File size: 9,942 Bytes
e33834b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 | # Standard libraries
import json
import os
from dotenv import load_dotenv
from typing import Dict, List, Any, Optional, Annotated
from typing_extensions import TypedDict
# Langchain and langgraph
from langchain_core.messages import HumanMessage, AIMessage, SystemMessage, ToolMessage, AnyMessage
from langgraph.graph import StateGraph, START, END, add_messages
from langgraph.prebuilt import ToolNode, tools_condition
# Custom modules
from prompts import MAIN_SYSTEM_PROMPT, QUESTION_DECOMPOSITION_PROMPT, TOOL_USE_INSTRUCTION, EXECUTION_INSTRUCTION
from utils import check_api_keys, setup_llm
from tools import (
calculator_tool, extract_text_from_image, transcribe_audio, execute_python_code,
read_file, web_search, wikipedia_search, arxiv_search, chess_board_image_analysis,
find_phrase_in_text, download_youtube_audio, web_content_extract, analyse_tabular_data
)
# AGENT STATE
class AgentState(TypedDict):
task_id: Optional[str]
file_name: Optional[str]
file_type: Optional[str]
file_path: Optional[str]
question_decomposition: Optional[str]
messages: Annotated[list[AnyMessage], add_messages]
tool_results: Dict
error_message: Optional[str]
# WORKFLOW CREATION
def create_workflow_for_final_agent():
"""
Creates and compiles the LangGraph workflow.
"""
llm_agent_management, llm_question_decomposition, _, _, _ = setup_llm()
tools = [
web_search,
web_content_extract,
wikipedia_search,
calculator_tool,
extract_text_from_image,
transcribe_audio,
execute_python_code,
read_file,
arxiv_search,
chess_board_image_analysis,
find_phrase_in_text,
download_youtube_audio,
analyse_tabular_data
]
llm_agent_management_with_tools = llm_agent_management.bind_tools(tools)
# Define nodes
def question_decomposition_node(state: AgentState):
new_state = state.copy()
messages = new_state.get("messages", []) # Use .get for safety, ensure it's a list
question = None
for msg in messages:
if isinstance(msg, HumanMessage):
question = msg.content
break
if not question:
new_state["error_message"] = "No question found for decomposition."
# Ensure messages list exists even if we return early
if "messages" not in new_state or not isinstance(new_state["messages"], list):
new_state["messages"] = []
return new_state
question_decomposition_prompt_messages = [
SystemMessage(content=QUESTION_DECOMPOSITION_PROMPT),
HumanMessage(content=f"Decompose this question: {question}")
]
question_decomposition_object = llm_question_decomposition.invoke(question_decomposition_prompt_messages)
question_decomposition_response = question_decomposition_object.content
new_state["question_decomposition"] = question_decomposition_response
# Ensure messages list exists
if "messages" not in new_state or not isinstance(new_state["messages"], list):
new_state["messages"] = []
return new_state
def call_model_node(state: AgentState):
new_state = state.copy()
messages = new_state.get("messages", []) # Use .get for safety
question_decomposition = new_state.get("question_decomposition", "")
llm_messages = list(messages) # Ensure it's a mutable list
add_decomposition = question_decomposition and (not llm_messages or not isinstance(llm_messages[-1], ToolMessage))
if add_decomposition:
decomposition_message = SystemMessage(content=f"Question decomposition: {question_decomposition}\\nUse this analysis to guide your actions.")
llm_messages.append(decomposition_message)
response = llm_agent_management_with_tools.invoke(llm_messages)
# Ensure new_state["messages"] exists and is a list before extending
current_messages = new_state.get("messages", [])
if not isinstance(current_messages, list):
current_messages = []
new_state["messages"] = current_messages + [response]
return new_state
workflow = StateGraph(AgentState)
workflow.add_node("decomposition", question_decomposition_node)
workflow.add_node("agent", call_model_node)
workflow.add_node("tools", ToolNode(tools))
workflow.add_edge(START, "decomposition")
workflow.add_edge("decomposition", "agent")
workflow.add_conditional_edges("agent", tools_condition)
workflow.add_edge("tools", "agent")
return workflow.compile()
class FinalAgent:
def __init__(self):
print("FinalAgent initializing...")
load_dotenv()
if not os.path.exists('.config'):
print("Warning: .config file not found. Using default values or expecting environment variables.")
