sagar008 commited on
Commit
2b2ba3d
Β·
verified Β·
1 Parent(s): 62169f2

Update vector_store.py

Browse files
Files changed (1) hide show
  1. vector_store.py +26 -25
vector_store.py CHANGED
@@ -113,38 +113,39 @@ class LegalDocumentVectorStore:
113
  print(f"❌ Error saving pre-computed embeddings: {e}")
114
  return False
115
 
116
- def get_retriever(self, clause_tagger, document_id: str = None):
117
- """Get retriever for chat functionality with improved settings"""
118
- try:
119
- self._initialize_pinecone()
 
120
 
121
- legal_embeddings = InLegalBERTEmbeddings(clause_tagger.embedding_model)
122
- index = self.pc.Index(self.index_name)
123
 
124
- vectorstore = PineconeVectorStore(
125
- index=index,
126
- embedding=legal_embeddings,
127
- text_key="text"
128
- )
129
 
130
  # More permissive search settings
131
- search_kwargs = {
132
- 'k': 10, # Increased from default 5
133
- 'include_metadata': True
134
- }
135
 
136
- if document_id:
137
- search_kwargs['filter'] = {'document_id': document_id}
138
 
139
- # Use similarity search without threshold initially
140
- return vectorstore.as_retriever(
141
- search_type="similarity", # Remove threshold for now
142
- search_kwargs=search_kwargs
143
- )
144
 
145
- except Exception as e:
146
- print(f"❌ Error creating retriever: {e}")
147
- return None
148
 
149
 
150
  # Global instance
 
113
  print(f"❌ Error saving pre-computed embeddings: {e}")
114
  return False
115
 
116
+
117
+ def get_retriever(self, clause_tagger, document_id: str = None):
118
+ """Get retriever for chat functionality with improved settings"""
119
+ try:
120
+ self._initialize_pinecone()
121
 
122
+ legal_embeddings = InLegalBERTEmbeddings(clause_tagger.embedding_model)
123
+ index = self.pc.Index(self.index_name)
124
 
125
+ vectorstore = PineconeVectorStore(
126
+ index=index,
127
+ embedding=legal_embeddings,
128
+ text_key="text"
129
+ )
130
 
131
  # More permissive search settings
132
+ search_kwargs = {
133
+ 'k': 10, # Increased from default 5
134
+ 'include_metadata': True
135
+ }
136
 
137
+ if document_id:
138
+ search_kwargs['filter'] = {'document_id': document_id}
139
 
140
+ # Use similarity search without threshold initially
141
+ return vectorstore.as_retriever(
142
+ search_type="similarity", # Remove threshold for now
143
+ search_kwargs=search_kwargs
144
+ )
145
 
146
+ except Exception as e:
147
+ print(f"❌ Error creating retriever: {e}")
148
+ return None
149
 
150
 
151
  # Global instance