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
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s
Build latest book artifacts / build (push) Canceled after 0s
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s
This commit is contained in:
@@ -0,0 +1,9 @@
|
||||
from aworld.prompt.base import (
|
||||
Prompt,
|
||||
create_prompt,
|
||||
)
|
||||
|
||||
__all__ = [
|
||||
"Prompt",
|
||||
"create_prompt",
|
||||
]
|
||||
@@ -0,0 +1,131 @@
|
||||
import logging
|
||||
import re
|
||||
from typing import Dict, Any, Optional, Callable, List
|
||||
|
||||
from aworld.core.context.base import Context
|
||||
from aworld.core.context.prompts.dynamic_variables import create_simple_field_getter, create_multiple_field_getters, \
|
||||
get_field_values_from_list
|
||||
|
||||
# Import system_prompt directly from the module
|
||||
try:
|
||||
from aworld.prompt.templates.prompt import system_prompt as default_system_prompt
|
||||
except ImportError:
|
||||
default_system_prompt = ""
|
||||
|
||||
class Prompt:
|
||||
"""
|
||||
Prompt processing class, responsible for loading templates and rendering variables
|
||||
"""
|
||||
|
||||
def __init__(self, template: str = None, context: Context = None) -> None:
|
||||
"""
|
||||
Initialize Prompt class
|
||||
|
||||
Args:
|
||||
template: Template content. If None, will load default system prompt.
|
||||
"""
|
||||
if template is None:
|
||||
self.template_content = default_system_prompt
|
||||
else:
|
||||
self.template_content = template
|
||||
|
||||
# Extract variables from the template
|
||||
self.prompt_variables = self._extract_variables(self.template_content)
|
||||
self.context = context
|
||||
|
||||
def _extract_variables(self, template_content: str) -> List[str]:
|
||||
"""
|
||||
Extract variables from the template
|
||||
|
||||
Args:
|
||||
template_content: Template content
|
||||
|
||||
Returns:
|
||||
List of variables
|
||||
"""
|
||||
# Use regular expression to extract {{variable}} format variables
|
||||
pattern = r'\{\{([a-zA-Z0-9_]+)\}\}'
|
||||
if not template_content:
|
||||
return []
|
||||
try:
|
||||
matches = re.findall(pattern, template_content)
|
||||
# Remove duplicates
|
||||
return list(set(matches))
|
||||
except Exception as e:
|
||||
logging.warning(f"Error extracting variables from template: {str(e)}")
|
||||
return []
|
||||
|
||||
|
||||
def _get_variable_values_from_api(self, variables: List[str]) -> Dict[str, Any]:
|
||||
"""
|
||||
Get variable values from API
|
||||
|
||||
Args:
|
||||
variables: List of variables
|
||||
|
||||
Returns:
|
||||
Dictionary of variable values, key is variable name, value is variable value
|
||||
"""
|
||||
# This is a mock implementation, should be replaced with actual API calls in real use
|
||||
result = {}
|
||||
context = self.context
|
||||
#getter = create_multiple_field_getters(variables, context)
|
||||
if context and variables:
|
||||
field_paths = variables
|
||||
try:
|
||||
result = get_field_values_from_list(context=context, field_paths=field_paths,default="")
|
||||
except Exception as e:
|
||||
logging.warning(f"Error getting variable values from API: {str(e)}")
|
||||
return None
|
||||
return result
|
||||
|
||||
def get_prompt(self, variables: Dict[str, Any] = None, variable_resolver: Optional[Callable[[List[str]], Dict[str, Any]]] = None) -> str:
|
||||
"""
|
||||
Render template, replace variables
|
||||
|
||||
Args:
|
||||
variables: Optional variable dictionary, if provided, will override API values
|
||||
variable_resolver: Optional variable resolver function, used to get variable values, higher priority than built-in API method
|
||||
|
||||
Returns:
|
||||
Rendered content
|
||||
"""
|
||||
if variables is None:
|
||||
variables = {}
|
||||
|
||||
# Get variable values
|
||||
if variable_resolver:
|
||||
# If custom variable resolver is provided, use it
|
||||
resolved_variables = variable_resolver(self.prompt_variables)
