icon={<MenuUnfoldOutlined />} onClick={handleOpenWorkspace}>
View Workspace
</Button>
)}
<Collapse activeKey={activeKeys} onChange={(keys) => setActiveKeys(Array.isArray(keys) ? keys : [keys])} items={items} />
</div>
);
};
export default React.memo(CardDefault);
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/cardLinkList/index.less
+54
View File
@@ -0,0 +1,54 @@
.cardwrap {
background-color: #eee;
border-radius: 10px;
padding: 10px;
position: relative;
.btn-workspace {
position: absolute;
top: -38px;
right: -6px;
}
.card-length {
font-size: 14px;
color: #333;
.ant-tag {
max-width: 480px;
white-space: nowrap;
overflow: hidden;
text-overflow: ellipsis;
padding: 0 10px;
border-radius: 8px;
line-height: 24px;
}
.check-icon {
color: #1890ff;
margin-right: 8px;
font-size: 16px;
}
}
}
.cardbox {
// width: 668px;
// width: 648px;
overflow-x: auto;
margin-top: 10px;
.card-item {
width: 175px;
min-width: 175px;
// margin-bottom: 16px;
.ant-card-head {
padding: 0 14px;
min-height: 50px;
}
.ant-card-body {
padding: 10px 14px 12px;
.desc {
margin-bottom: 0;
}
}
& + .card-item {
margin-left: 6px;
}
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/cardLinkList/index.tsx
+58
View File
@@ -0,0 +1,58 @@
import { CheckOutlined, MenuUnfoldOutlined, SearchOutlined } from '@ant-design/icons';
import { Button, Card, Flex, Tag, Typography } from 'antd';
import React, { useCallback } from 'react';
import type { ToolCardData } from '../utils';
import './index.less';
interface Props {
sessionId: string;
data: ToolCardData;
onOpenWorkspace?: (data: ToolCardData) => void;
}
interface ItemInterface {
title: string;
snippet: string;
link?: string;
}
const cardLinkList: React.FC<Props> = ({ sessionId, data, onOpenWorkspace }) => {
const items = data?.card_data?.search_items;
const cardItems = Array.isArray(items) ? items.filter((item) => item?.title && item?.link) : [];
// 打开workspace
const handleOpenWorkspace = useCallback(() => {
if (onOpenWorkspace) {
onOpenWorkspace(data);
}
}, [onOpenWorkspace, sessionId, data]);
return (
<div className="cardwrap bg">
<Button type="link" className="btn-workspace" icon={<MenuUnfoldOutlined />} onClick={handleOpenWorkspace}>
View Workspace
</Button>
<Flex justify="space-between" align="center" className="card-length">
<Tag icon={<SearchOutlined />}>{`search keywords: ${data?.card_data?.query || ''}`}</Tag>
<Flex align="center">
<CheckOutlined className="check-icon" />
{cardItems.length} results
</Flex>
</Flex>
<div className="border-box">
<Flex className="cardbox">
{cardItems?.map((item: ItemInterface, index: number) => (
<Card title={item?.title} key={index} className="card-item" onClick={() => item?.link && window.open(item?.link, '_blank', 'noopener,noreferrer')}>
<Typography.Paragraph className="desc" ellipsis={{ rows: 3, tooltip: typeof item?.snippet === 'string' ? item?.snippet : '' }}>
{item?.snippet}
</Typography.Paragraph>
<Typography.Text ellipsis={{ tooltip: typeof item?.link === 'string' ? item?.link : '' }}>{item?.link}</Typography.Text>
</Card>
))}
</Flex>
</div>
</div>
);
};
export default cardLinkList;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/index.less
+8
View File
@@ -0,0 +1,8 @@
.markdownbox{
p>strong{
padding-left: 5px;
}
pre{
white-space: pre-wrap;
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/index.tsx
+106
View File
@@ -0,0 +1,106 @@
import React, { useEffect, useRef } from 'react';
import ReactMarkdown from 'react-markdown';
import CardDefault from './cardDefault';
import CardLinkList from './cardLinkList';
import './index.less';
import type { ToolCardData } from './utils';
import { extractToolCards } from './utils';
interface BubbleItemProps {
sessionId: string;
data: string;
trace_id: string;
onOpenWorkspace?: (data: ToolCardData) => void;
isLoading?: boolean;
}
const BubbleItem: React.FC<BubbleItemProps> = ({ sessionId, data, onOpenWorkspace, isLoading = false }) => {
// 用于记录上次打开的workspace数据，避免重复调用
const lastWorkspaceDataRef = useRef<ToolCardData | null>(null);
// 修改openWorkspace函数，直接调用外部回调
const openWorkspace = (data: ToolCardData) => {
if (onOpenWorkspace) {
onOpenWorkspace(data);
}
};
const { segments } = extractToolCards(data);
// 比较两个workspace数据是否相同
const isWorkspaceDataEqual = (data1: ToolCardData | null, data2: ToolCardData | null): boolean => {
if (!data1 && !data2) return true;
if (!data1 || !data2) return false;
// 比较关键字段来判断是否为同一个workspace
return (
data1.tool_call_id === data2.tool_call_id &&
data1.artifacts?.length === data2.artifacts?.length &&
JSON.stringify(data1.artifacts) === JSON.stringify(data2.artifacts)
);
};
// 自动打开workspace的逻辑 - 只在流式输出过程中自动打开
useEffect(() => {
// 只有在流式输出过程中才自动打开workspace
if (!isLoading) {
return;
}
// 查找最新的具有workspace功能的tool_card（不区分card类型）
const toolCardSegments = segments.filter(segment => segment.type === 'tool_card');
// 从最后一个开始查找，找到第一个有artifacts的tool_card
const latestWorkspaceCard = toolCardSegments
.slice()
.reverse()
.find(segment => {
return segment.type === 'tool_card' &&
segment.data?.artifacts?.length > 0;
});
if (latestWorkspaceCard && latestWorkspaceCard.type === 'tool_card' && onOpenWorkspace) {
const currentWorkspaceData = latestWorkspaceCard.data;
// 检查当前workspace数据是否与上次相同
if (!isWorkspaceDataEqual(lastWorkspaceDataRef.current, currentWorkspaceData)) {
// 更新记录的workspace数据
lastWorkspaceDataRef.current = currentWorkspaceData;
// 使用requestAnimationFrame确保在下一帧渲染后打开workspace
const frameId = requestAnimationFrame(() => {
openWorkspace(currentWorkspaceData);
});
return () => cancelAnimationFrame(frameId);
} else {
console.log("latest workspace opened!", currentWorkspaceData, lastWorkspaceDataRef.current)
}
}
}, [segments, onOpenWorkspace, openWorkspace, isLoading]);
