Spaces:
Sleeping
Sleeping
File size: 18,338 Bytes
1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 e9c4983 1773e23 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 |
import os
import sys
import logging
from pathlib import Path
import json
import hashlib
from datetime import datetime
import threading
import queue
from typing import List, Dict, Any, Tuple, Optional
# Configure logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
# Importing necessary libraries
import torch
import numpy as np
from sentence_transformers import SentenceTransformer
import chromadb
from chromadb.utils import embedding_functions
import gradio as gr
from openai import OpenAI
import google.generativeai as genai
# Configuration class
class Config:
"""Configuration for vector store and RAG"""
def __init__(self,
local_dir: str = "./chroma_data",
batch_size: int = 20,
max_workers: int = 4,
embedding_model: str = "all-MiniLM-L6-v2",
collection_name: str = "markdown_docs"):
self.local_dir = local_dir
self.batch_size = batch_size
self.max_workers = max_workers
self.checkpoint_file = Path(local_dir) / "checkpoint.json"
self.embedding_model = embedding_model
self.collection_name = collection_name
# Create local directory for checkpoints and Chroma
Path(local_dir).mkdir(parents=True, exist_ok=True)
# Embedding engine
class EmbeddingEngine:
"""Handle embeddings with a lightweight model"""
def __init__(self, model_name="all-MiniLM-L6-v2"):
# Use GPU if available
self.device = "cuda" if torch.cuda.is_available() else "cpu"
logger.info(f"Using device: {self.device}")
# Try multiple model options in order of preference
model_options = [
model_name,
"all-MiniLM-L6-v2",
"paraphrase-MiniLM-L3-v2",
"all-mpnet-base-v2" # Higher quality but larger model
]
self.model = None
# Try each model in order until one works
for model_option in model_options:
try:
logger.info(f"Attempting to load model: {model_option}")
self.model = SentenceTransformer(model_option)
# Move model to device
self.model.to(self.device)
logger.info(f"Successfully loaded model: {model_option}")
self.model_name = model_option
self.vector_size = self.model.get_sentence_embedding_dimension()
break
except Exception as e:
logger.warning(f"Failed to load model {model_option}: {str(e)}")
if self.model is None:
logger.error("Failed to load any embedding model. Exiting.")
sys.exit(1)
def encode(self, text, batch_size=32):
"""Get embedding for a text or list of texts"""
# Handle single text
if isinstance(text, str):
texts = [text]
else:
texts = text
# Truncate texts if necessary to avoid tokenization issues
truncated_texts = [t[:50000] if len(t) > 50000 else t for t in texts]
# Generate embeddings
try:
embeddings = self.model.encode(truncated_texts, batch_size=batch_size,
show_progress_bar=False, convert_to_numpy=True)
return embeddings
except Exception as e:
logger.error(f"Error generating embeddings: {e}")
# Return zero embeddings as fallback
return np.zeros((len(truncated_texts), self.vector_size))
class VectorStoreManager:
"""Manage Chroma vector store operations - upload, query, etc."""
def __init__(self, config: Config):
self.config = config
# Initialize Chroma client (local persistence)
logger.info(f"Initializing Chroma at {config.local_dir}")
self.client = chromadb.PersistentClient(path=config.local_dir)
# Get or create collection
try:
# Initialize embedding model
logger.info("Loading embedding model...")
