[Feature] Add Database Selection Checkboxes (arxiv, indico, both) to Chatbot

[karthik18495]

Summary

Enhance the RAG4EIC chatbot interface to allow users to choose which vector database(s) to query: arxiv, indico, or both. Add two checkboxes, one for arxiv and one for indico. If both are checked, the chatbot should search across both databases for relevant responses.

Requirements

• Add two checkboxes to the chatbot UI (likely in the sidebar or top of the chat area):
  • "Search arxiv database"
  • "Search indico database"
• If both are checked, the query must be searched through both databases, and results merged for response generation.
• If only one is checked, search only the selected database.
• If neither is checked, show a warning or disable question submission until a selection is made.
• Results from both databases must be combined, deduplicated (if necessary), and passed as context to the LLM.
• Indicate to the user which database(s) were used in generating the answer (e.g., via a message or icon in the chat).

Implementation Plan

1. UI Changes

   • In the chatbot UI (likely streamlit_app/pages/2_RAG-ChatBot.py), add two checkboxes for "arxiv" and "indico" (using st.checkbox).
   • Store their states in st.session_state["use_arxiv"] and st.session_state["use_indico"].
   • If neither is checked, disable the chat input or display a warning.

2. Backend Logic

   • Instantiate or select the appropriate retriever(s) based on user selection:
     • If both are checked, run the query on both retrievers and merge results.
     • If only one is checked, run only that retriever.
   • For merging, combine results from both databases, deduplicate by unique document ID or content, and preserve source info for citations.
   • Ensure downstream code (LLM prompt, citation, etc.) can handle the merged results and source metadata.

3. Indico Database Integration

   • If not already, implement or update vectorstore/retriever logic for the indico database, similar to arxiv.
   • Ensure both retrievers provide results in compatible formats.

4. User Feedback

   • Display a message or icon in the chat indicating which database(s) were searched for each response.
   • Optionally, allow users to filter or expand/collapse results by source in the chat history.

5. Testing

   • Test all combinations: arxiv only, indico only, both checked, neither checked.
   • Check result merging, deduplication, and source indication in answers.

Future Improvements

• Allow users to prioritize one database over the other (e.g., arxiv first, then indico fallback)
• Add support for more databases in future (use a list of checkboxes or a multi-select)
• Allow user to see search results from each database separately before fusion

References

• See app_utilities.py and pages/2_RAG-ChatBot.py for retriever and chat logic
• Streamlit checkbox docs: https://docs.streamlit.io/library/api-reference/widgets/st.checkbox
• Vectorstore integration patterns (LangChain docs)