changing

app.py

@@ -1,13 +1,14 @@
 import os
 import streamlit as st
 from together import Together
-from langchain.vectorstores import Chroma
-from langchain.embeddings import HuggingFaceBgeEmbeddings
+from langchain_community.vectorstores import Chroma
+from langchain_community.embeddings import HuggingFaceBgeEmbeddings
 
 # --- Configuration ---
+# TogetherAI API key (env var name pilotikval)
 TOGETHER_API_KEY = os.environ.get("pilotikval")
 if not TOGETHER_API_KEY:
-    st.error("Missing TOGETHER_API_KEY environment variable.")
+    st.error("Missing pilotikval environment variable.")
     st.stop()
 
 # Initialize TogetherAI client
@@ -51,64 +52,71 @@ vectorstore = Chroma(
     persist_directory=persist_directory,
     embedding_function=embeddings
 )
-retriever = vectorstore.as_retriever(search_kwargs={"k":20})
+retriever = vectorstore.as_retriever(search_kwargs={"k": 20})  # k=20
 
-# System prompt template
+# System prompt template with instruction for detailed long answers
 def build_system(context: str) -> dict:
-    return {"role": "system", "content": (
-        "You are a helpful assistant for medical professionals. "
-        "Use the following context from medical documents to answer the question. "
-        "If you don't know, say you don't know.\n\nContext:\n" + context
-    )}
-
-st.title("🩺 DocChatter RAG (Streaming)")
+    return {
+        "role": "system",
+        "content": (
+            "You are an expert medical assistant. Provide a thorough, detailed, and complete answer. "
+            "If you don't know, say you don't know.\n"
+            "Use the following context from medical docs to answer.\n\n"
+            "Context:\n" + context
+        )
+    }
+
+st.title("🩺 DocChatter RAG (Streaming & Memory)")
 
 # Initialize chat history
 if 'chat_history' not in st.session_state:
-    st.session_state.chat_history = []
+    st.session_state.chat_history = []  # list of dicts {role, content}
 
-# Get user input (must be at top-level)
-prompt = st.chat_input("Ask anything about your docs...")
+# Get user input at top level
+user_prompt = st.chat_input("Ask anything about your docs…")
 
-# Tabs for display
+# Tabs for UI
 chat_tab, clear_tab = st.tabs(["Chat", "Clear History"])
 
 with chat_tab:
-    # Display conversation so far
+    # Display existing chat
     for msg in st.session_state.chat_history:
         st.chat_message(msg['role']).write(msg['content'])
 
-    # If user has just submitted
-    if prompt:
-        # Show user message
-        st.chat_message("user").write(prompt)
-        st.session_state.chat_history.append({"role": "user", "content": prompt})
+    # On new input
+    if user_prompt:
+        # Echo user
+        st.chat_message("user").write(user_prompt)
+        st.session_state.chat_history.append({"role": "user", "content": user_prompt})
 
-        # Retrieve context
-        docs = retriever.get_relevant_documents(prompt)
+        # Retrieve top-k docs
+        docs = retriever.get_relevant_documents(user_prompt)
         context = "\n---\n".join([d.page_content for d in docs])
 
-        # Assemble messages for streaming
+        # Build message sequence: system + full history
         messages = [build_system(context)]
-        # Include full history
         for m in st.session_state.chat_history:
-            if m['role'] in ('user','assistant'):
-                messages.append(m)
+            messages.append(m)
 
-        # Stream assistant response
+        # Prepare streaming response
         response_container = st.chat_message("assistant")
-        placeholder = response_container.empty()
+        stream_placeholder = response_container.empty()
         answer = ""
+
+        # Stream tokens
         for token in client.chat.completions.create(
             model="meta-llama/Llama-4-Scout-17B-16E-Instruct",
             messages=messages,
+            max_tokens=22048,
+            temperature=0.1,
             stream=True
         ):
-            if hasattr(token, 'choices'):
+            if hasattr(token, 'choices') and token.choices[0].delta.content:
                 delta = token.choices[0].delta.content
                 answer += delta
-                placeholder.write(answer)
-        # Save response
+                stream_placeholder.write(answer)
+
+        # Save assistant response
         st.session_state.chat_history.append({"role": "assistant", "content": answer})
 
 with clear_tab:
@@ -116,5 +124,5 @@ with clear_tab:
         st.session_state.chat_history = []
         st.experimental_rerun()
 
-# Optionally persist new embeddings
+# (Optional) persist new docs
 # vectorstore.persist()