Update agent.py

This commit is contained in:
2026-06-04 16:15:21 +00:00
parent 03a852d5ad
commit 1ae78b521f
+99 -165
View File
@@ -1,187 +1,121 @@
"""Deep Agent implementation based on LangGraph. """Deep Agent implementation based on LangGraph and LangChain.
This agent can: The module exposes two public functions:
1. Search the web for information using DuckDuckGo API.
2. Create virtual files in memory.
3. At the end of the run, persist virtual files to the real file system.
The code follows the requirements: * ``create_agent`` returns a :class:`langgraph.graph.StateGraph` that can be
- Uses langchain>=1.0.0 and langgraph>=1.0.0. executed.
- Correct imports for text splitters, Chroma, Ollama embeddings and chat model. * ``create_agent_executor`` returns an :class:`langchain.agents.AgentExecutor`
- Implements `create_agent` and `create_agent_executor` functions. that can be used directly.
- Uses the `write_file` tool to write virtual files.
Both functions lazily import the heavy LangChain/LangGraph dependencies so
that the module can be imported even if those packages are not installed.
""" """
import pathlib from __future__ import annotations
from typing import Dict, List, Any, Optional
# LangChain imports from typing import Any
from langchain_ollama import ChatOllama, OllamaEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_chroma import Chroma
# LangGraph imports # Lightweight imports can be imported eagerly
from langgraph.graph import Graph, State from virtual_fs import VirtualFileSystem
from tools import WebSearch, CreateVirtualFile, ExportVirtualFiles
# Requests for web search # Shared virtual file system instance used by all tools
import requests vfs = VirtualFileSystem()
# ----------------------------- # Define the tools
# Configuration web_search_tool = WebSearch()
# ----------------------------- create_file_tool = CreateVirtualFile(vfs)
OLLAMA_MODEL = "llama3" export_files_tool = ExportVirtualFiles(vfs)
EMBEDDINGS_MODEL = "llama3"
SEARCH_URL = "https://api.duckduckgo.com/"
OUTPUT_DIR = pathlib.Path("output_files")
OUTPUT_DIR.mkdir(exist_ok=True)
# ----------------------------- # ---------------------------------------------------------------------------
# Helper functions # Public API
# ----------------------------- # ---------------------------------------------------------------------------
def search_web(query: str) -> str: def create_agent() -> Any:
"""Return a short summary of the search results using DuckDuckGo.""" """Return a LangGraph graph.
params = {
"q": query, The function performs a lazy import of :mod:`langgraph` and related
"format": "json", dependencies. If the imports fail, a clear ``ImportError`` is raised.
"no_html": 1, """
"skip_disambig": 1,
}
try: try:
r = requests.get(SEARCH_URL, params=params, timeout=10) from langgraph.graph import StateGraph
r.raise_for_status() from langgraph.prebuilt import create_react_agent
data = r.json() from langchain_ollama import ChatOllama
abstract = data.get("AbstractText") from langchain_core.prompts import ChatPromptTemplate
if abstract: from langchain_core.output_parsers import StrOutputParser
return abstract except Exception as exc: # pragma: no cover
topics = data.get("RelatedTopics", []) raise ImportError(
snippets = [t.get("Text", "") for t in topics if "Text" in t] "Failed to import LangGraph/LangChain dependencies. Ensure that the\n"
return "\n".join(snippets[:5]) "required packages are installed and the environment is correctly\n"
except Exception as e: "configured."
