Implement NL2SQL tooling

.gitignore

@@ -21,7 +21,7 @@ sbin/**
 !opentofu/examples/manual-test.sh
 !src/entrypoint.sh
 !src/client/spring_ai/templates/env.sh
-tests/db_startup_temp/**
+test*/db_startup_temp
 
 ##############################################################################
 # Environment (PyVen, IDE, etc.)

.pylintrc

@@ -52,7 +52,7 @@ ignore=CVS,.venv
 # ignore-list. The regex matches against paths and can be in Posix or Windows
 # format. Because '\\' represents the directory delimiter on Windows systems,
 # it can't be used as an escape character.
-ignore-paths=.*[/\\]wip[/\\].*,src/client/mcp,docs/themes/relearn,docs/public,docs/static/demoware
+ignore-paths=.*[/\\]wip[/\\].*,src/client/mcp,docs/themes/relearn,docs/public,docs/static/demoware,src/server/agents/chatbot.py
 
 # Files or directories matching the regular expression patterns are skipped.
 # The regex matches against base names, not paths. The default value ignores

src/client/content/chatbot.py

@@ -16,7 +16,7 @@
 from streamlit import session_state as state
 
 from client.content.config.tabs.models import get_models
-from client.utils import st_common, api_call, client
+from client.utils import st_common, api_call, client, vs_options
 from client.utils.st_footer import render_chat_footer
 from common import logging_config
 
@@ -26,15 +26,15 @@
 #############################################################################
 # Functions
 #############################################################################
-def show_vector_search_refs(context):
+def show_vector_search_refs(context, vs_metadata=None):
     """When Vector Search Content Found, show the references"""
     st.markdown("**References:**")
     ref_src = set()
     ref_cols = st.columns([3, 3, 3])
     # Create a button in each column
-    for i, (ref_col, chunk) in enumerate(zip(ref_cols, context[0])):
+    for i, (ref_col, chunk) in enumerate(zip(ref_cols, context["documents"])):
         with ref_col.popover(f"Reference: {i + 1}"):
-            chunk = context[0][i]
+            chunk = context["documents"][i]
             logger.debug("Chunk Content: %s", chunk)
             st.subheader("Reference Text", divider="red")
             st.markdown(chunk["page_content"])
@@ -46,9 +46,32 @@ def show_vector_search_refs(context):
             except KeyError:
                 logger.error("Chunk Metadata NOT FOUND!!")
 
-    for link in ref_src:
-        st.markdown("- " + link)
-    st.markdown(f"**Notes:** Vector Search Query - {context[1]}")
+    # Display Vector Search details in expander
+    if vs_metadata or ref_src:
+        with st.expander("Vector Search Details", expanded=False):
+            if ref_src:
+                st.markdown("**Source Documents:**")
+                for link in ref_src:
+                    st.markdown(f"- {link}")
+
+            if vs_metadata and vs_metadata.get("searched_tables"):
+                st.markdown("**Tables Searched:**")
+                for table in vs_metadata["searched_tables"]:
+                    st.markdown(f"- {table}")
+
+            if vs_metadata and vs_metadata.get("context_input"):
+                st.markdown(f"**Search Query:** {vs_metadata.get('context_input')}")
+            elif context.get("context_input"):
+                st.markdown(f"**Search Query:** {context.get('context_input')}")
+
+
+def show_token_usage(token_usage):
+    """Display token usage for AI responses using caption"""
+    if token_usage:
+        prompt_tokens = token_usage.get("prompt_tokens", 0)
+        completion_tokens = token_usage.get("completion_tokens", 0)
+        total_tokens = token_usage.get("total_tokens", 0)
+        st.caption(f"Token usage: {prompt_tokens} prompt + {completion_tokens} completion = {total_tokens} total")
 
 
 def setup_sidebar():
@@ -62,7 +85,7 @@ def setup_sidebar():
     st_common.tools_sidebar()
     st_common.history_sidebar()
     st_common.ll_sidebar()
-    st_common.vector_search_sidebar()
+    vs_options.vector_search_sidebar()
 
     if not state.enable_client:
         st.stop()
@@ -80,22 +103,33 @@ def create_client():
 
 
 def display_chat_history(history):
-    """Display chat history messages"""
+    """Display chat history messages with metadata"""
     st.chat_message("ai").write("Hello, how can I help you?")
     vector_search_refs = []
 
     for message in history or []:
         if not message["content"]:
             continue
 
