152 lines
5.0 KiB
Python
152 lines
5.0 KiB
Python
import os
|
|
import sys
|
|
import asyncio
|
|
from pathlib import Path
|
|
from typing import List
|
|
|
|
from langchain_ollama import Ollama, OllamaEmbeddings
|
|
from langchain_core.documents import Document
|
|
from langchain_qdrant import QdrantVectorStore
|
|
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
|
from langchain.tools import tool
|
|
from langchain_core.messages import HumanMessage
|
|
from deepagents import create_deep_agent
|
|
from deepagents.backends import FilesystemBackend, LocalShellBackend, CompositeBackend
|
|
from qdrant_client import QdrantClient
|
|
|
|
# Инициализация эмбеддингов и LLM через Ollama
|
|
embeddings = OllamaEmbeddings(model="nomic-embed-text")
|
|
llm = Ollama(model="llama3")
|
|
|
|
# Инициализация Qdrant
|
|
client = QdrantClient(host="localhost", port=6333)
|
|
collection_name = "knowledge"
|
|
vector_store = QdrantVectorStore(
|
|
client=client,
|
|
collection_name=collection_name,
|
|
embedding=embeddings,
|
|
)
|
|
|
|
# Чанкинг
|
|
splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
|
|
|
|
# Инструмент: поиск в базе знаний
|
|
@tool
|
|
def search_knowledge_base(query: str, max_results: int = 3) -> str:
|
|
"""Search the knowledge base for relevant information."""
|
|
docs: List[Document] = vector_store.similarity_search(query, k=max_results)
|
|
if not docs:
|
|
return "No results."
|
|
return "\n".join(doc.page_content for doc in docs)
|
|
|
|
# Инструмент: добавление документа в базу знаний
|
|
@tool
|
|
def add_to_knowledge_base(content: str, title: str = "doc") -> str:
|
|
"""Add content to the knowledge base."""
|
|
chunks = splitter.split_text(content)
|
|
docs = [Document(page_content=chunk, metadata={"title": title}) for chunk in chunks]
|
|
vector_store.add_documents(docs)
|
|
return f"Added: {title} ({len(chunks)} chunks)."
|
|
|
|
# Backend для deepagents
|
|
backend = CompositeBackend(
|
|
[
|
|
LocalShellBackend(workspace_dir="./workspace"),
|
|
FilesystemBackend(),
|
|
]
|
|
)
|
|
|
|
system_prompt = (
|
|
"You are a helpful agent with access to a knowledge base. "
|
|
"Use the tools to search and add information. "
|
|
"When you need to retrieve information, call search_knowledge_base. "
|
|
"When you need to store new information, call add_to_knowledge_base."
|
|
)
|
|
|
|
agent = create_deep_agent(
|
|
model=llm,
|
|
tools=[search_knowledge_base, add_to_knowledge_base],
|
|
backend=backend,
|
|
system_prompt=system_prompt,
|
|
)
|
|
|
|
# Загрузка документов из директории
|
|
def load_documents(dir_path: str) -> None:
|
|
"""Load all .txt files from dir_path into the knowledge base."""
|
|
path = Path(dir_path)
|
|
if not path.is_dir():
|
|
print(f"Directory not found: {dir_path}")
|
|
return
|
|
for file_path in path.glob("*.txt"):
|
|
try:
|
|
content = file_path.read_text(encoding="utf-8")
|
|
title = file_path.stem
|
|
result = add_to_knowledge_base(content, title)
|
|
print(f"Loaded {file_path.name}: {result}")
|
|
except Exception as e:
|
|
print(f"Error loading {file_path.name}: {e}")
|
|
|
|
# Интерактивный клиент
|
|
async def run_cli() -> None:
|
|
thread_id = "session-1"
|
|
print("Welcome to the RAG agent CLI.")
|
|
print("Commands:")
|
|
print(" /add - add a new document")
|
|
print(" /search - search the knowledge base")
|
|
print(" /quit - exit")
|
|
while True:
|
|
try:
|
|
user_input = input("\n> ").strip()
|
|
except (EOFError, KeyboardInterrupt):
|
|
print("\nExiting.")
|
|
break
|
|
|
|
if not user_input:
|
|
continue
|
|
|
|
if user_input.lower() == "/quit":
|
|
print("Goodbye.")
|
|
break
|
|
|
|
if user_input.lower() == "/add":
|
|
title = input("Title: ").strip()
|
|
print("Enter content (end with a single line containing only 'END'):")
|
|
lines: List[str] = []
|
|
while True:
|
|
line = input()
|
|
if line.strip() == "END":
|
|
break
|
|
lines.append(line)
|
|
content = "\n".join(lines)
|
|
result = add_to_knowledge_base(content, title)
|
|
print(result)
|
|
continue
|
|
|
|
if user_input.lower() == "/search":
|
|
query = input("Enter search query: ").strip()
|
|
if not query:
|
|
print("Empty query.")
|
|
continue
|
|
result = search_knowledge_base(query)
|
|
print("\nSearch results:")
|
|
print(result)
|
|
continue
|
|
|
|
# Any other message is sent to the agent
|
|
response = await agent.ainvoke(
|
|
{"messages": [HumanMessage(content=user_input)]},
|
|
{"configurable": {"thread_id": thread_id}},
|
|
)
|
|
agent_reply = response["messages"][-1].content
|
|
print(f"\nAgent: {agent_reply}")
|
|
|
|
def main() -> None:
|
|
# Optional: load documents from a directory passed as first argument
|
|
if len(sys.argv) > 1:
|
|
dir_path = sys.argv[1]
|
|
print(f"Loading documents from {dir_path}...")
|
|
load_documents(dir_path)
|
|
asyncio.run(run_cli())
|
|
|
|
if __name__ == "__main__":
|
|
main() |