// console.log('segments:', segments);
return (
<div className="card">
{segments.map((segment, index) => {
if (segment.type === 'text') {
return (
<div className="markdownbox" key={`text-${index}`}>
<ReactMarkdown>{segment.content}</ReactMarkdown>
</div>
);
} else if (segment.type === 'tool_card') {
const cardType = segment.data?.card_type;
if (cardType === 'tool_call_card_link_list') {
return <CardLinkList key={`tool-${index}`} sessionId={sessionId} data={segment.data} onOpenWorkspace={openWorkspace} />;
} else {
return <CardDefault key={`tool-${index}`} sessionId={sessionId} data={segment.data} onOpenWorkspace={openWorkspace} />;
}
}
})}
{/* 移除内部的Drawer */}
</div>
);
};
export default BubbleItem;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/BubbleItem/utils.ts
+67
View File
@@ -0,0 +1,67 @@
export interface ToolCardData {
tool_type: string;
tool_name: string;
function_name: string;
tool_call_id: string;
arguments: string;
results: string;
card_type: string;
card_data: any;
artifacts: any[];
}
type ContentSegment =
| { type: 'text'; content: string }
| { type: 'tool_card'; data: ToolCardData; raw: string };
export interface ParsedContent {
segments: ContentSegment[];
}
export const extractToolCards = (content: string): ParsedContent => {
const toolCardRegex = /(.*?)(```tool_card\s*({[\s\S]*?})\s*```)/gs;
const segments: ContentSegment[] = [];
let lastIndex = 0;
let match;
while ((match = toolCardRegex.exec(content)) !== null) {
const [, textBefore, fullToolCard, toolCardJson] = match;
// 添加文本内容
if (textBefore) {
segments.push({
type: 'text',
content: textBefore.trim()
});
}
// 添加工具卡片
try {
segments.push({
type: 'tool_card',
data: JSON.parse(toolCardJson),
raw: fullToolCard.trim()
});
} catch (e) {
console.error('Failed to parse tool_card JSON:', e);
// 如果解析失败，仍保留原始文本
segments.push({
type: 'text',
content: fullToolCard.trim()
});
}
lastIndex = toolCardRegex.lastIndex;
}
// 添加最后剩余的文本内容
const remainingText = content.slice(lastIndex);
if (remainingText.trim()) {
segments.push({
type: 'text',
content: remainingText.trim()
});
}
return { segments };
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/Trace/index.less
+13
View File
@@ -0,0 +1,13 @@
.tracebox {
padding: 16px;
.mermaid {
width: 80%;
max-width: 700px;
margin: 0 auto;
text-align: center;
}
.trace-id{
text-align: center;
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/Trace/index.tsx
+88
View File
@@ -0,0 +1,88 @@
import React, { useEffect, useRef, useState, useCallback } from 'react';
import mermaid from 'mermaid';
import { fetchTraceData } from '@/api/trace';
import { treeToMermaid } from './mermaidUtils';
import './index.less';
interface TraceProps {
traceId?: string;
drawerVisible?: boolean;
}
const Trace: React.FC<TraceProps> = ({ traceId, drawerVisible }) => {
const diagramRef = useRef<HTMLDivElement>(null);
const [mermaidCode, setMermaidCode] = useState<string>('');
const isFetching = useRef(false);
const renderError = (message: string) => {
return `graph TD\n A[${message}]`;
};
const handleFetchTrace = useCallback(async () => {
if (!traceId || isFetching.current) return;
isFetching.current = true;
try {
const result = await fetchTraceData(traceId);
if (!result?.data) throw new Error('Invalid trace data format');
const mermaidData = treeToMermaid(result.data);
if (!mermaidData.includes('graph') && !mermaidData.includes('flowchart')) {
throw new Error(`Invalid mermaid data format`);
}
setMermaidCode(mermaidData);
} catch (error) {
console.error('Trace processing error:', error);
setMermaidCode(renderError(error instanceof Error ? error.message : 'Data Processing Error'));
} finally {
isFetching.current = false;
}
}, [traceId]);
useEffect(() => {
if (traceId && drawerVisible) {
handleFetchTrace();
}
return () => {
// Cleanup if component unmounts during fetch
};
}, [traceId, drawerVisible, handleFetchTrace]);
useEffect(() => {
if (!mermaidCode) return;
const renderMermaid = async () => {
try {
mermaid.initialize({
startOnLoad: false,
securityLevel: 'loose'
});
if (diagramRef.current) {
diagramRef.current.innerHTML = mermaidCode;
await mermaid.run({
nodes: [diagramRef.current],
suppressErrors: true
});
}
} catch (error) {
console.error('Mermaid error:', error);
setMermaidCode(renderError(error instanceof Error ? error.message : 'Rendering Error'));
}
};
renderMermaid();
}, [mermaidCode]);
return (
<div className="tracebox">
<div ref={diagramRef} className="mermaid">
{mermaidCode ||
`graph TD
A[loading...]`}
</div>
<p className='trace-id'>traceId: {traceId}</p>
</div>
);
};
export default Trace;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/Trace/mermaidUtils.ts
+57
View File
@@ -0,0 +1,57 @@
interface TraceNode {
show_name: string;
span_id?: string;
duration_ms?: number;
children?: TraceNode[];
}
export function treeToMermaid(input: any): string {
let output = 'flowchart TD\n';
const processedNodes = new Set<string>();
function processNode(node: TraceNode, parentId?: string) {
if (!node?.show_name) return;
const rawNodeId = `${node.show_name}_${node.span_id || ''}`.replace(/\s+/g, '_');
const cleanNodeId = rawNodeId.replace(/[^a-zA-Z0-9_]/g, '_');
if (!processedNodes.has(cleanNodeId)) {
const cleanName = node.show_name
.replace(/[^a-zA-Z0-9-\s\-_.,]/g, '')
.trim();
output += ` ${cleanNodeId}["${cleanName}"]\n`;
processedNodes.add(cleanNodeId);
}
if (parentId) {
const cleanParentId = parentId.replace(/[^a-zA-Z0-9_]/g, '_');
const duration = node.duration_ms ? `${node.duration_ms.toFixed(2)}ms` : '';
output += ` ${cleanParentId} -->|${duration}| ${cleanNodeId}\n`;
}
if (node.children && node.children.length > 0) {
node.children.forEach((child: TraceNode) => processNode(child, cleanNodeId));
}
}
