rag-chat / main.py
pvanand's picture
Update main.py
0f54295 verified
from fastapi import FastAPI, HTTPException, Header, Depends, BackgroundTasks, Query
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import StreamingResponse
from pydantic import BaseModel, Field
from typing import List, Optional, Dict, AsyncGenerator
import json
import os
import logging
from txtai.embeddings import Embeddings
import pandas as pd
import glob
import uuid
import httpx
import asyncio
import requests
# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
app = FastAPI(
title="Embeddings API",
description="An API for creating and querying text embeddings indexes.",
version="1.0.0"
)
CHAT_AUTH_KEY = os.environ.get("CHAT_AUTH_KEY", "default_secret_key")
# Enable CORS
app.add_middleware(
CORSMiddleware,
allow_origins=["*"], # Allows all origins
allow_credentials=True,
allow_methods=["*"], # Allows all methods
allow_headers=["*"], # Allows all headers
)
embeddings = Embeddings({"path": "avsolatorio/GIST-all-MiniLM-L6-v2"})
class DocumentRequest(BaseModel):
index_id: str = Field(..., description="Unique identifier for the index")
documents: List[str] = Field(..., description="List of documents to be indexed")
class QueryRequest(BaseModel):
index_id: str = Field(..., description="Unique identifier for the index to query")
query: str = Field(..., description="The search query")
num_results: int = Field(..., description="Number of results to return", ge=1)
def save_embeddings(index_id: str, document_list: List[str]):
try:
folder_path = f"/app/indexes/{index_id}"
os.makedirs(folder_path, exist_ok=True)
# Save embeddings
embeddings.save(f"{folder_path}/embeddings")
# Save document_list
with open(f"{folder_path}/document_list.json", "w") as f:
json.dump(document_list, f)
logger.info(f"Embeddings and document list saved for index_id: {index_id}")
except Exception as e:
logger.error(f"Error saving embeddings for index_id {index_id}: {str(e)}")
raise HTTPException(status_code=500, detail=f"Error saving embeddings: {str(e)}")
def load_embeddings(index_id: str) -> List[str]:
try:
folder_path = f"/app/indexes/{index_id}"
if not os.path.exists(folder_path):
logger.error(f"Index not found for index_id: {index_id}")
raise HTTPException(status_code=404, detail="Index not found")
# Load embeddings
embeddings.load(f"{folder_path}/embeddings")
# Load document_list
with open(f"{folder_path}/document_list.json", "r") as f:
document_list = json.load(f)
logger.info(f"Embeddings and document list loaded for index_id: {index_id}")
return document_list
except Exception as e:
logger.error(f"Error loading embeddings for index_id {index_id}: {str(e)}")
raise HTTPException(status_code=500, detail=f"Error loading embeddings: {str(e)}")
@app.post("/create_index/", response_model=dict, tags=["Index Operations"])
async def create_index(request: DocumentRequest):
"""
Create a new index with the given documents.
- **index_id**: Unique identifier for the index
- **documents**: List of documents to be indexed
"""
try:
document_list = [(i, text, None) for i, text in enumerate(request.documents)]
embeddings.index(document_list)
save_embeddings(request.index_id, request.documents) # Save the original documents
logger.info(f"Index created successfully for index_id: {request.index_id}")
return {"message": "Index created successfully"}
except Exception as e:
logger.error(f"Error creating index: {str(e)}")
raise HTTPException(status_code=500, detail=f"Error creating index: {str(e)}")
@app.post("/query_index/", response_model=dict, tags=["Index Operations"])
async def query_index(request: QueryRequest):
"""
Query an existing index with the given search query.