self.embedding_engine = EmbeddingEngine(config.embedding_model)
logger.info(f"Using model: {self.embedding_engine.model_name}")
# Create embedding function
sentence_transformer_ef = embedding_functions.SentenceTransformerEmbeddingFunction(
model_name=self.embedding_engine.model_name
)
# Try to get existing collection
try:
self.collection = self.client.get_collection(
name=config.collection_name,
embedding_function=sentence_transformer_ef
)
logger.info(f"Using existing collection: {config.collection_name}")
except:
# Create new collection if it doesn't exist
self.collection = self.client.create_collection(
name=config.collection_name,
embedding_function=sentence_transformer_ef,
metadata={"hnsw:space": "cosine"}
)
logger.info(f"Created new collection: {config.collection_name}")
except Exception as e:
logger.error(f"Error initializing Chroma collection: {e}")
sys.exit(1)
def query(self, query_text: str, n_results: int = 5) -> List[Dict]:
"""
Query the vector store with a text query
"""
try:
# Query the collection
search_results = self.collection.query(
query_texts=[query_text],
n_results=n_results,
include=["documents", "metadatas", "distances"]
)
# Format results
results = []
if search_results["documents"] and len(search_results["documents"][0]) > 0:
for i in range(len(search_results["documents"][0])):
results.append({
'document': search_results["documents"][0][i],
'metadata': search_results["metadatas"][0][i],
'score': 1.0 - search_results["distances"][0][i] # Convert distance to similarity
})
return results
except Exception as e:
logger.error(f"Error querying collection: {e}")
return []
def get_statistics(self) -> Dict[str, Any]:
"""Get statistics about the vector store"""
stats = {}
try:
# Get collection count
collection_info = self.collection.count()
stats['total_documents'] = collection_info
# Estimate unique files - with no chunking, each document is a file
stats['unique_files'] = collection_info
except Exception as e:
logger.error(f"Error getting statistics: {e}")
stats['error'] = str(e)
return stats
class RAGSystem:
"""Retrieval-Augmented Generation with multiple LLM providers"""
def __init__(self, vector_store: VectorStoreManager):
self.vector_store = vector_store
self.openai_client = None
self.gemini_configured = False
def setup_openai(self, api_key: str):
"""Set up OpenAI client with API key"""
try:
self.openai_client = OpenAI(api_key=api_key)
return True
except Exception as e:
logger.error(f"Error initializing OpenAI client: {e}")
return False
def setup_gemini(self, api_key: str):
"""Set up Gemini with API key"""
try:
genai.configure(api_key=api_key)
self.gemini_configured = True
return True
except Exception as e:
logger.error(f"Error configuring Gemini: {e}")
return False
def format_context(self, documents: List[Dict]) -> str:
"""Format retrieved documents into context for the LLM"""
if not documents:
return "No relevant documents found."
context_parts = []
for i, doc in enumerate(documents):
metadata = doc['metadata']
title = metadata.get('title', metadata.get('filename', 'Unknown document'))
# For readability, limit length of context document
doc_text = doc['document']
if len(doc_text) > 10000: # Limit long documents in context
doc_text = doc_text[:10000] + "... [Document truncated for context]"
context_parts.append(f"Document {i+1} - {title}:\n{doc_text}\n")
return "\n".join(context_parts)
def generate_response_openai(self, query: str, context: str) -> str:
"""Generate a response using OpenAI model with context"""
if not self.openai_client:
return "Error: OpenAI API key not configured. Please enter an API key in the settings tab."
system_prompt = """
You are a helpful assistant that answers questions based on the context provided.
Use the information from the context to answer the user's question.
If the context doesn't contain the information needed, say so clearly.
Always cite the specific sections from the context that you used in your answer.
"""
try:
response = self.openai_client.chat.completions.create(
model="gpt-4o-mini", # Use GPT-4o mini
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": f"Context:\n{context}\n\nQuestion: {query}"}
],
temperature=0.3, # Lower temperature for more factual responses
max_tokens=1000,
)
return response.choices[0].message.content
except Exception as e:
logger.error(f"Error generating response with OpenAI: {e}")
return f"Error generating response with OpenAI: {str(e)}"
def generate_response_gemini(self, query: str, context: str) -> str:
"""Generate a response using Gemini with context"""
if not self.gemini_configured:
return "Error: Google AI API key not configured. Please enter an API key in the settings tab."
prompt = f"""
You are a helpful assistant that answers questions based on the context provided.
Use the information from the context to answer the user's question.
If the context doesn't contain the information needed, say so clearly.
Always cite the specific sections from the context that you used in your answer.
Context:
{context}
Question: {query}
"""
try:
model = genai.GenerativeModel('gemini-1.5-flash')
response = model.generate_content(prompt)
return response.text
except Exception as e:
logger.error(f"Error generating response with Gemini: {e}")
return f"Error generating response with Gemini: {str(e)}"
def query_and_generate(self, query: str, n_results: int = 5, model: str = "openai") -> str:
"""Retrieve relevant documents and generate a response using the specified model"""
# Query vector store
documents = self.vector_store.query(query, n_results=n_results)
if not documents:
return "No relevant documents found to answer your question."