if (!input) return output;
const rootNode: TraceNode = {
show_name: 'Trace Root',
span_id: 'root',
children: [] as TraceNode[]
};
if (input.data && Array.isArray(input.data)) {
rootNode.children = input.data;
} else if (Array.isArray(input)) {
rootNode.children = input;
} else {
rootNode.children = [input];
}
processNode(rootNode);
return output;
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/TraceThoughtChain/index.tsx
+111
View File
@@ -0,0 +1,111 @@
import React, { useState, useMemo, useEffect, useCallback } from 'react';
import { ThoughtChain } from '@ant-design/x';
import type { ThoughtChainProps, ThoughtChainItem } from '@ant-design/x';
import { Card, Typography, message } from 'antd';
import { fetchTraceData } from '@/api/trace';
const { Paragraph } = Typography;
interface TraceProps {
traceId?: string;
drawerVisible?: boolean;
}
type TraceNodeStatus = 'success' | 'pending' | 'error';
interface TraceNode {
id: string;
status?: TraceNodeStatus;
show_name: string;
children?: TraceNode[];
description?: string;
event_id: string;
summary?: string;
token_usage?: number;
input_tokens?: number;
output_tokens?: number;
use_tools?: string[];
}
const Trace: React.FC<TraceProps> = ({ traceId, drawerVisible }) => {
const [expandedKeys, setExpandedKeys] = useState<string[]>([]);
const [traceData, setTraceData] = useState<TraceNode[]>([]);
const fetchData = useCallback(async () => {
if (!traceId || !drawerVisible) return;
try {
const res = await fetchTraceData(traceId);
const validateStatus = (status?: string): TraceNodeStatus | undefined => {
return status === 'success' || status === 'pending' || status === 'error' ? (status as TraceNodeStatus) : undefined;
};
const validatedData = (res.data || []).map((item: TraceNode) => ({
...item,
status: validateStatus(item.status)
}));
setTraceData(validatedData);
// Expand the first node by default
if (validatedData?.[0]?.event_id) {
setExpandedKeys([validatedData[0].event_id]);
}
} catch (err) {
message.error('Failed to fetch trace data');
console.error(err);
}
}, [traceId, drawerVisible]);
useEffect(() => {
fetchData();
}, [fetchData]);
const renderNodeContent = useCallback(
(node: TraceNode) => (
<>
{node.token_usage && <p>token_usage: {node.token_usage}</p>}
{node.input_tokens && <p>input_tokens: {node.input_tokens}</p>}
{node.output_tokens && <p>output_tokens: {node.output_tokens}</p>}
{node.use_tools?.length && <p>use_tools: {node.use_tools.join(', ')}</p>}
{node.summary && (
<Typography>
<Paragraph>
<pre>{JSON.stringify(JSON.parse(node.summary), null, 2)}</pre>
</Paragraph>
</Typography>
)}
{node.children?.length ? <ThoughtChain items={convertToItems(node.children)} /> : null}
</>
),
[]
);
const convertToItems = useCallback(
(nodes: TraceNode[]): ThoughtChainItem[] => {
return nodes.map((node) => ({
key: node.event_id,
title: node.show_name,
description: node.event_id,
content: renderNodeContent(node),
status: node.status || 'pending'
}));
},
[renderNodeContent]
);
const items = useMemo(() => convertToItems(traceData), [traceData, convertToItems]);
const collapsible: ThoughtChainProps['collapsible'] = useMemo(() => {
return {
expandedKeys,
onExpand: (keys: string[]) => setExpandedKeys(keys)
};
}, [expandedKeys]);
return (
<Card style={{ width: 650 }}>
<ThoughtChain items={items} collapsible={collapsible} />
</Card>
);
};
export default Trace;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/TraceXY/CustomNode.tsx
+57
View File
@@ -0,0 +1,57 @@
import React from 'react';
import { Tooltip, Typography } from 'antd';
import { Position, Handle } from '@xyflow/react';
import type { CustomNodeData } from './TraceXY.types';
interface CustomNodeProps {
data: {
data: CustomNodeData;
};
isFirst?: boolean;
isLast?: boolean;
}
const CustomNode: React.FC<CustomNodeProps> = ({ data, isFirst, isLast }) => {
const nodeData: CustomNodeData = data || {};
const summary = nodeData.summary
? (typeof nodeData.summary === 'string'
? JSON.parse(nodeData.summary).summary
: nodeData.summary?.summary) || ''
: '';
const tooltipContent = nodeData.event_id ? (
<div className="Tooltipbox">
{summary.length > 100 ? summary : ''}
<div>{nodeData.event_id}</div>
</div>
) : null;
return (
<Tooltip title={tooltipContent} placement="bottom" className="Tooltipbox">
<div className="custom-node">
<Typography.Paragraph className="summary" ellipsis={{ rows: 4 }}>
{summary}
</Typography.Paragraph>
<div className="name">{nodeData.show_name || 'Unnamed Node'}</div>
{!isFirst && (
<Handle
type="target"
position={Position.Top}
/>
)}
{!isLast && (
<Handle
type="source"
position={Position.Bottom}
id="bottom"
/>
)}
{nodeData.sourceHandle?.includes('right') && (
<Handle type="source" position={Position.Right} id="right" />
)}
{nodeData.sourceHandle?.includes('left') && (
<Handle type="source" position={Position.Left} id="left" />
)}
</div>
</Tooltip>
);
};
export default CustomNode;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/TraceXY/TraceXY.types.ts
+23
View File
@@ -0,0 +1,23 @@
import type { Node, Edge } from '@xyflow/react';
export interface CustomNodeData {
show_name?: string;
event_id?: string;
summary?: string | { summary: string };
[key: string]: any;
}
export interface NodeData extends Node {
data: CustomNodeData;
type: string;
}
export interface EdgeData extends Edge {
[key: string]: any;
}
export interface TraceXYProps {
traceId?: string;
traceQuery?: string;
drawerVisible?: boolean;
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/TraceXY/index.less
+137
View File
@@ -0,0 +1,137 @@
.traceXYbox {
@box-shadow: 0 2px 8px rgba(0, 0, 0, 0.1);
@border-radius: 8px;
@transition: all 0.3s ease;
@text-color: #222;
@border-color: #d9d9d9;
@light-bg: #f8f9fa;
