Spaces:
Sleeping
Sleeping
File size: 8,805 Bytes
9679fcd |
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 |
"""
Complete RAG Engine
Integrates hybrid retrieval, GraphRAG, and Groq LLM for Ireland Q&A
"""
import json
import time
from typing import List, Dict, Optional
from hybrid_retriever import HybridRetriever, RetrievalResult
from groq_llm import GroqLLM
import hashlib
class IrelandRAGEngine:
"""Complete RAG engine for Ireland knowledge base"""
def __init__(
self,
chunks_file: str = "dataset/wikipedia_ireland/chunks.json",
graphrag_index_file: str = "dataset/wikipedia_ireland/graphrag_index.json",
groq_api_key: Optional[str] = None,
groq_model: str = "llama-3.3-70b-versatile",
use_cache: bool = True
):
"""Initialize RAG engine"""
print("[INFO] Initializing Ireland RAG Engine...")
# Initialize retriever
self.retriever = HybridRetriever(
chunks_file=chunks_file,
graphrag_index_file=graphrag_index_file
)
# Try to load pre-built indexes, otherwise build them
try:
self.retriever.load_indexes()
except:
print("[INFO] Pre-built indexes not found, building new ones...")
self.retriever.build_semantic_index()
self.retriever.build_keyword_index()
self.retriever.save_indexes()
# Initialize LLM
self.llm = GroqLLM(api_key=groq_api_key, model=groq_model)
# Cache for instant responses
self.use_cache = use_cache
self.cache = {}
self.cache_hits = 0
self.cache_misses = 0
print("[SUCCESS] RAG Engine ready!")
def _hash_query(self, query: str) -> str:
"""Create hash of query for caching"""
return hashlib.md5(query.lower().strip().encode()).hexdigest()
def answer_question(
self,
question: str,
top_k: int = 5,
semantic_weight: float = 0.7,
keyword_weight: float = 0.3,
use_community_context: bool = True,
return_debug_info: bool = False
) -> Dict:
"""
Answer a question about Ireland using GraphRAG
Args:
question: User's question
top_k: Number of chunks to retrieve
semantic_weight: Weight for semantic search (0-1)
keyword_weight: Weight for keyword search (0-1)
use_community_context: Whether to include community summaries
return_debug_info: Whether to return detailed debug information
Returns:
Dict with answer, citations, and metadata
"""
start_time = time.time()
# Check cache
query_hash = self._hash_query(question)
if self.use_cache and query_hash in self.cache:
self.cache_hits += 1
cached_result = self.cache[query_hash].copy()
cached_result['cached'] = True
cached_result['response_time'] = time.time() - start_time
return cached_result
self.cache_misses += 1
# Step 1: Hybrid retrieval
retrieval_start = time.time()
retrieved_chunks = self.retriever.hybrid_search(
query=question,
top_k=top_k,
semantic_weight=semantic_weight,
keyword_weight=keyword_weight
)
retrieval_time = time.time() - retrieval_start
# Step 2: Prepare contexts for LLM
contexts = []
for result in retrieved_chunks:
context = {
'text': result.text,
'source_title': result.source_title,
'source_url': result.source_url,
'combined_score': result.combined_score,
'semantic_score': result.semantic_score,
'keyword_score': result.keyword_score,
'community_id': result.community_id
}
contexts.append(context)
# Step 3: Add community context if enabled
community_summaries = []
if use_community_context:
# Get unique communities from results
communities = set(result.community_id for result in retrieved_chunks if result.community_id >= 0)
for comm_id in list(communities)[:2]: # Use top 2 communities
comm_context = self.retriever.get_community_context(comm_id)
if comm_context:
community_summaries.append({
'community_id': comm_id,
'num_chunks': comm_context.get('num_chunks', 0),
'top_entities': [e['entity'] for e in comm_context.get('top_entities', [])[:5]],
'sources': comm_context.get('sources', [])[:3]
})
# Step 4: Generate answer with citations
generation_start = time.time()
llm_result = self.llm.generate_with_citations(
question=question,
contexts=contexts,
max_contexts=top_k
)
generation_time = time.time() - generation_start
# Step 5: Build response
response = {
'question': question,
'answer': llm_result['answer'],
'citations': llm_result['citations'],
'num_contexts_used': llm_result['num_contexts_used'],
'communities': community_summaries if use_community_context else [],
'cached': False,
'response_time': time.time() - start_time,
'retrieval_time': retrieval_time,
'generation_time': generation_time
}
# Add debug info if requested
if return_debug_info:
response['debug'] = {
'retrieved_chunks': [
{
'rank': r.rank,
'source': r.source_title,
'semantic_score': f"{r.semantic_score:.3f}",
'keyword_score': f"{r.keyword_score:.3f}",
'combined_score': f"{r.combined_score:.3f}",
'community': r.community_id,
'text_preview': r.text[:150] + "..."
}
for r in retrieved_chunks
],
'cache_stats': {
'hits': self.cache_hits,
'misses': self.cache_misses,
'hit_rate': f"{self.cache_hits / (self.cache_hits + self.cache_misses) * 100:.1f}%" if (self.cache_hits + self.cache_misses) > 0 else "0%"
}
}
# Cache the response
if self.use_cache:
self.cache[query_hash] = response.copy()
return response
def get_cache_stats(self) -> Dict:
"""Get cache statistics"""
total_queries = self.cache_hits + self.cache_misses
hit_rate = (self.cache_hits / total_queries * 100) if total_queries > 0 else 0
return {
'cache_size': len(self.cache),
'cache_hits': self.cache_hits,
'cache_misses': self.cache_misses,
'total_queries': total_queries,
'hit_rate': f"{hit_rate:.1f}%"
}
def clear_cache(self):
"""Clear the response cache"""
self.cache.clear()
self.cache_hits = 0
self.cache_misses = 0
print("[INFO] Cache cleared")
def get_stats(self) -> Dict:
"""Get engine statistics"""
return {
'total_chunks': len(self.retriever.chunks),
'total_communities': len(self.retriever.graphrag_index['communities']),
'cache_stats': self.get_cache_stats()
}
if __name__ == "__main__":
# Test RAG engine
engine = IrelandRAGEngine()
# Test questions
questions = [
"What is the capital of Ireland?",
"When did Ireland join the European Union?",
"Who is the current president of Ireland?",
"What is the oldest university in Ireland?"
]
for question in questions:
print("\n" + "=" * 80)
print(f"Question: {question}")
print("=" * 80)
result = engine.answer_question(question, top_k=5, return_debug_info=True)
print(f"\nAnswer:\n{result['answer']}")
print(f"\nResponse Time: {result['response_time']:.2f}s")
print(f" - Retrieval: {result['retrieval_time']:.2f}s")
print(f" - Generation: {result['generation_time']:.2f}s")
print(f"\nCitations:")
for cite in result['citations']:
print(f" [{cite['id']}] {cite['source']} (score: {cite['relevance_score']:.3f})")
if result.get('communities'):
print(f"\nRelated Topics:")
for comm in result['communities']:
print(f" - {', '.join(comm['top_entities'][:3])}")
print("\n" + "=" * 80)
print("Cache Stats:", engine.get_cache_stats())
print("=" * 80)
|