Update main.py with Ollama and LangGraph implementation
This commit is contained in:
@@ -1,124 +1,166 @@
|
||||
"""main.py
|
||||
|
||||
Простой FAQ‑бот, использующий ChromaDB для локальных конспектов и один MCP‑подобный инструмент – HTTP‑запрос к статическому JSON.
|
||||
|
||||
Требования:
|
||||
- Python 3.10+
|
||||
- Ollama модели: `nomic-embed-text` и `llama3`
|
||||
- Пакеты из requirements.txt
|
||||
|
||||
Запуск:
|
||||
```bash
|
||||
python main.py
|
||||
```
|
||||
|
||||
В интерактивном режиме можно задавать вопросы. В примере уже есть три готовых вопроса – два из Chroma, один из MCP‑тул.
|
||||
"""
|
||||
|
||||
import json
|
||||
import os
|
||||
import asyncio
|
||||
import sys
|
||||
from pathlib import Path
|
||||
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
|
||||
from typing import List
|
||||
|
||||
import httpx
|
||||
from langchain_ollama import ChatOllama, OllamaEmbeddings
|
||||
from langchain_chroma import Chroma
|
||||
from langchain_core.documents import Document
|
||||
from langchain.tools import tool
|
||||
from deepagents import create_deep_agent
|
||||
from deepagents.backends import FilesystemBackend, LocalShellBackend, CompositeBackend
|
||||
from langchain_core.messages import HumanMessage
|
||||
from langchain_core.tools import BaseTool, tool
|
||||
from langchain_core.prompts import ChatPromptTemplate
|
||||
from langchain_core.output_parsers import StrOutputParser
|
||||
from langchain_core.runnables import Runnable
|
||||
from langchain.agents import create_agent, AgentExecutor, Tool
|
||||
|
||||
# --------------------- Configuration ---------------------
|
||||
BASE_DIR = Path(__file__).parent
|
||||
DATA_DIR = BASE_DIR / "data"
|
||||
CHROMA_DIR = BASE_DIR / "chroma_faq"
|
||||
MOCK_META_FILE = BASE_DIR / "course_meta.json"
|
||||
# ---------------------------------------------------------------------------
|
||||
# 1. Векторная база
|
||||
# ---------------------------------------------------------------------------
|
||||
CHROMA_PATH = Path("./chroma_faq")
|
||||
DATA_PATH = Path("./data")
|
||||
|
||||
# --------------------- LLM and Embeddings ---------------------
|
||||
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,
|
||||
)
|
||||
# Создаём embeddings
|
||||
embeddings = OllamaEmbeddings(model="nomic-embed-text")
|
||||
|
||||
embeddings = OpenAIEmbeddings(
|
||||
model="text-embedding-3-small",
|
||||
base_url="https://openrouter.ai/api/v1",
|
||||
api_key=os.getenv("OPENAI_API_KEY"),
|
||||
)
|
||||
# ---------------------------------------------------------------------------
|
||||
# Чтение markdown‑файлов и загрузка в Chroma
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
# --------------------- Chroma Vector Store ---------------------
|
||||
vector_store = Chroma(
|
||||
collection_name="faq_collection",
|
||||
embedding_function=embeddings,
|
||||
persist_directory=str(CHROMA_DIR),
|
||||
)
|
||||
def load_faq_to_chroma() -> Chroma:
|
||||
"""Загружает все .md файлы из DATA_PATH в Chroma.
|
||||
Если коллекция уже существует – просто возвращаем её.
|
||||
"""
|
||||
if CHROMA_PATH.exists():
|
||||
return Chroma(
|
||||
collection_name="faq",
|
||||
embedding_function=embeddings,
|
||||
persist_directory=str(CHROMA_PATH),
|
||||
)
|
||||
# Если нет, создаём новую коллекцию
|
||||
from langchain_text_splitters import MarkdownTextSplitter
|
||||
|
||||
# --------------------- Tools ---------------------
|
||||
splitter = MarkdownTextSplitter(chunk_size=500, chunk_overlap=50)
|
||||
docs = []
|
||||
for md_file in DATA_PATH.glob("*.md"):
|
||||
text = md_file.read_text(encoding="utf-8")
|
||||
docs.extend(splitter.split_text(text))
|
||||
vector_store = Chroma.from_texts(
|
||||
texts=docs,
|
||||
embedding=embeddings,
|
||||
collection_name="faq",
|
||||
persist_directory=str(CHROMA_PATH),
|
||||
)
|
||||
return vector_store
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 2. Tool: поиск по Chroma
|
||||
# ---------------------------------------------------------------------------
|
||||
@tool
|
||||
def search_course_docs(query: str, k: int = 3) -> str:
|
||||
"""Search the local FAQ collection for relevant passages."""