# Format context
context = self.format_context(documents)
# Generate response with the appropriate model
if model == "openai":
return self.generate_response_openai(query, context)
elif model == "gemini":
return self.generate_response_gemini(query, context)
else:
return f"Unknown model: {model}"
def rag_chat(query, n_results, model_choice, rag_system):
"""Function to handle RAG chat queries"""
return rag_system.query_and_generate(query, n_results=int(n_results), model=model_choice)
def simple_query(query, n_results, vector_store):
"""Function to handle simple vector store queries"""
results = vector_store.query(query, n_results=int(n_results))
# Format results for display
formatted = []
for i, res in enumerate(results):
metadata = res['metadata']
title = metadata.get('title', metadata.get('filename', 'Unknown'))
# Limit preview text for display
preview = res['document'][:800] + '...' if len(res['document']) > 800 else res['document']
formatted.append(f"**Result {i+1}** (Similarity: {res['score']:.2f})\n\n"
f"**Source:** {title}\n\n"
f"**Content:**\n{preview}\n\n"
f"---\n")
return "\n".join(formatted) if formatted else "No results found."
def get_db_stats(vector_store):
"""Function to get vector store statistics"""
stats = vector_store.get_statistics()
return (f"Total documents: {stats.get('total_documents', 0)}\n"
f"Unique files: {stats.get('unique_files', 0)}")
def update_api_keys(openai_key, gemini_key, rag_system):
"""Update API keys for the RAG system"""
success_msg = []
if openai_key:
if rag_system.setup_openai(openai_key):
success_msg.append("β
OpenAI API key configured successfully")
else:
success_msg.append("β Failed to configure OpenAI API key")
if gemini_key:
if rag_system.setup_gemini(gemini_key):
success_msg.append("β
Google AI API key configured successfully")
else:
success_msg.append("β Failed to configure Google AI API key")
if not success_msg:
return "Please enter at least one API key"
return "\n".join(success_msg)
# Main function to run the application
def main():
# Set up paths for existing Chroma database
chroma_dir = Path("./chroma_data")
# Initialize the system
config = Config(
local_dir=str(chroma_dir),
collection_name="markdown_docs"
)
# Initialize vector store manager with existing collection
vector_store = VectorStoreManager(config)
# Initialize RAG system without API keys initially
rag_system = RAGSystem(vector_store)
# Define Gradio app
def rag_chat_wrapper(query, n_results, model_choice):
return rag_chat(query, n_results, model_choice, rag_system)
def simple_query_wrapper(query, n_results):
return simple_query(query, n_results, vector_store)
def update_api_keys_wrapper(openai_key, gemini_key):
return update_api_keys(openai_key, gemini_key, rag_system)
# Create the Gradio interface
with gr.Blocks(title="Markdown RAG System") as app:
gr.Markdown("# RAG System with Multiple LLM Providers")
with gr.Tab("Chat with Documents"):
with gr.Row():
with gr.Column(scale=3):
query_input = gr.Textbox(label="Question", placeholder="Ask a question about your documents...")
num_results = gr.Slider(minimum=1, maximum=10, value=3, step=1, label="Number of documents to retrieve")
model_choice = gr.Radio(
choices=["openai", "gemini"],
value="openai",
label="Choose LLM Provider",
info="Select which model to use for generating answers"
)
query_button = gr.Button("Ask", variant="primary")
with gr.Column(scale=7):
response_output = gr.Markdown(label="Response")
# Database stats
stats_display = gr.Textbox(label="Database Statistics", value=get_db_stats(vector_store))
refresh_button = gr.Button("Refresh Statistics")
with gr.Tab("Document Search"):
search_input = gr.Textbox(label="Search Query", placeholder="Search your documents...")
search_num = gr.Slider(minimum=1, maximum=20, value=5, step=1, label="Number of results")
search_button = gr.Button("Search", variant="primary")
search_output = gr.Markdown(label="Search Results")
with gr.Tab("Settings"):
gr.Markdown("""
## API Keys Configuration
This application can use either OpenAI's GPT-4o-mini or Google's Gemini 1.5 Flash for generating responses.
You need to provide at least one API key to use the chat functionality.
""")
openai_key_input = gr.Textbox(
label="OpenAI API Key",
placeholder="Enter your OpenAI API key here...",
type="password"
)
gemini_key_input = gr.Textbox(
label="Google AI API Key",
placeholder="Enter your Google AI API key here...",
type="password"
)
save_keys_button = gr.Button("Save API Keys", variant="primary")
api_status = gr.Markdown("")
# Set up events
query_button.click(
fn=rag_chat_wrapper,
inputs=[query_input, num_results, model_choice],
outputs=response_output
)
refresh_button.click(
fn=lambda: get_db_stats(vector_store),
inputs=None,
outputs=stats_display
)
search_button.click(
fn=simple_query_wrapper,
inputs=[search_input, search_num],
outputs=search_output
)
save_keys_button.click(
fn=update_api_keys_wrapper,
inputs=[openai_key_input, gemini_key_input],
outputs=api_status
)
# Launch the interface
app.launch()
if __name__ == "__main__":
main() |