Files
cucumbers-solutions/solutions/6a02e23da6fe2e4ac16acf65/solution.py
T

129 lines
4.5 KiB
Python

from pathlib import Path
import os
from typing import List
# LLM and embeddings via Ollama
from langchain_ollama import ChatOllama, OllamaEmbeddings
# Tools
from langchain.tools import tool
# Vector store
from langchain_qdrant import QdrantVectorStore
from qdrant_client import QdrantClient
from qdrant_client.http.models import Distance, VectorParams
# Text splitter
from langchain.text_splitter import RecursiveCharacterTextSplitter
# Agent
from langchain.agents import create_agent
# Documents
from langchain_core.documents import Document
# -------------------- 1. RAG tools --------------------
@tool
def search_knowledge_base(query: str, max_results: int = 5) -> str:
"""Search the knowledge base for relevant documents."""
results = vector_store.similarity_search_with_score(query, k=max_results)
if not results:
return "No relevant documents found."
response_lines = []
for doc, score in results:
title = doc.metadata.get("title", "N/A")
snippet = doc.page_content[:200] + ("..." if len(doc.page_content) > 200 else "")
response_lines.append(f"Score: {score:.4f}\nTitle: {title}\nContent: {snippet}")
return "\n\n".join(response_lines)
@tool
def add_to_knowledge_base(content: str, title: str) -> str:
"""Add a new document to the knowledge base."""
doc = Document(page_content=content, metadata={"title": title})
vector_store.add_documents([doc])
return f"Document '{title}' added successfully."
# -------------------- 2. Qdrant setup --------------------
qdrant_client = QdrantClient(":memory:")
qdrant_client.create_collection(
collection_name="knowledge_base",
vectors_config=VectorParams(size=384, distance=Distance.COSINE),
)
embeddings = OllamaEmbeddings(model="nomic-embed-text")
vector_store = QdrantVectorStore(
client=qdrant_client,
collection_name="knowledge_base",
embedding=embeddings,
)
# -------------------- 3. Text splitter --------------------
splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=100)
def load_and_index(directory: str):
"""Load all .txt files from directory and index them."""
docs: List[Document] = []
for file_path in Path(directory).glob("*.txt"):
text = file_path.read_text(encoding="utf-8")
chunks = splitter.split_text(text)
for i, chunk in enumerate(chunks):
docs.append(
Document(
page_content=chunk,
metadata={
"title": f"{file_path.stem} #{i+1}",
"source": str(file_path),
},
)
)
vector_store.add_documents(docs)
# -------------------- 4. Agent --------------------
system_prompt = """
You are an assistant that can search and add documents to a knowledge base.
Use the tools `search_knowledge_base` and `add_to_knowledge_base` as needed.
"""
agent = create_agent(
model=ChatOllama(model="llama3", temperature=0.2),
tools=[search_knowledge_base, add_to_knowledge_base],
system_prompt=system_prompt,
)
# -------------------- 5. CLI client --------------------
def main():
# Load initial documents
load_and_index("docs") # ensure a 'docs' folder with .txt files
print(
"RAG Agent ready. Commands: /add <title> <content>, /search <query>, /quit"
)
while True:
user_input = input("> ").strip()
if not user_input:
continue
if user_input.lower() in ("exit", "quit", "/quit"):
break
if user_input.startswith("/add"):
try:
_, title, content = user_input.split(" ", 2)
result_msg = add_to_knowledge_base(content=content, title=title)
print(result_msg)
except ValueError:
print("Usage: /add <title> <content>")
elif user_input.startswith("/search"):
query = user_input[len("/search") :].strip()
if not query:
print("Provide a search query.")
continue
response = agent.invoke({"messages": [{"role": "human", "content": query}]})
for msg in response["messages"]:
if hasattr(msg, "content"):
print(msg.content)
else:
# Regular chat with the agent
response = agent.invoke(
{"messages": [{"role": "human", "content": user_input}]}
)
for msg in response["messages"]:
if hasattr(msg, "content"):
print(msg.content)
if __name__ == "__main__":
main()