deploy vector store example

Author: James Liounis

perplexity-llamaindex/memory/chat_with_persistence/scripts/chat_store/docstore.json

@@ -0,0 +1 @@
+{}

perplexity-llamaindex/memory/chat_with_persistence/scripts/chat_store/graph_store.json

@@ -0,0 +1 @@
+{"graph_dict": {}}

perplexity-llamaindex/memory/chat_with_persistence/scripts/chat_store/image__vector_store.json

@@ -0,0 +1 @@
+{"embedding_dict": {}, "text_id_to_ref_doc_id": {}, "metadata_dict": {}}

perplexity-llamaindex/memory/chat_with_persistence/scripts/chat_store/index_store.json

@@ -0,0 +1 @@
+{"index_store/data": {"b20b1210-c462-4280-9ca8-690293aa7e07": {"__type__": "vector_store", "__data__": "{\"index_id\": \"b20b1210-c462-4280-9ca8-690293aa7e07\", \"summary\": null, \"nodes_dict\": {}, \"doc_id_dict\": {}, \"embeddings_dict\": {}}"}}}

perplexity-llamaindex/memory/chat_with_persistence/scripts/chat_with_persistence.py

@@ -0,0 +1,105 @@
+from llama_index.core import VectorStoreIndex, StorageContext, Document
+from llama_index.core.node_parser import SentenceSplitter
+from llama_index.vector_stores.lancedb import LanceDBVectorStore
+from openai import OpenAI as PerplexityClient
+from llama_index.core.vector_stores import MetadataFilters, MetadataFilter, FilterOperator
+import lancedb
+import pyarrow as pa
+import os
+from datetime import datetime
+
+# Initialize Perplexity Sonar client
+sonar_client = PerplexityClient(
+    api_key=os.environ["PERPLEXITY_API_KEY"],
+    base_url="https://api.perplexity.ai"
+)
+
+# Define explicit schema matching metadata structure
+schema = pa.schema([
+    pa.field("id", pa.string()),
+    pa.field("text", pa.string()),
+    pa.field("metadata", pa.map_(pa.string(), pa.string())),  # Store metadata as key-value map
+    pa.field("embedding", pa.list_(pa.float32(), 768))  # Match your embedding dimension
+])
+
+# Initialize persistent vector store with clean slate
+lancedb_uri = "./lancedb"
+if os.path.exists(lancedb_uri):
+    import shutil
+    shutil.rmtree(lancedb_uri)
+    
+db = lancedb.connect(lancedb_uri)
+vector_store = LanceDBVectorStore(uri=lancedb_uri, table_name="chat_history")
+storage_context = StorageContext.from_defaults(vector_store=vector_store)
+
+# Configure node parser with metadata support
+node_parser = SentenceSplitter(
+    chunk_size=1024,
+    chunk_overlap=100,
+    include_metadata=True
+)
+
+def initialize_chat_session():
+    """Create new session with proper schema"""
+    return VectorStoreIndex(
+        [],
+        storage_context=storage_context,
+        node_parser=node_parser
+    )
+
+def chat_with_persistence(user_query: str, index: VectorStoreIndex):
+    # Store user query
+    user_doc = Document(
+        text=user_query,
+        metadata={
+            "role": "user",
+            "timestamp": datetime.now().isoformat()
+        }
+    )
+    index.insert_nodes(node_parser.get_nodes_from_documents([user_doc]))
+    
+    # Retrieve context nodes
+    retriever = index.as_retriever(similarity_top_k=3)
+    context_nodes = retriever.retrieve(user_query)
+    
+    # Ensure context relevance by filtering for recent queries
+    context_text = "\n".join([
+        f"{n.metadata['role'].title()}: {n.text}" 
+        for n in context_nodes if n.metadata["role"] == "user"
+    ])
+    
+    # Generate Sonar API request
+    messages = [
+        {
+            "role": "system",
+            "content": f"Conversation History:\n{context_text}\n\nAnswer the latest query using this context."
+        },
+        {"role": "user", "content": user_query}
+    ]
+    
+    response = sonar_client.chat.completions.create(
+        model="sonar-pro",
+        messages=messages,
+        temperature=0.3
+    )
+    
+    assistant_response = response.choices[0].message.content
+    
+    # Store assistant response
+    assistant_doc = Document(
+        text=assistant_response,
+        metadata={
+            "role": "assistant",
+            "timestamp": datetime.now().isoformat()
+        }
+    )
+    index.insert_nodes(node_parser.get_nodes_from_documents([assistant_doc]))
+    
+    # Persist conversation state
+    storage_context.persist(persist_dir="./chat_store")
+    return assistant_response
+
+# Usage
+index = initialize_chat_session()
+print("Response:", chat_with_persistence("What's the current weather in London?", index))
+print("Follow-up:", chat_with_persistence("What about tomorrow's forecast?", index))

perplexity-llamaindex/memory/chat_with_persistence/scripts/example_usage.py

@@ -0,0 +1,20 @@
+# example_usage.py
+from chat_with_persistence import initialize_chat_session, chat_with_persistence
+
+def main():
+    # Initialize a new chat session
+    index = initialize_chat_session()
+    
+    # First query
+    print("### Initial Query ###")
+    response = chat_with_persistence("What's the current weather in London?", index)
+    print(f"Assistant: {response}")
+    
+    # Follow-up query
+    print("\n### Follow-Up Query ###")
+    follow_up = chat_with_persistence("What about tomorrow's forecast?", index)
+    print(f"Assistant: {follow_up}")
+    
+if __name__ == "__main__":
+    main()
+