- **index_id**: Unique identifier for the index to query
- **query**: The search query
- **num_results**: Number of results to return
"""
try:
document_list = load_embeddings(request.index_id)
results = embeddings.search(request.query, request.num_results)
queried_texts = [document_list[idx[0]] for idx in results]
logger.info(f"Query executed successfully for index_id: {request.index_id}")
return {"queried_texts": queried_texts}
except Exception as e:
logger.error(f"Error querying index: {str(e)}")
raise HTTPException(status_code=500, detail=f"Error querying index: {str(e)}")
def process_csv_file(file_path):
try:
df = pd.read_csv(file_path)
df_rows = df.apply(lambda row: ' '.join(row.values.astype(str)), axis=1)
txtai_data = [(i, row, None) for i, row in enumerate(df_rows)]
return txtai_data, df_rows.tolist()
except Exception as e:
logger.error(f"Error processing CSV file {file_path}: {str(e)}")
return None, None
def check_and_index_csv_files():
index_data_folder = "/app/index_data"
if not os.path.exists(index_data_folder):
logger.warning(f"index_data folder not found: {index_data_folder}")
return
csv_files = glob.glob(os.path.join(index_data_folder, "*.csv"))
for csv_file in csv_files:
index_id = os.path.splitext(os.path.basename(csv_file))[0]
if not os.path.exists(f"/app/indexes/{index_id}"):
logger.info(f"Processing CSV file: {csv_file}")
txtai_data, documents = process_csv_file(csv_file)
if txtai_data and documents:
embeddings.index(txtai_data)
save_embeddings(index_id, documents)
logger.info(f"CSV file indexed successfully: {csv_file}")
else:
logger.warning(f"Failed to process CSV file: {csv_file}")
else:
logger.info(f"Index already exists for: {csv_file}")
# ... [Previous code for DocumentRequest, QueryRequest, save_embeddings, load_embeddings, create_index, query_index, process_csv_file, check_and_index_csv_files remains the same]
class ChatRequest(BaseModel):
query: str = Field(..., description="The user's query")
index_id: str = Field(..., description="Unique identifier for the index to query")
conversation_id: Optional[str] = Field(None, description="Unique identifier for the conversation")
model_id: str = Field(..., description="Identifier for the LLM model to use")
user_id: str = Field(..., description="Unique identifier for the user")
enable_followup: bool = Field(default=False, description="Flag to enable follow-up questions")
def get_api_key(x_api_key: str = Header(...)) -> str:
if x_api_key != CHAT_AUTH_KEY:
raise HTTPException(status_code=403, detail="Invalid API key")
return x_api_key
def stream_llm_request(api_key: str, llm_request: dict, endpoint_url: str):
"""
Make a streaming request to the LLM service using requests.
"""
try:
headers = {
"accept": "text/event-stream",
"X-API-Key": api_key,
"Content-Type": "application/json"
}
with requests.post(endpoint_url, headers=headers, json=llm_request, stream=True) as response:
if response.status_code != 200:
raise HTTPException(status_code=response.status_code, detail="Error from LLM service")
for chunk in response.iter_content(chunk_size=None, decode_unicode=True):
if chunk:
yield chunk
except requests.RequestException as e:
logger.error(f"HTTP error occurred while making LLM request: {str(e)}")
raise HTTPException(status_code=500, detail=f"HTTP error occurred while making LLM request: {str(e)}")
except Exception as e:
logger.error(f"Unexpected error occurred while making LLM request: {str(e)}")
raise HTTPException(status_code=500, detail=f"Unexpected error occurred while making LLM request: {str(e)}")
@app.post("/chat/", response_class=StreamingResponse, tags=["Chat"])
async def chat(request: ChatRequest, background_tasks: BackgroundTasks, api_key: str = Depends(get_api_key)):
try:
document_list = load_embeddings(request.index_id)
search_results = embeddings.search(request.query, 6)
context = "\n".join([document_list[idx[0]] for idx in search_results])
rag_prompt = f"Based on the following context, please answer the user's question:\n\nContext:\n{context}\n\nUser's question: {request.query}\n\nAnswer:"
system_prompt = "You are a helpful assistant tasked with providing answers using the context provided"
conversation_id = request.conversation_id or str(uuid.uuid4())
if request.enable_followup:
llm_request = {
"query": rag_prompt,
"model_id": 'openai/gpt-4o-mini',
"conversation_id": conversation_id,
"user_id": request.user_id
}
endpoint_url = "https://pvanand-general-chat.hf.space/v2/followup-agent"
else:
llm_request = {
"prompt": rag_prompt,
"system_message": system_prompt,
"model_id": request.model_id,
"conversation_id": conversation_id,
"user_id": request.user_id
}
endpoint_url = "https://pvanand-audio-chat.hf.space/llm-agent"
logger.info(f"Starting chat response generation for user: {request.user_id} Full request: {llm_request}")
def response_generator():
full_response = ""
for chunk in stream_llm_request(api_key, llm_request, endpoint_url):
full_response += chunk
yield chunk
logger.info(f"Finished chat response generation for user: {request.user_id} Full response: {full_response}")
return StreamingResponse(response_generator(), media_type="text/event-stream")
except Exception as e:
logger.error(f"Error in chat endpoint: {str(e)}")
raise HTTPException(status_code=500, detail=f"Error in chat endpoint: {str(e)}")
@app.on_event("startup")
async def startup_event():
check_and_index_csv_files()
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=7860)