@node-bg: linear-gradient(135deg, #fff, #f8f8f8);
@primary-color: #1890ff;
width: 80%;
max-width: 700px;
height: 100%;
position: relative;
top: -20px;
background: @light-bg;
border-radius: @border-radius;
box-shadow: @box-shadow;
overflow: hidden;
.react-flow__node {
width: 300px;
min-width: 14.5%;
text-align: center;
max-width: 30%;
@node-shadow: 0 2px 6px rgba(0, 0, 0, 0.1);
@node-hover-shadow: 0 4px 12px rgba(0, 0, 0, 0.15);
@node-selected-shadow: 0 0 0 2px fade(@primary-color, 20%);
border: 1px solid @border-color;
border-radius: @border-radius;
// padding: 12px;
background: @node-bg;
box-shadow: @node-shadow;
font-size: 10px;
// transition: @transition;
margin-bottom: 25px;
&:hover {
box-shadow: @node-hover-shadow;
transform: translateY(-2px);
}
&-selected {
border-color: @primary-color;
box-shadow: @node-selected-shadow;
}
.desc {
margin: 0;
font-size: 12px;
}
}
.react-flow__handle{
background-color: #ccc;
}
.react-flow__edge-path {
stroke: #ddd;
stroke-width: 2;
animation: dashdraw 0.5s linear;
}
.react-flow__controls {
box-shadow: @box-shadow;
border-radius: 4px;
overflow: hidden;
}
.trace-id {
position: absolute;
bottom: 15px;
right: 15px;
background: rgba(255, 255, 255, 0.9);
padding: 6px 12px;
border-radius: 20px;
font-size: 10px;
color: #666;
box-shadow: 0 1px 4px rgba(0, 0, 0, 0.1);
border: 1px solid #eee;
}
@keyframes dashdraw {
from {
stroke-dashoffset: 100;
}
}
}
// .ant-tooltip-content {
// width: 420px;
// }
.Tooltipbox {
padding: 5px 8px;
.summary {
margin: 0;
line-height: 1.4;
font-size: 12px;
text-align: left;
}
pre {
white-space: pre-wrap;
word-break: break-word;
word-wrap: break-word;
}
}
.empty-state {
display: flex;
justify-content: center;
align-items: center;
height: 100%;
color: #888;
}
//edge click no changes
.virtual-node-edge,
.node-edge {
&:hover,
&-selected {
box-shadow: none !important;
transform: none !important;
border-color: transparent !important;
}
pointer-events: none !important;
}
//virtual-node hidden handle
// .react-flow__handle {
// background-color: #999;
// &.virtual-handle-target {
// width: 0px;
// height: 0px;
// min-width: 0;
// min-height: 0;
// border: none;
// }
// }
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/TraceXY/index.tsx
+126
View File
@@ -0,0 +1,126 @@
import React, { useState, useEffect, useCallback } from 'react';
import {
ReactFlow,
Background,
Controls,
ReactFlowProvider,
applyNodeChanges
} from '@xyflow/react';
import type { NodeChange } from '@xyflow/react';
import CustomNode from './CustomNode';
import '@xyflow/react/dist/style.css';
import { fetchTraceData } from '@/api/trace';
import { getLayoutedElements } from './layoutUtils';
import './index.less';
import type { TraceXYProps, NodeData, EdgeData } from './TraceXY.types';
const nodeTypes = {
customNode: CustomNode
};
const TraceXY: React.FC<TraceXYProps> = ({ traceId, drawerVisible }) => {
const [nodes, setNodes] = useState<NodeData[]>([]);
const [edges, setEdges] = useState<EdgeData[]>([]);
const [loading, setLoading] = useState(false);
const [error, setError] = useState<string | null>(null);
const onNodesChange = useCallback((changes: NodeChange[]) => {
setNodes((nds) => {
const updatedNodes = applyNodeChanges(changes, nds);
return updatedNodes.map((node) => ({
...node,
type: node.type || 'customNode',
data: (node as NodeData).data
})) as NodeData[];
});
}, []);
const processNodes = useCallback((rawNodes: any[] = []): NodeData[] => {
return rawNodes.map((node) => ({
id: node.span_id || node.id || '',
type: 'customNode',
position: node.position || { x: 0, y: 0 },
data: {
...node.data,
label: node.show_name,
summary: node.summary || '',
show_name: node.show_name,
event_id: node.event_id
}
}));
}, []);
const processEdges = useCallback((rawEdges: any[] = []): EdgeData[] => {
return rawEdges.map((edge) => ({
id: `${edge.source}-${edge.target}`,
source: edge.source,
target: edge.target,
className: 'node-edge',
type: 'smoothstep'
}));
}, []);
const loadAndLayoutElements = useCallback(async () => {
if (!traceId || !drawerVisible) return;
setLoading(true);
setError(null);
try {
const result = await fetchTraceData(traceId);
const nodesWithPosition = processNodes(result?.nodes || []);
const edgesWithId = processEdges(result?.edges || []);
const { nodes: layoutedNodes, edges: layoutedEdges } = await getLayoutedElements(
nodesWithPosition,
edgesWithId
);
setNodes(layoutedNodes);
setEdges(layoutedEdges);
} catch (err) {
setError('Failed to load trace data, please try again later.');
console.error('Failed to fetch and build trace elements:', err);
} finally {
setLoading(false);
}
}, [traceId, drawerVisible, processNodes, processEdges]);
useEffect(() => {
loadAndLayoutElements();
}, [loadAndLayoutElements]);
return (
<div className="traceXYbox" style={{ height: '100%', width: '100%' }}>
{loading && <div className="loading-indicator">Loading...</div>}
{error && <div className="error-message">{error}</div>}
{!loading && !error && nodes.length === 0 && (
<div className="empty-state">No trace data available</div>
)}
{nodes.length > 0 && (
<ReactFlow
nodes={nodes}
edges={edges}
nodeTypes={nodeTypes}
nodesDraggable
onNodesChange={onNodesChange}
snapToGrid={true}
snapGrid={[15, 15]}
fitView
minZoom={0.1}
maxZoom={2}
>
<Background gap={16} />
<Controls />
</ReactFlow>
)}
</div>
);
};
const TraceXYWithProvider: React.FC<TraceXYProps> = (props) => (
<ReactFlowProvider>
<TraceXY {...props} />
</ReactFlowProvider>
);
export default TraceXYWithProvider;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/TraceXY/layoutUtils.ts
+62
View File
@@ -0,0 +1,62 @@
import dagre from 'dagre';
const calculateEdgeLength = (
sourcePos: { x: number; y: number },
targetPos: { x: number; y: number }