-        if message["role"] == "tool" and message["name"] == "oraclevs_tool":
+        if message["role"] == "tool" and message["name"] == "optimizer_vs-retriever":
             vector_search_refs = json.loads(message["content"])
 
         elif message["role"] in ("ai", "assistant"):
             with st.chat_message("ai"):
                 st.markdown(message["content"])
+
+                # Extract metadata from response_metadata
+                response_metadata = message.get("response_metadata", {})
+                vs_metadata = response_metadata.get("vs_metadata", {})
+                token_usage = response_metadata.get("token_usage", {})
+
+                # Show token usage immediately after message
+                if token_usage:
+                    show_token_usage(token_usage)
+
+                # Show vector search references if available
                 if vector_search_refs:
-                    show_vector_search_refs(vector_search_refs)
+                    show_vector_search_refs(vector_search_refs, vs_metadata)
                     vector_search_refs = []
 
         elif message["role"] in ("human", "user"):
@@ -131,9 +165,32 @@ async def handle_chat_input(user_client):
         try:
             message_placeholder = st.chat_message("ai").empty()
             full_answer = ""
-            async for chunk in user_client.stream(message=human_request.text, image_b64=file_b64):
-                full_answer += chunk
-                message_placeholder.markdown(full_answer)
+
+            # Animated thinking indicator
+            async def animate_thinking():
+                """Animate the thinking indicator with increasing dots"""
+                dots = 0
+                while True:
+                    message_placeholder.markdown(f"🤔 Thinking{'.' * (dots % 4)}")
+                    dots += 1
+                    await asyncio.sleep(0.5)  # Update every 500ms
+
+            # Start the thinking animation
+            thinking_task = asyncio.create_task(animate_thinking())
+
+            try:
+                async for chunk in user_client.stream(message=human_request.text, image_b64=file_b64):
+                    # Cancel thinking animation on first chunk
+                    if thinking_task and not thinking_task.done():
+                        thinking_task.cancel()
+                        thinking_task = None
+                    full_answer += chunk
+                    message_placeholder.markdown(full_answer)
+            finally:
+                # Ensure thinking task is cancelled
+                if thinking_task and not thinking_task.done():
+                    thinking_task.cancel()
+
             st.rerun()
         except (ConnectionError, TimeoutError, api_call.ApiError) as ex:
             logger.exception("Error during chat streaming: %s", ex)

src/client/content/testbed.py

@@ -17,7 +17,7 @@
 
 from client.content.config.tabs.models import get_models
 
-from client.utils import st_common, api_call
+from client.utils import st_common, api_call, vs_options
 
 from common import logging_config
 
@@ -493,7 +493,7 @@ def render_evaluation_ui(available_ll_models: list) -> None:
     st.info("Use the sidebar settings for chatbot evaluation parameters", icon="⬅️")
     st_common.tools_sidebar()
     st_common.ll_sidebar()
-    st_common.vector_search_sidebar()
+    vs_options.vector_search_sidebar()
     st.write("Choose a model to judge the correctness of the chatbot answer, then start evaluation.")
     col_left, col_center, _ = st.columns([4, 3, 3])
 
@@ -510,20 +510,13 @@ def render_evaluation_ui(available_ll_models: list) -> None:
         on_change=st_common.update_client_settings("testbed"),
     )
 
-    # Check if vector search is enabled but no vector store is selected
-    evaluation_disabled = False
-    if state.client_settings.get("vector_search", {}).get("enabled", False):
-        # If vector search is enabled, check if a vector store is selected
-        if not state.client_settings.get("vector_search", {}).get("vector_store"):
-            evaluation_disabled = True
-
     if col_center.button(
         "Start Evaluation",
         type="primary",
         key="evaluate_button",
         help="Evaluation will automatically save the TestSet to the Database",
         on_click=qa_update_db,
-        disabled=evaluation_disabled,
+        disabled=not state.enable_client,
     ):
         with st.spinner("Starting Q&A evaluation... please be patient.", show_time=True):
             st_common.clear_state_key("testbed_evaluations")