|
||||
docs = vector_store.similarity_search(query, k=k)
|
||||
if not docs:
|
||||
return "No relevant information found in the course materials."
|
||||
return "\n\n---\n\n".join(f"**{doc.metadata.get('title', 'Document')}**\n{doc.page_content}" for doc in docs)
|
||||
"""Возвращает кортеж из k наиболее релевантных фрагментов.
|
||||
Формат ответа: ``source: chroma`` + текст.
|
||||
"""
|
||||
vector_store = load_faq_to_chroma()
|
||||
results = vector_store.similarity_search(query, k=k)
|
||||
snippets = "\n---\n".join([r.page_content for r in results])
|
||||
return f"source: chroma\n{snippets}"
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 3. MCP‑подобный инструмент – HTTP‑запрос к статическому JSON
|
||||
# ---------------------------------------------------------------------------
|
||||
MCP_JSON_PATH = Path("./course_meta.json")
|
||||
|
||||
@tool
|
||||
def fetch_course_meta(query: str) -> str:
|
||||
"""Mock MCP-style tool that returns course metadata.
|
||||
In production this would be an HTTP call to an MCP server.
|
||||
Here we read a local JSON file for simplicity.
|
||||
"""Имитирует вызов MCP‑сервера. Читает локальный JSON и возвращает
|
||||
информацию, содержащуюся в ключе, совпадающем с query.
|
||||
"""
|
||||
import json
|
||||
if not MOCK_META_FILE.exists():
|
||||
return "Metadata source not available."
|
||||
with open(MOCK_META_FILE, "r", encoding="utf-8") as f:
|
||||
data = json.load(f)
|
||||
# Simple keyword search in the metadata
|
||||
results = [f"{k}: {v}" for k, v in data.items() if query.lower() in k.lower() or query.lower() in str(v).lower()]
|
||||
return "\n".join(results) if results else "No metadata matches your query."
|
||||
if not MCP_JSON_PATH.exists():
|
||||
return "source: mcp_meta\nMeta file not found."
|
||||
data = json.loads(MCP_JSON_PATH.read_text(encoding="utf-8"))
|
||||
# простая логика: ищем ключ, содержащий query (case‑insensitive)
|
||||
for key, value in data.items():
|
||||
if query.lower() in key.lower():
|
||||
return f"source: mcp_meta\n{key}: {value}"
|
||||
return "source: mcp_meta\nNo matching metadata found."
|
||||
|
||||
# --------------------- Backend ---------------------
|
||||
backend = CompositeBackend([
|
||||
LocalShellBackend(workspace_dir="./workspace"),
|
||||
FilesystemBackend(),
|
||||
])
|
||||
|
||||
# --------------------- Agent ---------------------
|
||||
agent = create_deep_agent(
|
||||
model=llm,
|
||||
tools=[search_course_docs, fetch_course_meta],
|
||||
backend=backend,
|
||||
system_prompt=(
|
||||
"You are a helpful FAQ bot for the course.\n"
|
||||
"Use the search_course_docs tool for questions about lecture materials.\n"
|
||||
"Use the fetch_course_meta tool for questions about schedule or metadata.\n"
|
||||
"Do not call both tools unless absolutely necessary.\n"
|
||||
"In your final answer, prefix the source with 'source: chroma' or 'source: mcp_meta'."
|
||||
),
|
||||
# ---------------------------------------------------------------------------
|
||||
# 4. Создание агента
|
||||
# ---------------------------------------------------------------------------
|
||||
# Системный промпт, который заставляет агент выбирать нужный инструмент
|
||||
SYSTEM_PROMPT = (
|
||||
"You are a helpful assistant for a course FAQ. Use the provided tools to answer the user. "
|
||||
"If the question is about course materials, use search_course_docs. "
|
||||
"If it is about schedule or metadata, use fetch_course_meta. "
|
||||
"Do not use both tools unless necessary. "
|
||||
"Always prefix your answer with the source: chroma or source: mcp_meta."