): number => Math.hypot(targetPos.x - sourcePos.x, targetPos.y - sourcePos.y);
export const getLayoutedElements = (nodes: any[], edges: any[]) => {
const dagreGraph = new dagre.graphlib.Graph();
dagreGraph.setDefaultEdgeLabel(() => ({}));
dagreGraph.setGraph({
rankdir: 'TB',
nodesep: 50,
ranksep: 50
});
nodes.forEach((node) => {
dagreGraph.setNode(node.id, { width: 200, height: 100 });
});
edges.forEach((edge) => {
dagreGraph.setEdge(edge.source, edge.target);
});
dagre.layout(dagreGraph);
edges.forEach((edge) => {
const sourceNode = nodes.find((n) => n.id === edge.source);
const targetNode = nodes.find((n) => n.id === edge.target);
if (!sourceNode || !targetNode) return;
const sourcePos = dagreGraph.node(edge.source);
const targetPos = dagreGraph.node(edge.target);
const length = calculateEdgeLength(sourcePos, targetPos);
if (length > 300) {
const direction = targetPos.x > sourcePos.x ? 'right' : 'left';
sourceNode.data = sourceNode.data || {};
sourceNode.data.sourceHandle = sourceNode.data.sourceHandle || [];
sourceNode.data.sourceHandle.push(direction);
edge.sourceHandle = direction;
}
});
const updatedNodes = nodes.map((node) => {
const position = dagreGraph.node(node.id);
return {
...node,
position: {
x: position.x - 100,
y: position.y - 50
}
};
});
return {
nodes: updatedNodes,
edges: edges
};
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/Workspace/index.less
+95
View File
@@ -0,0 +1,95 @@
.workspacebox {
width: 100%;
box-sizing: border-box;
.btn {
color: #555;
height: 28px;
background-color: #daffd5;
border-radius: 10px;
position: fixed;
top: 14px;
right: 380px;
&:hover {
color: #555 !important;
border: 1px solid #daffd5 !important;
background-color: #f6ffed !important;
}
}
&.border,
.border {
border: 1px solid #c1c1c1;
border-radius: 10px;
}
.tabbox {
width: 100%;
box-sizing: border-box;
margin-bottom: 12px;
.num {
width: 30px;
height: 30px;
text-align: center;
line-height: 30px;
border-radius: 50%;
margin-right: 10px;
background-color: #efefef;
}
.tab {
width: 29%;
padding: 5px 10px;
cursor: pointer;
&.active {
.num {
background-color: #c4efa6;
color: #555;
}
}
.name {
font-size: 14px;
}
.desc {
font-size: 12px;
color: #999;
}
}
}
.listwrap {
background-color: #fafafa;
.title {
text-align: center;
line-height: 40px;
border-bottom: 1px solid #a7a7a7;
}
.listbox {
.list {
padding: 10px 14px;
.name {
font-size: 14px;
margin-bottom: 3px;
display: flex;
align-items: center;
&::before {
display: inline-block;
width: 12px;
height: 12px;
margin-right: 5px;
border-radius: 50%;
border: 1px solid #999;
background-color: #d8d8d8;
}
}
.desc,
.link {
color: #999;
font-size: 12px;
}
.desc {
margin-bottom: 0;
}
&:not(:last-child) {
border-bottom: 1px solid #a7a7a7;
}
}
}
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Drawer/Workspace/index.tsx
+114
View File
@@ -0,0 +1,114 @@
import { getWorkspaceArtifacts } from '@/api/workspace';
import { Image, Typography } from 'antd';
import React, { useEffect, useRef, useState } from 'react';
import type { ToolCardData } from '../../BubbleItem/utils';
import './index.less';
interface ArtifactItem {
snippet: string;
link: string;
key: string;
title: string;
content: string;
}
interface WorkspaceProps {
sessionId: string;
toolCardData: ToolCardData;
}
const Workspace: React.FC<WorkspaceProps> = ({ sessionId, toolCardData }) => {
const [artifacts, setArtifacts] = useState<ArtifactItem[]>([]);
const [imgUrl, setImgUrl] = useState<string | undefined>();
const isLinkListCard = toolCardData?.card_type === 'tool_call_card_link_list';
// 用于缓存上次的请求参数，避免重复调用
const lastRequestRef = useRef<{
sessionId: string;
artifactType: string;
artifactId: string;
} | null>(null);
useEffect(() => {
if (!toolCardData) return; // 如果没有 toolCardData，直接退出
const fetchWorkspaceArtifacts = async () => {
try {
const artifactType = toolCardData.artifacts?.[0]?.artifact_type;
const artifactId = toolCardData.artifacts?.[0]?.artifact_id;
if (!artifactType || !artifactId) {
console.warn('Invalid artifact data');
return;
}
// 检查是否与上次请求参数相同
const currentRequest = {
sessionId,
artifactType,
artifactId
};
if (lastRequestRef.current &&
lastRequestRef.current.sessionId === currentRequest.sessionId &&
lastRequestRef.current.artifactType === currentRequest.artifactType &&
lastRequestRef.current.artifactId === currentRequest.artifactId) {
// 参数相同，跳过重复请求
return;
}
// 更新缓存的请求参数
lastRequestRef.current = currentRequest;
const data = await getWorkspaceArtifacts(sessionId, {
artifact_types: [artifactType],
artifact_ids: [artifactId]
});
const content = data?.data?.[0]?.content;
if (isLinkListCard) {
setArtifacts(Array.isArray(content) ? content : []);
} else {
setImgUrl(content);
}
} catch (error) {
console.error('Failed to fetch workspace artifacts:', error);
}
};
fetchWorkspaceArtifacts();
}, [sessionId, toolCardData, isLinkListCard]);
const renderArtifactsList = () => (
<div className="listbox">
{artifacts.map((item, index) => (
<div className="list" key={index}>
<Typography.Link href={item?.link} target="_blank">
<Typography.Paragraph className="name" ellipsis={{ rows: 1 }}>
{item?.title}
</Typography.Paragraph>
<Typography.Paragraph className="desc" ellipsis={{ rows: 3 }}>
{item?.snippet}
</Typography.Paragraph>
<Typography.Paragraph className="link" ellipsis={{ rows: 1 }}>
{item?.link}
</Typography.Paragraph>
</Typography.Link>
</div>
))}
</div>
);
const renderImage = () => <Image preview={false} src={imgUrl} alt="Workspace Artifact" />;
return (
<div className="workspacebox">
<div className="border listwrap">
{isLinkListCard ? renderArtifactsList() : renderImage()}
</div>
</div>
);
};
export default Workspace;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Prompts/index.less
