Fix ChromaDB permissions with PersistentClient

app.py

@@ -1,7 +1,6 @@
-# app.py 
-
 import os
 import gradio as gr
+import chromadb # Added import
 
 from dotenv import load_dotenv 
 import requests
@@ -17,10 +16,11 @@ from langchain_core.output_parsers import StrOutputParser
 # --- DEPLOYMENT-ONLY FUNCTION ---
 def build_brain_if_needed():
     """Checks if the ChromaDB exists and builds it if it doesn't."""
-    if not os.path.exists("./chroma_db"):
+    # Use an absolute path inside the container for consistency
+    db_path = "/app/chroma_db"
+    if not os.path.exists(db_path):
         print("Database not found. Building now... (This will run only once on the server's first startup)")
         from langchain_community.document_loaders import TextLoader
-        
         from langchain.text_splitter import RecursiveCharacterTextSplitter
         
         loader = TextLoader('knowledge.txt', encoding='utf-8')
@@ -29,10 +29,15 @@ def build_brain_if_needed():
         docs = text_splitter.split_documents(documents)
         embedding_function = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
         
+        # Explicitly create a persistent client pointing to the absolute path
+        persistent_client = chromadb.PersistentClient(path=db_path)
+
+        # Create the Chroma vector store using the client
         db = Chroma.from_documents(
-            docs,
-            embedding_function,
-            persist_directory="./chroma_db"
+            client=persistent_client,
+            documents=docs,
+            embedding=embedding_function, # Correct parameter name is 'embedding'
+            collection_name="churchill_collection" # Good practice to name the collection
         )
         print("Database built successfully.")
     else:
@@ -59,7 +64,16 @@ if not ELEVENLABS_VOICE_ID:
 # Load RAG chain 
 def load_and_build_chain():
     embedding_function = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
-    vectorstore = Chroma(persist_directory="./chroma_db", embedding_function=embedding_function)
+    
+    # Use the same persistent client to load the existing DB
+    persistent_client = chromadb.PersistentClient(path="/app/chroma_db")
+    
+    vectorstore = Chroma(
+        client=persistent_client,
+        embedding_function=embedding_function,
+        collection_name="churchill_collection" # Must use the same collection name
+    )
+    
     retriever = vectorstore.as_retriever()
 
     persona_prompt_template = """
@@ -168,7 +182,8 @@ def process_user_turn(user_input, chat_history):
         return chat_history, audio_file
     except Exception as e:
         print(f"Processing Error: {e}")
-        chat_history.append((user_input, "I'm terribly sorry, something went wrong."))
+        chat_history.append({"role": "user", "content": user_input})
+        chat_history.append({"role": "assistant", "content": "I'm terribly sorry, something went wrong."})
         return chat_history, None
 
 # Gradio UI 
@@ -195,21 +210,18 @@ with gr.Blocks(css="""
 
     def handle_text_submission(message, history):
         history, audio = process_user_turn(message, history)
-        return history, audio
+        return history, audio, ""
 
     def handle_audio_submission(audio_file, history):
         if not audio_file:
-            return history, None
+            return history, None, ""
         transcribed = transcribe_speech(audio_file)
         history, audio = process_user_turn(transcribed, history)
-        return history, audio
-
-    text_in.submit(handle_text_submission, [text_in, chatbot], [chatbot, audio_out])
-    send_btn.click(handle_text_submission, [text_in, chatbot], [chatbot, audio_out])
-    audio_in.stop_recording(handle_audio_submission, [audio_in, chatbot], [chatbot, audio_out])
+        return history, audio, ""
 
-    text_in.submit(lambda: "", None, text_in)
-    send_btn.click(lambda: "", None, text_in)
+    text_in.submit(handle_text_submission, [text_in, chatbot], [chatbot, audio_out, text_in])
+    send_btn.click(handle_text_submission, [text_in, chatbot], [chatbot, audio_out, text_in])
+    audio_in.stop_recording(handle_audio_submission, [audio_in, chatbot], [chatbot, audio_out, text_in])
 
 # Launch app 
 demo.launch(server_name="0.0.0.0")