|
||||
)
|
||||
|
||||
# --------------------- Data Loading ---------------------
|
||||
async def load_faq_to_chroma():
|
||||
"""Load all .md files from data/ into the Chroma collection."""
|
||||
if not DATA_DIR.exists():
|
||||
print("Data directory not found.")
|
||||
return
|
||||
docs = []
|
||||
for md_file in DATA_DIR.glob("*.md"):
|
||||
text = md_file.read_text(encoding="utf-8")
|
||||
docs.append(Document(page_content=text, metadata={"title": md_file.stem}))
|
||||
if docs:
|
||||
vector_store.add_documents(docs)
|
||||
vector_store.persist()
|
||||
print(f"Loaded {len(docs)} documents into Chroma.")
|
||||
else:
|
||||
print("No markdown files found in data/.")
|
||||
# Создаём LLM
|
||||
llm = ChatOllama(model="llama3", temperature=0.2)
|
||||
|
||||
# --------------------- CLI ---------------------
|
||||
PRESET_QUESTIONS = [
|
||||
"What is the deadline for the final project?", # chroma
|
||||
"Explain the concept of tokenization in NLP.", # chroma
|
||||
"When is the next lecture scheduled?", # mcp_meta
|
||||
# Prompt template
|
||||
prompt = ChatPromptTemplate.from_messages([
|
||||
("system", SYSTEM_PROMPT),
|
||||
("human", "{input}"),
|
||||
])
|
||||
|
||||
# Создаём Runnable, который будет использовать инструменты
|
||||
agent = create_agent(
|
||||
llm=llm,
|
||||
tools=[search_course_docs, fetch_course_meta],
|
||||
system_message=SYSTEM_PROMPT,
|
||||
verbose=True,
|
||||
)
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 5. CLI
|
||||
# ---------------------------------------------------------------------------
|
||||
PREDEFINED_QUESTIONS = [
|
||||
"What topics are covered in the first lecture?", # Chroma
|
||||
"Explain the concept of polymorphism in the context of the course.", # Chroma
|
||||
"What is the schedule for the next week?", # MCP
|
||||
]
|
||||
|
||||
async def run_cli():
|
||||
await load_faq_to_chroma()
|
||||
print("\n--- FAQ Bot CLI ---\n")
|
||||
for i, q in enumerate(PRESET_QUESTIONS, 1):
|
||||
def run_cli():
|
||||
print("--- FAQ Bot ---")
|
||||
print("Predefined questions:")
|
||||
for i, q in enumerate(PREDEFINED_QUESTIONS, 1):
|
||||
print(f"{i}. {q}")
|
||||
print("\nEnter your own question (or 'exit' to quit):")
|
||||
print("\nEnter 1-3 to ask a predefined question, or type your own question.")
|
||||
while True:
|
||||
user_input = input("> ")
|
||||
user_input = input("\n> ")
|
||||
if user_input.lower() in {"exit", "quit"}:
|
||||
print("Goodbye!")
|
||||
break
|
||||
result = await agent.ainvoke(
|
||||
{"messages": [HumanMessage(content=user_input)]},
|
||||
{"configurable": {"thread_id": "session-1"}},
|
||||
)
|
||||
print(result["messages"][-1].content)
|
||||
if user_input.isdigit() and 1 <= int(user_input) <= len(PREDEFINED_QUESTIONS):
|
||||
question = PREDEFINED_QUESTIONS[int(user_input) - 1]
|
||||
else:
|
||||
question = user_input
|
||||
try:
|
||||
response = agent.invoke({"input": question})
|
||||
print("\nAnswer:\n", response)
|
||||
except Exception as e:
|
||||
print("Error:", e)
|
||||
|
||||
if __name__ == "__main__":
|
||||
asyncio.run(run_cli())
|
||||
# Убедимся, что данные загружены
|
||||
load_faq_to_chroma()
|
||||
run_cli()
|
||||
|
||||
Reference in New Issue
Block a user