+12
View File
@@ -0,0 +1,12 @@
.chatPrompt{
.ant-prompts-label {
color: #000000e0 !important;
}
.ant-prompts-desc {
color: #000000a6 !important;
width: 100%;
}
.ant-prompts-icon {
color: #000000a6 !important;
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Prompts/index.tsx
+36
View File
@@ -0,0 +1,36 @@
import {
Prompts as AntDesignPrompts,
} from '@ant-design/x';
import './index.less';
interface IPromptsProps {
items: any[];
onItemClick: (item: any) => void;
className?: string;
}
const Prompts = (props: IPromptsProps) => {
const { items, onItemClick, className } = props;
return (
<AntDesignPrompts
items={items}
styles={{
item: {
flex: 1,
backgroundImage: 'linear-gradient(123deg, #e5f4ff 0%, #efe7ff 100%)',
borderRadius: 12,
border: 'none',
},
subItem: { background: '#ffffffa6' },
}}
onItemClick={(info) => {
onItemClick(info.data.description as string )
}}
className={className || "chatPrompt"}
/>
)
}
export default Prompts;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Welcome/index.less
+123
View File
@@ -0,0 +1,123 @@
.welcome-container {
display: flex;
justify-content: center;
background-color: #ffffff;
align-items: center;
position: relative;
bottom: 50px;
}
.content {
width: 100%;
display: flex;
flex-direction: column;
align-items: center;
}
.logo-title-container {
display: flex;
align-items: center;
justify-content: center;
gap: 12px;
margin-bottom: 8px;
img {
transition: transform 0.2s ease;
&:hover {
transform: scale(1.1);
}
}
.aworld-link {
color: inherit;
text-decoration: none;
transition: color 0.2s ease;
&:hover {
color: #1677ff;
}
}
}
.input-area {
position: relative;
width: 100%;
margin-top: 24px;
}
.text-input {
border-radius: 20px;
padding: 12px 50px 50px 20px;
border: 1px solid #d9d9d9;
font-size: 16px;
}
.submit-button {
position: absolute;
right: 12px;
bottom: 12px;
width: 40px !important;
height: 40px !important;
background-color: #000000;
border: none;
transition: opacity 0.2s;
}
.submit-button:hover,
.submit-button:focus {
background-color: rgba(0, 0, 0, 0.7) !important;
}
.submit-button:disabled {
opacity: 0.5;
cursor: not-allowed;
background-color: rgba(0, 0, 0, 0.1) !important;
}
.submit-button:disabled:hover,
.submit-button:disabled:focus {
opacity: 0.5;
background-color: rgba(0, 0, 0, 0.1) !important;
}
.controls-area {
width: 100%;
margin-top: 20px;
}
.model-select {
width: 100%;
// width: fit-content;
height: 44px;
border-radius: 50px;
.ant-select-selector {
border-radius: 12px !important;
padding-left: 12px !important;
border: 1px solid #d9d9d9 !important;
}
.ant-select-selection-item {
padding-right: 24px !important;
}
.ant-select-arrow {
right: 15px;
}
}
.select-item {
line-height: 30px;
small {
margin-left: 10px;
color: #b8b8b8;
font-weight: normal;
}
.icon-right {
color: #d9d9d9;
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/components/Welcome/index.tsx
+102
View File
@@ -0,0 +1,102 @@
import { ArrowUpOutlined, RightOutlined } from '@ant-design/icons';
import { Button, Col, Flex, Input, Row, Select, Typography } from 'antd';
import React, { useState } from 'react';
import logo from '../../../assets/aworld_logo.png';
import './index.less';
const { Title } = Typography;
interface WelcomeProps {
onSubmit: (value: string) => void;
models: Array<{ label: string; value: string }>;
selectedModel: string;
onModelChange: (value: string) => void;
modelsLoading: boolean;
}
const Welcome: React.FC<WelcomeProps> = ({
onSubmit,
models,
selectedModel,
onModelChange,
modelsLoading,
}) => {
const [inputValue, setInputValue] = useState('');
const handleKeyDown = (e: React.KeyboardEvent<HTMLTextAreaElement>) => {
if (e.key === 'Enter' && !e.shiftKey) {
e.preventDefault();
if (inputValue.trim()) onSubmit(inputValue);
}
};
return (
<div className="welcome-container">
<div className="content">
<Row justify="center">
<Col>
<div className="logo-title-container">
<img src={logo} alt="AWorld Logo" width="46" height="46" />
<Title level={1} style={{ margin: 0 }}>
<a
href="https://github.com/inclusionAI/AWorld"
target="_blank"
rel="noopener noreferrer"
className="aworld-link"
>
Hello{' '}AWorld
</a>
</Title>
</div>
</Col>
</Row>
<div className="input-area">
<Input.TextArea
value={inputValue}
onChange={(e) => setInputValue(e.target.value)}
onKeyDown={handleKeyDown}
placeholder="Ask or input / use skills"
autoSize={{ minRows: 3, maxRows: 5 }}
className="text-input"
/>
<Button
type="primary"
shape="circle"
onClick={() => {
if (inputValue.trim()) onSubmit(inputValue);
}}
icon={<ArrowUpOutlined />}
className="submit-button"
disabled={inputValue.trim() === ''}
/>
</div>
<div className="controls-area">
<Select
value={selectedModel}
onChange={onModelChange}
options={models}
loading={modelsLoading}
placeholder="Select a model"
className="model-select"
showSearch
filterOption={(input, option) =>
(option?.label ?? '').toLowerCase().includes(input.toLowerCase())
}
optionRender={(option) => (
<div className="select-item">
<Flex justify="space-between">
<div>
<strong>{option.label}</strong>
<small>{option.value}</small>
</div>
<RightOutlined className="icon-right" />
</Flex>
</div>
)}
/>
</div>
</div>
</div>
);
};
export default Welcome;
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/index.less
View File
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/components/CustomNode/index.less
+29
View File
@@ -0,0 +1,29 @@
.react-flow__node-customNode {
border-radius: 6px;
.custom-node {
// background: #fadddb;
// border: 2px solid #E6A5AD;
border-radius: 4px;
padding: 10px;
box-shadow: 0 2px 8px rgba(0, 0, 0, 0.1);
min-width: 200px;
max-width: 360px;
&-header {
font-weight: bold;
// color: #d58690;
border-bottom: 1px solid #eee;
padding-bottom: 5px;