self.config = {} # Default to empty config
else:
with open('.config', 'r') as f:
self.config = json.load(f)
self.base_url = self.config.get('BASE_URL', os.getenv('BASE_URL'))
self.debug_mode = self.config.get('DEBUG_MODE', str(os.getenv('DEBUG_MODE', 'False')).lower() == 'true')
if not check_api_keys():
# check_api_keys itself prints messages
raise ValueError("API keys are missing or invalid. Please set the required environment variables.")
self.workflow = create_workflow_for_final_agent()
print("FinalAgent initialized successfully.")
def __call__(self, question: str, task_id: Optional[str] = None) -> str:
print(f"FinalAgent received question for task_id '{task_id}': {question[:100]}...")
initial_messages = [
SystemMessage(content=MAIN_SYSTEM_PROMPT + "\\n\\n" + TOOL_USE_INSTRUCTION + "\\n\\n" + EXECUTION_INSTRUCTION),
HumanMessage(content=question)
]
initial_state: AgentState = {
"messages": initial_messages,
"task_id": task_id,
"file_name": None,
"file_path": None,
"file_type": None,
"question_decomposition": None,
"tool_results": {},
"error_message": None
}
try:
result_state = self.workflow.invoke(initial_state)
except Exception as e:
print(f"Error invoking workflow for task {task_id}: {e}")
import traceback
traceback.print_exc()
return f"AGENT ERROR: Failed to process question due to an internal error: {e}"
messages = result_state.get("messages", [])
final_answer = ""
if not messages:
print(f"No messages found in the result state for task {task_id}.")
return "AGENT ERROR: No messages returned by the agent."
for msg in reversed(messages):
if hasattr(msg, "content") and msg.content:
content = msg.content
if isinstance(content, str):
if "FINAL ANSWER:" in content:
final_answer = content.split("FINAL ANSWER:", 1)[1].strip()
break
elif isinstance(msg, AIMessage):
# If it's an AIMessage and no "FINAL ANSWER:" has been found yet,
# tentatively set it. This will be overridden if a "FINAL ANSWER:" is found later.
if not final_answer:
final_answer = content
# If after checking all messages, final_answer is still from a non-"FINAL ANSWER:" AIMessage, that's our best guess.
# If final_answer is empty, it means no AIMessage with content or "FINAL ANSWER:" was found.
if not final_answer: # This means no "FINAL ANSWER:" and no AIMessage content was suitable
final_answer = "AGENT ERROR: Could not extract a final answer from the agent's messages."
print(f"Could not extract final answer for task {task_id}. Messages: {messages}")
print(f"FinalAgent returning answer for task_id '{task_id}': {final_answer[:100]}...")
return final_answer
if __name__ == '__main__':
print("Running a simple test for FinalAgent...")
if not os.path.exists('.config'):
print("Creating a dummy .config file for testing.")
with open('.config', 'w') as f:
json.dump({"DEBUG_MODE": True, "BASE_URL": "http://localhost:8000"}, f)
# Check for .env and API keys
if not load_dotenv(): # Attempts to load .env and returns True if successful
print("Warning: .env file not found or failed to load. API keys might be missing.")
if not (os.getenv("OPENAI_API_KEY") or os.getenv("DEEPSEEK_API_KEY") or os.getenv("TAVILY_API_KEY")):
print("\\nWARNING: Required API key no found in environment variables (OPENAI_API_KEY, DEEPSEEK_API_KEY, TAVILY_API_KEY).")
print("The agent will likely fail to initialize or run properly without at least one.")
print("Please set them in your .env file or environment for testing.\\n")
try:
agent = FinalAgent()
test_question = "What is the capital of France? And what is the weather like there today?"
print(f"Test Question 1: {test_question}")
answer = agent(test_question, task_id="test_001")
print(f"Test Answer 1: {answer}")
test_question_calc = "What is 123 * 4 / 2 + 6?"
print(f"\\nTest Question 2 (Calc): {test_question_calc}")
answer_calc = agent(test_question_calc, task_id="test_002")
print(f"Test Answer 2 (Calc): {answer_calc}")
except ValueError as ve:
print(f"Initialization Error: {ve}")
except Exception as e:
print(f"An error occurred during the test: {e}")
import traceback
traceback.print_exc() |