File size: 20,732 Bytes
8629355 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 |
import os
import time
from typing import List, Dict, Any, Optional
# LangChain imports using modern patterns
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain_chroma import Chroma
from langchain_core.documents import Document
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough, RunnableParallel
from langchain_core.messages import HumanMessage, AIMessage
# Local imports
from config import Config, PromptTemplates, validate_config
from models import RouteQuery, QueryResult, RetrievalStats
from document_processor import DocumentProcessor
from chat_logger import ChatLogger
class RAGService:
"""Modern RAG service using LangChain Expression Language (LCEL)."""
def __init__(self, base_path: str = None):
"""Initialize the RAG service.
Args:
base_path: Base path for documents and vector store
"""
# Validate configuration
validate_config()
self.base_path = base_path or Config.DATA_BASE_PATH
self.chat_logger = ChatLogger()
self.conversation_memory = [] # Simple in-memory conversation storage
# Initialize components
self._initialize_models()
self._initialize_vector_store()
self._setup_chains()
# Track last generated queries for logging
self.last_generated_queries = []
def _initialize_models(self):
"""Initialize LLM and embedding models."""
print("Initializing AI models...")
# Initialize LLM
self.llm = ChatOpenAI(
model=Config.MODEL_NAME,
temperature=Config.TEMPERATURE,
api_key=Config.OPENAI_API_KEY
)
# Initialize embeddings with better error handling
self.embeddings = OpenAIEmbeddings(
api_key=Config.OPENAI_API_KEY,
model=Config.EMBEDDING_MODEL,
chunk_size=Config.EMBEDDING_CHUNK_SIZE,
max_retries=Config.EMBEDDING_MAX_RETRIES,
request_timeout=Config.EMBEDDING_REQUEST_TIMEOUT
)
print("β
AI models initialized successfully")
def _initialize_vector_store(self):
"""Initialize the vector store (empty initially)."""
self.vector_store = None
print("Vector store placeholder initialized")
def _setup_chains(self):
"""Set up all the LCEL chains."""
print("Setting up LangChain LCEL chains...")
# Router chain
router_prompt = ChatPromptTemplate.from_messages([
("system", PromptTemplates.ROUTER_SYSTEM_TEMPLATE),
("human", "{question}")
])
self.router_chain = router_prompt | self.llm.with_structured_output(RouteQuery)
# Query generation chains for different content types
self.query_generation_chains = {}
for content_type in ["course", "program", "both"]:
template = PromptTemplates.get_query_template(content_type)
prompt = ChatPromptTemplate.from_template(template)
self.query_generation_chains[content_type] = prompt | self.llm | StrOutputParser()
# Main QA chain
qa_prompt = ChatPromptTemplate.from_messages([
("system", PromptTemplates.SYSTEM_TEMPLATE),
("human", "{question}")
])
# This will be completed when vector store is loaded
self.qa_chain = None
print("β
LCEL chains set up successfully")
def load_documents(self) -> int:
"""Load and process documents, create or load vector store.
Returns:
Number of document chunks processed
"""
try:
print(f"Checking for existing database at: {Config.CHROMA_DB_PATH}")
if os.path.exists(Config.CHROMA_DB_PATH) and os.listdir(Config.CHROMA_DB_PATH):
print("Existing database found, attempting to load...")
count = self._load_existing_database()
if count == 0:
print("β οΈ Existing database is empty, rebuilding...")
return self._create_new_database()
return count
else:
print("No existing database found, creating new one...")
return self._create_new_database()
except Exception as e:
print(f"Error loading documents: {str(e)}")
raise
def _load_existing_database(self) -> int:
"""Load existing vector database.
Returns:
Number of documents in the database
"""
print("Loading existing embeddings from Chroma database...")
try:
self.vector_store = Chroma(
persist_directory=Config.CHROMA_DB_PATH,
embedding_function=self.embeddings,
collection_metadata={"hnsw:space": "cosine"},
collection_name=Config.COLLECTION_NAME
)
# Get collection size
collection_data = self.vector_store.get()
collection_size = len(collection_data['ids'])
if collection_size == 0:
print("Database exists but is empty")
return 0
print(f"β
Loaded {collection_size} existing document chunks from database")
self._setup_qa_chain()
return collection_size
except Exception as e:
print(f"Error loading existing database: {str(e)}")
return 0
def _create_new_database(self) -> int:
"""Create new vector database from documents.
Returns:
Number of document chunks processed
"""
print("Creating new embeddings (this will incur OpenAI API costs)...")