|
||||
else:
|
||||
# Otherwise use the built-in API method
|
||||
resolved_variables = self._get_variable_values_from_api(self.prompt_variables)
|
||||
|
||||
# Override API values with provided variables
|
||||
for var_name, var_value in variables.items():
|
||||
if var_value is not None: # Only override when value is not None
|
||||
resolved_variables[var_name] = var_value
|
||||
|
||||
# Render template
|
||||
rendered_content = self.template_content
|
||||
|
||||
# Replace variables
|
||||
for var_name, var_value in resolved_variables.items():
|
||||
if var_value:
|
||||
placeholder = f"{{{{{var_name}}}}}"
|
||||
rendered_content = rendered_content.replace(placeholder, str(var_value))
|
||||
|
||||
return rendered_content
|
||||
|
||||
# Simplified API, only keep the prompt creation functionality
|
||||
def create_prompt(template_content: str) -> Prompt:
|
||||
"""
|
||||
Create Prompt from string
|
||||
|
||||
Args:
|
||||
template_content: Template content
|
||||
|
||||
Returns:
|
||||
Prompt object
|
||||
"""
|
||||
return Prompt(template_content)
|
||||
@@ -0,0 +1,56 @@
|
||||
system_prompt="""
|
||||
<system_instruction>
|
||||
You are an advanced AI assistant powered by a large language model, operating within the AWorld framework. Your purpose is to assist users with a wide range of tasks by leveraging your knowledge and capabilities.
|
||||
|
||||
## Core Capabilities
|
||||
You are designed to:
|
||||
1. **Understand and respond** to user queries with accurate, helpful information
|
||||
2. **Reason** through complex problems step by step
|
||||
3. **Generate** creative content based on user requirements
|
||||
4. **Execute** tasks using available tools when appropriate
|
||||
5. **Learn** from interactions to better serve users over time
|
||||
|
||||
## Task Approach
|
||||
When addressing user requests:
|
||||
1. **Analyze the request** carefully to understand the user's intent and needs
|
||||
2. **Plan your approach** by breaking down complex tasks into manageable steps
|
||||
3. **Use available tools** when necessary to gather information or perform actions
|
||||
4. **Provide clear explanations** of your reasoning and actions
|
||||
5. **Verify your responses** for accuracy, relevance, and completeness before delivering them
|
||||
|
||||
## Communication Guidelines
|
||||
1. **Be concise** but thorough in your responses
|
||||
2. **Use appropriate formatting** to enhance readability (headings, bullet points, code blocks)
|
||||
3. **Adapt your tone** to match the context and user's communication style
|
||||
4. **Acknowledge limitations** when you're uncertain or when a request is beyond your capabilities
|
||||
5. **Seek clarification** when user requests are ambiguous or incomplete
|
||||
|
||||
## Ethical Guidelines
|
||||
1. **Prioritize user safety** and well-being in all interactions
|
||||
2. **Respect privacy** and confidentiality of user information
|
||||
3. **Provide balanced perspectives** on controversial topics
|
||||
4. **Decline** to assist with harmful, illegal, or unethical requests
|
||||
5. **Be transparent** about your limitations and capabilities
|
||||
|
||||
## Tool Usage
|
||||
When using tools:
|
||||
1. **Select the appropriate tool** based on the task requirements
|
||||
2. **Explain your reasoning** for using a particular tool
|
||||
3. **Use tools efficiently** to minimize unnecessary operations
|
||||
4. **Interpret tool outputs** accurately and incorporate them into your response
|
||||
5. **Handle errors gracefully** if tools fail or return unexpected results
|
||||
|
||||
## Available Tools:
|
||||
{{available_tools}}
|
||||
|
||||
## Task Description:
|
||||
{{task_description}}
|
||||
|
||||
##Test Context variables:
|
||||
{{current_time}}
|
||||
{{session_id}}
|
||||
{{is_task}}
|
||||
{{trajectories}}
|
||||
|
||||
</system_instruction>
|
||||
"""
|
||||
Reference in New Issue
Block a user