return f"Error during search: {e}" ) from exc
# ----------------------------- # LLM and prompt
# State definition llm = ChatOllama(model="llama3.1")
# ----------------------------- prompt = ChatPromptTemplate.from_messages(
class AgentState(State): [
query: str ("system", "You are a helpful assistant that can search the web, create virtual files, and export them to disk."),
virtual_files: Dict[str, str] ("human", "{input}"),
history: List[Dict[str, str]] ]
answer: Optional[str] = None )
parser = StrOutputParser()
# ----------------------------- # Build a simple react agent using LangGraph prebuilt
# Tool: write_file react_agent = create_react_agent(
# ----------------------------- llm=llm,
tools=[web_search_tool, create_file_tool, export_files_tool],
def write_file_tool(state: AgentState, file_name: str, content: str) -> AgentState: prompt=prompt,
new_files = state.virtual_files.copy() output_parser=parser,
new_files[file_name] = content )
return state.copy(update={"virtual_files": new_files})
# -----------------------------
# Tool: search
# -----------------------------
def search_tool(state: AgentState, query: str) -> AgentState:
result = search_web(query)
new_history = state.history + [{"role": "tool", "name": "search", "content": result}]
return state.copy(update={"history": new_history})
# -----------------------------
# Agent logic
# -----------------------------
def create_agent() -> Graph:
llm = ChatOllama(model=OLLAMA_MODEL, temperature=0.7)
embeddings = OllamaEmbeddings(model=EMBEDDINGS_MODEL)
splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
def agent(state: AgentState) -> Dict[str, Any]:
# Build prompt from history
messages = []
for msg in state.history:
if msg["role"] == "user":
messages.append({"role": "user", "content": msg["content"]})
elif msg["role"] == "assistant":
messages.append({"role": "assistant", "content": msg["content"]})
elif msg["role"] == "tool":
messages.append({"role": "assistant", "content": f"[Tool: {msg['name']}] {msg['content']}"})
messages.append({"role": "assistant", "content": f"User query: {state.query}"})
response = llm.invoke(messages)
text = response.content.strip()
if text.upper().startswith("ANSWER:"):
answer = text[7:].strip()
new_history = state.history + [{"role": "assistant", "content": answer}]
return {"final_answer": answer, "history": new_history}
elif text.upper().startswith("SEARCH:"):
query = text[7:].strip()
return {"search_query": query}
elif text.upper().startswith("WRITE:"):
try:
rest = text[6:].strip()
file_name, content = rest.split("|", 1)
return {"write_file": {"file_name": file_name.strip(), "content": content.strip()}}
except Exception:
return {"final_answer": "Could not parse WRITE command."}
else:
answer = text
new_history = state.history + [{"role": "assistant", "content": answer}]
return {"final_answer": answer, "history": new_history}
graph = Graph()
graph.add_node("agent", agent)
graph.add_node("search", search_tool)
graph.add_node("write_file", write_file_tool)
def final_answer_node(state: AgentState) -> AgentState:
for msg in reversed(state.history):
if msg["role"] == "assistant":
state = state.copy(update={"answer": msg["content"]})
break
return state
graph.add_node("final_answer", final_answer_node)
graph.add_edge("agent", "search", condition=lambda out: "search_query" in out)
graph.add_edge("agent", "write_file", condition=lambda out: "write_file" in out)
graph.add_edge("agent", "final_answer", condition=lambda out: "final_answer" in out)
graph.add_edge("search", "agent", condition=lambda out: True)
graph.add_edge("write_file", "agent", condition=lambda out: True)
graph.set_start("agent")
graph.set_end("final_answer")
# Create the graph
graph = StateGraph()
graph.add_node("react_agent", react_agent)
graph.set_entry_point("react_agent")
graph.set_finish_point("react_agent")
return graph return graph
# -----------------------------
# Executor helper
# -----------------------------
def create_agent_executor() -> Graph: def create_agent_executor() -> Any:
return create_agent() """Return an :class:`langchain.agents.AgentExecutor`.
# ----------------------------- The function lazily imports the required classes. It is useful for
# Main execution quick experimentation and for environments where the full graph is
# ----------------------------- unnecessary.
if __name__ == "__main__": """
import argparse try:
parser = argparse.ArgumentParser(description="Run the deep agent.") from langchain.agents import AgentExecutor
parser.add_argument("query", type=str, help="User query to process") from langgraph.prebuilt import create_react_agent
args = parser.parse_args() from langchain_ollama import ChatOllama
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
except Exception as exc: # pragma: no cover
raise ImportError(
"Failed to import LangChain dependencies for the AgentExecutor."
) from exc
graph = create_agent_executor() llm = ChatOllama(model="llama3.1")
initial_state = AgentState(query=args.query, virtual_files={}, history=[{"role": "user", "content": args.query}]) prompt = ChatPromptTemplate.from_messages(
[
("system", "You are a helpful assistant that can search the web, create virtual files, and export them to disk."),
("human", "{input}"),
]
)
parser = StrOutputParser()
final_state = graph.invoke(initial_state) # Build the react agent
react_agent = create_react_agent(
llm=llm,
tools=[web_search_tool, create_file_tool, export_files_tool],
prompt=prompt,
output_parser=parser,
)
print("\n=== Final Answer ===\n") return AgentExecutor.from_agent_and_tools(
print(final_state.answer if final_state.answer else "No answer produced.") agent=react_agent,
tools=[web_search_tool, create_file_tool, export_files_tool],
verbose=True,
)
for fname, content in final_state.virtual_files.items(): # Expose public names for tests
out_path = OUTPUT_DIR / fname __all__ = ["create_agent", "create_agent_executor", "vfs"]
out_path.parent.mkdir(parents=True, exist_ok=True)
with open(out_path, "w", encoding="utf-8") as f:
f.write(content)
print(f"Virtual file written to {out_path}")