margin-bottom: 5px;
}
&-content {
color: #666;
font-size: 12px;
.custom-node-io {
font-size: 12px;
margin-top: 5px;
}
}
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/components/CustomNode/index.tsx
+162
View File
@@ -0,0 +1,162 @@
import React, { useState } from 'react';
import { Handle, Position, useNodes, useReactFlow } from '@xyflow/react';
import type { Node, NodeProps } from '@xyflow/react';
import { deleteNode } from '@/pages/xyflow/utils/nodeUtils';
import { Tag, Drawer, Dropdown } from 'antd';
import { EllipsisOutlined, DeleteOutlined, CopyOutlined } from '@ant-design/icons';
import { NodeEditor } from '../NodeEditor';
interface NodeIOItem {
id: string;
label: string;
type: 'string' | 'number' | 'boolean';
defaultValue?: string;
}
interface CustomNodeData
extends Node<{
id: string;
label: string;
content?: React.ReactNode;
input?: NodeIOItem[];
output?: NodeIOItem[];
nodeType?: 'start' | 'end' | 'default';
}> {}
interface CustomNodeProps extends NodeProps<CustomNodeData> {}
export const CustomNode: React.FC<CustomNodeProps> = ({ id, data }) => {
const { label, content, input, output } = data;
const nodes = useNodes();
const reactFlowInstance = useReactFlow();
const { setNodes } = reactFlowInstance;
const [isDrawerOpen, setIsDrawerOpen] = useState(false);
const [pendingData, setPendingData] = useState<Partial<CustomNodeData['data']>>({});
const [editingData, setEditingData] = useState({
content: typeof content === 'string' ? content : '',
input: input || []
});
React.useEffect(() => {
setEditingData({
content: typeof content === 'string' ? content : '',
input: input || []
});
}, [content, input]);
const handleNodeClick = (e: React.MouseEvent) => {
e.stopPropagation();
setIsDrawerOpen(true);
};
const handleDrawerClose = (e: React.MouseEvent | React.KeyboardEvent) => {
if ('stopPropagation' in e) {
e.stopPropagation();
}
if (Object.keys(pendingData).length > 0) {
setNodes((nds) =>
nds.map((node) => {
if (node.id === id) {
return {
...node,
data: {
...node.data,
...pendingData
}
};
}
return node;
})
);
}
setIsDrawerOpen(false);
};
const renderIO = (title: string, items?: NodeIOItem[]) => {
return (
<div className="custom-node-io">
<span>{title}：</span>
{items?.map((item) => (
<Tag key={item.label}>
{item.type}.<strong>{item.label}</strong>
</Tag>
))}
</div>
);
};
return (
<div className="custom-node" onClick={handleNodeClick}>
<div className="custom-node-header">
<div style={{ display: 'flex', justifyContent: 'space-between', width: '100%' }}>
<span>{label}</span>
{data.nodeType !== 'start' && data.nodeType !== 'end' && (
<Dropdown
menu={{
items: [
{
key: 'delete',
label: '删除',
icon: <DeleteOutlined />,
onClick: (e) => {
e.domEvent.stopPropagation();
deleteNode(nodes, setNodes, id);
}
},
{
key: 'duplicate',
label: '创建副本',
icon: <CopyOutlined />,
onClick: (e) => {
e.domEvent.stopPropagation();
alert('暂不支持');
}
}
]
}}
trigger={['click']}
>
<EllipsisOutlined
style={{ cursor: 'pointer' }}
onClick={(e) => e.stopPropagation()}
/>
</Dropdown>
)}
</div>
</div>
<div className="custom-node-body">
<div className="custom-node-content">
<div>{editingData.content || 'Custom Node Content'}</div>
{data.nodeType !== 'end' && renderIO('输入', input)}
{data.nodeType !== 'start' && renderIO('输出', output)}
</div>
</div>
{data.nodeType !== 'start' && <Handle type="target" position={Position.Left} />}
{data.nodeType !== 'end' && <Handle type="source" position={Position.Right} />}
<Drawer
title={label}
placement="right"
closable={true}
maskClosable={true}
onClose={handleDrawerClose}
open={isDrawerOpen}
width={500}
keyboard={true}
>
<NodeEditor
node={{
id,
position: { x: 0, y: 0 },
data: { ...data, ...editingData }
}}
onUpdate={(updatedNode) => {
setPendingData((prev) => ({
...prev,
...updatedNode.data
}));
}}
onClose={() => handleDrawerClose({ stopPropagation: () => {} } as React.MouseEvent)}
/>
</Drawer>
</div>
);
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/components/FlowControls.tsx
+57
View File
@@ -0,0 +1,57 @@
import { Controls, ControlButton } from '@xyflow/react';
import { PlusOutlined, SaveOutlined, FolderOutlined, ReloadOutlined, GlobalOutlined, UndoOutlined, RedoOutlined } from '@ant-design/icons';
import type { FC } from 'react';
interface FlowControlsProps {
isStraightLine: boolean;
showMinimap: boolean;
onToggleLine: () => void;
onSave: () => void;
onLoad: () => void;
onAutoLayout: () => void;
onToggleMinimap: () => void;
onAddNode: () => void;
onUndo: () => void;
onRedo: () => void;
}
export const FlowControls: FC<FlowControlsProps> = ({
isStraightLine,
showMinimap,
onToggleLine,
onSave,
onLoad,
onAutoLayout,
onToggleMinimap,
onAddNode,
onUndo,
onRedo
}) => {
return (
<Controls style={{ left: '50%', transform: 'translateX(-50%)' }}>
<ControlButton onClick={onToggleLine} title={isStraightLine ? 'Switch to curved line' : 'Switch to straight line'}>
{isStraightLine ? '—' : '~'}
</ControlButton>
<ControlButton onClick={onSave} title="Save flowchart">
<SaveOutlined />
</ControlButton>
<ControlButton onClick={onLoad} title="Load flowchart">
<FolderOutlined />
</ControlButton>
<ControlButton onClick={onAutoLayout} title="Auto Layout">
<ReloadOutlined />
</ControlButton>
<ControlButton onClick={onUndo} title="Undo">
<UndoOutlined />
</ControlButton>
<ControlButton onClick={onRedo} title="Redo">
<RedoOutlined />
</ControlButton>
<ControlButton onClick={onToggleMinimap} title={showMinimap ? 'Hide minimap' : 'Show minimap'}>
<GlobalOutlined />
</ControlButton>
<ControlButton onClick={onAddNode} title="Add Node">
<PlusOutlined />
</ControlButton>
</Controls>
);
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/components/NodeEditor/index.less
+16
View File
@@ -0,0 +1,16 @@