# Process documents
processor = DocumentProcessor(self.base_path)
documents = processor.process_all_documents()
if not documents:
raise ValueError("No documents found to process")
# Chunk documents
chunks = processor.chunk_documents(documents)
# Initialize empty vector store
self.vector_store = Chroma(
embedding_function=self.embeddings,
persist_directory=Config.CHROMA_DB_PATH,
collection_metadata={"hnsw:space": "cosine"},
collection_name=Config.COLLECTION_NAME
)
# Process documents in batches to avoid token limits
total_processed = self._process_documents_in_batches(chunks)
print(f"β
Database creation completed! Processed {total_processed} documents.")
self._setup_qa_chain()
return total_processed
def _process_documents_in_batches(self, chunks: List[Document]) -> int:
"""Process documents in batches to avoid API limits.
Args:
chunks: List of document chunks to process
Returns:
Number of successfully processed chunks
"""
batch_size = Config.BATCH_SIZE
total_processed = 0
print(f"Processing {len(chunks)} document chunks in batches of {batch_size}...")
for i in range(0, len(chunks), batch_size):
batch = chunks[i:i + batch_size]
batch_num = i // batch_size + 1
total_batches = (len(chunks) + batch_size - 1) // batch_size
print(f"Processing batch {batch_num}/{total_batches} ({len(batch)} documents)")
try:
self.vector_store.add_documents(batch)
total_processed += len(batch)
print(f"β
Successfully processed {len(batch)} documents (Total: {total_processed})")
# Small delay to be nice to the API
time.sleep(1)
except Exception as e:
print(f"β Error processing batch {batch_num}: {str(e)}")
# Continue with next batch instead of failing completely
continue
return total_processed
def _setup_qa_chain(self):
"""Set up the main QA chain with retriever."""
if not self.vector_store:
raise ValueError("Vector store not initialized")
# Create the main QA chain using LCEL
qa_prompt = ChatPromptTemplate.from_messages([
("system", PromptTemplates.SYSTEM_TEMPLATE),
("human", "{question}")
])
def format_docs(docs):
"""Format retrieved documents for the prompt."""
return "\n\n".join([d.page_content for d in docs])
def format_chat_history(memory):
"""Format chat history for the prompt."""
if not memory:
return "No previous conversation."
formatted = []
for msg in memory[-6:]: # Last 6 messages (3 exchanges)
if isinstance(msg, dict):
role = msg.get('role', 'unknown')
content = msg.get('content', '')
elif hasattr(msg, 'type') and hasattr(msg, 'content'):
role = msg.type
content = msg.content
else:
continue
formatted.append(f"{role}: {content}")
return "\n".join(formatted)
# Create retriever (will be configured per query)
self.base_retriever = self.vector_store.as_retriever()
# The QA chain will be constructed per query with specific retriever config
self.qa_prompt = qa_prompt
self.format_docs = format_docs
self.format_chat_history = format_chat_history
print("β
QA chain set up successfully")
def route_query(self, question: str) -> str:
"""Route the query to determine content type.
Args:
question: User's question
Returns:
Content type: 'course', 'program', or 'both'
"""
try:
result = self.router_chain.invoke({"question": question})
return result.content_type
except Exception as e:
print(f"Error in query routing: {str(e)}")
return "both" # Default to both if routing fails
def generate_query_variations(self, question: str, content_type: str) -> List[str]:
"""Generate multiple query variations for better retrieval.
Args:
question: Original question
content_type: Content type from routing
Returns:
List of query variations
"""
try:
chain = self.query_generation_chains[content_type]
variations = chain.invoke({"question": question})
# Process and clean the variations
queries = [q.strip() for q in variations.split('\n') if q.strip()]
# Always include the original question
if question not in queries:
queries.append(question)
# Store for logging
self.last_generated_queries = queries
return queries
except Exception as e:
print(f"Error generating query variations: {str(e)}")
# Fallback to original question
self.last_generated_queries = [question]
return [question]
def retrieve_documents(self, question: str, content_type: str) -> List[Document]:
"""Retrieve relevant documents using multiple query variations.
Args:
question: Original question
content_type: Content type from routing
Returns:
List of relevant documents
"""
if not self.vector_store:
raise ValueError("Vector store not initialized. Please load documents first.")
# Generate query variations
queries = self.generate_query_variations(question, content_type)
print(f"\nGenerated queries for '{question}':")
for q in queries:
print(f" β’ {q}")
# Configure retriever based on content type
k = Config.RETRIEVAL_K_VALUES[content_type]
# Create metadata filter if needed
search_kwargs = {
"k": k,
"fetch_k": k * 3 # Fetch more candidates for MMR
}
if content_type != "both":
search_kwargs["filter"] = {"doc_type": content_type}
# Configure retriever
retriever = self.vector_store.as_retriever(
search_type="mmr",
search_kwargs=search_kwargs
)
# Retrieve documents for each query variation
all_docs = []
for query in queries:
try:
docs = retriever.invoke(query)
all_docs.extend(docs)
except Exception as e:
print(f"Error retrieving for query '{query}': {str(e)}")
continue
# Remove duplicates while preserving order
unique_docs = []
seen_content = set()
for doc in all_docs:
# Create a unique identifier from content and source
doc_id = f"{doc.page_content[:100]}_{doc.metadata.get('source', '')}"
if doc_id not in seen_content:
seen_content.add(doc_id)
unique_docs.append(doc)
# Log retrieval statistics
doc_types = [doc.metadata.get('doc_type', 'unknown') for doc in unique_docs]
print(f"\nRetrieved {len(unique_docs)} unique documents:")
print(f" β’ Courses: {doc_types.count('course')}")
print(f" β’ Programs: {doc_types.count('program')}")
return unique_docs
def query(self, question: str) -> QueryResult:
"""Process a user query and return response.
