145 lines
5.2 KiB
Python
145 lines
5.2 KiB
Python
import os
|
||
import asyncio
|
||
import json
|
||
from pathlib import Path
|
||
from typing import List
|
||
|
||
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
|
||
from langchain_chroma import Chroma
|
||
from langchain_core.documents import Document
|
||
from langchain_core.messages import HumanMessage
|
||
from langchain.tools import tool
|
||
from deepagents import create_deep_agent
|
||
from deepagents.backends import FilesystemBackend, LocalShellBackend, CompositeBackend
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 1. Настройка LLM и Embeddings (OpenRouter)
|
||
# ---------------------------------------------------------------------------
|
||
llm = ChatOpenAI(
|
||
model="openai/gpt-oss-20b:free",
|
||
base_url="https://openrouter.ai/api/v1",
|
||
api_key=os.getenv("OPENAI_API_KEY"),
|
||
temperature=0.0,
|
||
)
|
||
|
||
embeddings = OpenAIEmbeddings(
|
||
model="text-embedding-3-small",
|
||
base_url="https://openrouter.ai/api/v1",
|
||
api_key=os.getenv("OPENAI_API_KEY"),
|
||
)
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 2. ChromaDB: загрузка .md файлов и поиск
|
||
# ---------------------------------------------------------------------------
|
||
CHROMA_PATH = Path("./chroma_faq")
|
||
CHROMA_COLLECTION = "faq_collection"
|
||
|
||
vector_store = Chroma(
|
||
collection_name=CHROMA_COLLECTION,
|
||
embedding_function=embeddings,
|
||
persist_directory=str(CHROMA_PATH),
|
||
)
|
||
|
||
# Если коллекция пуста, загрузим данные из data/*.md
|
||
if not vector_store.get_collection().list_documents():
|
||
md_files = list(Path("data").glob("*.md"))
|
||
docs: List[Document] = []
|
||
for f in md_files:
|
||
text = f.read_text(encoding="utf-8")
|
||
docs.append(Document(page_content=text, metadata={"source": f.name}))
|
||
vector_store.add_documents(docs)
|
||
vector_store.persist()
|
||
|
||
@tool
|
||
def search_course_docs(query: str, k: int = 3) -> str:
|
||
"""Search the local FAQ collection for relevant passages."""
|
||
results = vector_store.similarity_search(query, k=k)
|
||
if not results:
|
||
return "No relevant information found in the course materials."
|
||
return "\n\n---\n\n".join(f"{doc.metadata.get('source', 'unknown')}\n{doc.page_content}" for doc in results)
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 3. MCP‑style tool (mocked via local JSON file)
|
||
# ---------------------------------------------------------------------------
|
||
META_JSON = Path("course_meta.json")
|
||
if not META_JSON.exists():
|
||
# Создаём простую статическую мета‑информацию
|
||
META_JSON.write_text(json.dumps({
|
||
"schedule": {
|
||
"Monday": "Lecture 1",
|
||
"Wednesday": "Lecture 2",
|
||
"Friday": "Lab"
|
||
},
|
||
"instructor": "Dr. Example"
|
||
}, indent=2))
|
||
|
||
@tool
|
||
def fetch_course_meta(query: str) -> str:
|
||
"""Return course metadata that matches the query.
|
||
The function simply looks for the query string in the keys of the JSON.
|
||
"""
|
||
data = json.loads(META_JSON.read_text())
|
||
for key, value in data.items():
|
||
if query.lower() in key.lower():
|
||
return json.dumps({key: value}, indent=2)
|
||
return "No metadata found for the given query."
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 4. DeepAgent с маршрутизацией
|
||
# ---------------------------------------------------------------------------
|
||
backend = CompositeBackend([
|
||
LocalShellBackend(workspace_dir="./workspace"),
|
||
FilesystemBackend(),
|
||
])
|
||
|
||
agent = create_deep_agent(
|
||
model=llm,
|
||
tools=[search_course_docs, fetch_course_meta],
|
||
backend=backend,
|
||
system_prompt=(
|
||
"You are a helpful FAQ bot for a course.\n"
|
||
"If the question is about course content, use the search_course_docs tool.\n"
|
||
"If the question is about schedule, instructor, or other metadata, use fetch_course_meta.\n"
|
||
"Do not call both tools unless absolutely necessary.\n"
|
||
"In your final answer, prepend 'source: chroma' or 'source: mcp_meta' to indicate where the answer came from."
|
||
),
|
||
)
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 5. CLI
|
||
# ---------------------------------------------------------------------------
|
||
PRESET_QUESTIONS = [
|
||
"What topics are covered in Lecture 1?", # should hit chroma
|
||
"Who is the instructor for this course?", # should hit mcp_meta
|
||
"Explain the concept of polymorphism."
|
||
]
|
||
|
||
async def run_agent(question: str) -> str:
|
||
result = await agent.ainvoke(
|
||
{"messages": [HumanMessage(content=question)]},
|
||
{"configurable": {"thread_id": "session-1"}},
|
||
)
|
||
return result["messages"][-1].content
|
||
|
||
async def main():
|
||
print("\n--- FAQ Bot Demo ---\n")
|
||
for i, q in enumerate(PRESET_QUESTIONS, 1):
|
||
print(f"Q{i}: {q}")
|
||
ans = await run_agent(q)
|
||
print(f"A{i}: {ans}\n")
|
||
|
||
print("Enter your own question (or press Ctrl+C to exit):")
|
||
while True:
|
||
try:
|
||
user_q = input("> ")
|
||
if not user_q.strip():
|
||
continue
|
||
ans = await run_agent(user_q)
|
||
print(ans)
|
||
except KeyboardInterrupt:
|
||
print("\nExiting.")
|
||
break
|
||
|
||
if __name__ == "__main__":
|
||
asyncio.run(main())
|