.node-editor {
&-content {
margin-bottom: 16px;
}
&-header {
display: flex;
justify-content: space-between;
align-items: center;
margin-bottom: 16px;
}
&-collapse {
margin-top: 16px;
}
}
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/components/NodeEditor/index.tsx
+175
View File
@@ -0,0 +1,175 @@
import React, { useCallback, useMemo } from 'react';
import { Button, Input, Table, Select, Collapse } from 'antd';
import type { ColumnType } from 'antd/es/table';
import './index.less';
import { PlusOutlined } from '@ant-design/icons';
import type { Node } from '@xyflow/react';
const { Option } = Select;
interface NodeIOItem {
id: string;
label: string;
type: 'string' | 'number' | 'boolean';
defaultValue?: string;
}
interface NodeEditorProps {
node: Node<{
id: string;
label: string;
content?: React.ReactNode;
input?: NodeIOItem[];
output?: NodeIOItem[];
}>;
onUpdate: (node: Node) => void;
onClose: () => void;
}
export const NodeEditor: React.FC<NodeEditorProps> = ({ node, onUpdate }) => {
const [editingContent, setEditingContent] = React.useState(
typeof node.data.content === 'string' ? node.data.content : ''
);
const [editingInputs, setEditingInputs] = React.useState<NodeIOItem[]>(node.data.input || []);
React.useEffect(() => {
setEditingContent(typeof node.data.content === 'string' ? node.data.content : '');
setEditingInputs(node.data.input || []);
}, [node.data.content, node.data.input]);
const handleUpdate = useCallback(
(newData: Partial<typeof node.data>) => {
onUpdate({
...node,
data: {
...node.data,
...newData
}
});
},
[node, onUpdate]
);
const handleInputChange = useCallback(
<K extends keyof NodeIOItem>(index: number, field: K, value: NodeIOItem[K]) => {
const newInputs = [...editingInputs];
newInputs[index][field] = value;
setEditingInputs(newInputs);
handleUpdate({ input: newInputs });
},
[editingInputs, handleUpdate]
);
return (
<>
<div>{editingContent}</div>
<Collapse defaultActiveKey={['input']} bordered={false} className="node-editor-collapse">
<Collapse.Panel
header="输入"
key="input"
extra={
<Button
className="node-editor-collapse-btn"
icon={<PlusOutlined />}
onClick={(e) => {
e.stopPropagation();
const newInputs: NodeIOItem[] = [
...editingInputs,
{
id: `input-${Date.now()}-${Math.random().toString(36).substr(2, 9)}`,
label: '',
type: 'string' as const,
defaultValue: ''
}
];
setEditingInputs(newInputs);
handleUpdate({ input: newInputs });
}}
/>
}
>
<Table
dataSource={editingInputs}
rowKey={(record) => record.id}
pagination={false}
columns={useMemo<Array<ColumnType<NodeIOItem>>>(
() => [
{
title: '变量名',
dataIndex: 'label',
render: (text: string, _: NodeIOItem, index: number) => (
<Input
key={index}
value={text as 'string' | 'number' | 'boolean'}
onChange={(e) => handleInputChange(index, 'label', e.target.value)}
placeholder="Variable name"
/>
)
},
{
title: '变量值',
dataIndex: 'type',
render: (text: string, _: NodeIOItem, index: number) => (
<Select
value={text as 'string' | 'number' | 'boolean'}
style={{ width: '100%' }}
onChange={(value: 'string' | 'number' | 'boolean') =>
handleInputChange(index, 'type', value)
}
>
<Option value="string">String</Option>
<Option value="number">Number</Option>
<Option value="boolean">Boolean</Option>
</Select>
)
},
{
title: '',
dataIndex: 'defaultValue',
render: (text: string | undefined, _record: NodeIOItem, index: number) => (
<Input
value={text}
onChange={(e) => handleInputChange(index, 'defaultValue', e.target.value)}
placeholder="Default value"
/>
)
},
{
title: '',
render: (_text, _record: NodeIOItem, index: number) => (
<Button
danger
onClick={() => {
const newInputs = editingInputs.filter((_, i) => i !== index);
setEditingInputs(newInputs);
handleUpdate({ input: newInputs });
}}
>
Delete
</Button>
)
}
],
[handleInputChange, editingInputs, handleUpdate]
)}
/>
</Collapse.Panel>
</Collapse>
<Collapse defaultActiveKey={['output']} bordered={false} className="node-editor-collapse">
<Collapse.Panel header="输出" key="output">
<Input.TextArea
value={editingContent || ''}
onChange={(e) => {
const newValue = e.target.value;
setEditingContent(newValue);
handleUpdate({ content: newValue });
}}
placeholder="Enter node content"
autoSize={{ minRows: 3, maxRows: 10 }}
/>
</Collapse.Panel>
</Collapse>
</>
);
};
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/constants.ts
+39
View File
@@ -0,0 +1,39 @@
import { Position } from '@xyflow/react';
export const initialNodes = [
{
id: '1',
type: 'customNode',
data: {
label: 'Start Node',
nodeType: 'start',
content: '开始节点，用于设定工作流启动变量',
input: [
// { label: 'name', type: 'int' },
// { label: 'age', type: 'Boolean' }
]
},
position: { x: 0, y: 0 },
style: { background: '#E8F8F5', border: '2px solid #1ABC9C', color: '#16A085' },
sourcePosition: Position.Right,
targetPosition: Position.Left
},
{
id: '2',
type: 'customNode',
data: {
label: 'End Node',
nodeType: 'end',
content: '结束节点，用于返回工作流运行结果',
output: [
// { label: 'name', type: 'int' },
// { label: 'age', type: 'Boolean' }
]
},
position: { x: 400, y: 0 },
style: { background: '#FEF9E7', border: '2px solid #F7DC6F', color: '#D4AC0D' },
sourcePosition: Position.Right,
targetPosition: Position.Left
}
];
export const initialEdges = [];
chapter9/gaia-experience/AWorld/aworld/cmd/web/webui/src/pages/xyflow/index.less
+18
View File
@@ -0,0 +1,18 @@
@import '@xyflow/react/dist/style.css';
@import './components/CustomNode/index.less';
.react-flow__controls {
display: flex;
flex-direction: row;
gap: 8px;
padding: 8px;
border-radius: 4px;
box-shadow: 0 1px 4px rgba(0, 0, 0, 0.2);
}
.react-flow__controls-button {
width: 32px;
height: 32px;
padding: 6px;
font-size: 16px;
}

Some files were not shown because too many files have changed in this diff Show More