Args:
question: User's question
Returns:
QueryResult with answer and metadata
"""
if not self.vector_store:
raise ValueError("Model not initialized. Please load documents first.")
start_time = time.time()
try:
# Route the query
content_type = self.route_query(question)
print(f"Query routed as: {content_type}")
# Retrieve relevant documents
docs = self.retrieve_documents(question, content_type)
# Format context and chat history
context = self.format_docs(docs)
chat_history = self.format_chat_history(self.conversation_memory)
# Generate answer using LCEL
chain = self.qa_prompt | self.llm | StrOutputParser()
answer = chain.invoke({
"context": context,
"question": question,
"chat_history": chat_history
})
# Update conversation memory
self.conversation_memory.extend([
{"role": "human", "content": question},
{"role": "assistant", "content": answer}
])
# Keep memory within reasonable size
if len(self.conversation_memory) > 12: # Keep last 6 exchanges
self.conversation_memory = self.conversation_memory[-12:]
# Format sources
sources = self._format_sources(docs)
if sources:
answer += sources
# Calculate processing time
processing_time = time.time() - start_time
# Create result
result = QueryResult(
answer=answer,
source_documents=docs,
content_type=content_type,
processing_time=processing_time,
generated_queries=self.last_generated_queries
)
# Log the interaction
self._log_interaction(question, result)
return result
except Exception as e:
error_msg = f"Error processing query: {str(e)}"
print(error_msg)
return QueryResult(
answer=error_msg,
source_documents=[],
content_type="error",
processing_time=time.time() - start_time
)
def _format_sources(self, docs: List[Document]) -> str:
"""Format source documents for display.
Args:
docs: Retrieved documents
Returns:
Formatted sources string
"""
if not docs:
return ""
# Get unique sources
sources = list(set(
os.path.basename(doc.metadata.get("source", ""))
for doc in docs if doc.metadata.get("source")
))
sources = sorted(sources)
if not sources:
return ""
sources_text = ""
if len(sources) > 2:
# Show only first 2 sources with expandable section for more
visible_sources = sources[:2]
hidden_sources = sources[2:]
sources_text += "\n\nSources:"
for source in visible_sources:
sources_text += f"\nβ’ {source}"
sources_text += f"\n<details><summary>**See {len(hidden_sources)} more sources...**</summary>\n"
for source in hidden_sources:
sources_text += f"\nβ’ {source}"
sources_text += "\n</details>"
else:
# If 2 or fewer sources, show all
sources_text += "\n\nSources:"
for source in sources:
sources_text += f"\nβ’ {source}"
return sources_text
def _log_interaction(self, question: str, result: QueryResult):
"""Log the interaction for analysis.
Args:
question: User's question
result: Query result
"""
try:
system_info = {
"model_version": Config.MODEL_NAME,
"embedding_version": Config.EMBEDDING_MODEL,
"search_config": {
"search_type": "mmr",
"k_value": Config.RETRIEVAL_K_VALUES.get(result.content_type),
"content_type": result.content_type
}
}
self.chat_logger.log_interaction(
question=question,
answer=result.answer,
source_documents=result.source_documents,
content_type=result.content_type,
generated_queries=result.generated_queries or [],
processing_time=result.processing_time or 0,
chat_history=self.conversation_memory,
system_info=system_info
)
except Exception as e:
print(f"Error logging interaction: {str(e)}")
def get_system_status(self) -> Dict[str, Any]:
"""Get current system status.
Returns:
Dictionary with system status information
"""
status = {
"database_initialized": self.vector_store is not None,
"model_version": Config.MODEL_NAME,
"embedding_version": Config.EMBEDDING_MODEL,
"conversation_length": len(self.conversation_memory),
"last_queries": self.last_generated_queries
}
if self.vector_store:
try:
collection_data = self.vector_store.get()
status["documents_loaded"] = len(collection_data['ids'])
except:
status["documents_loaded"] = "unknown"
else:
status["documents_loaded"] = 0
return status
def clear_conversation_memory(self):
"""Clear the conversation memory."""
self.conversation_memory = []
print("Conversation memory cleared")
def get_conversation_history(self) -> List[Dict[str, str]]:
"""Get the current conversation history.
Returns:
List of conversation messages
"""
return self.conversation_memory.copy() |