Implement Chat with RAG and Suggested Questions for Summarized News

[hem210]

Description
We need to implement a chat box on the platform that allows users to query the news summaries and receive answers based on RAG (Retrieval-Augmented Generation). Additionally, we want to display suggested questions derived from the summaries, offering a more guided interaction for the user.

Requirements:

1. Chat Box with RAG:

   • Implement a chat box on the Streamlit frontend where users can ask questions related to the displayed news summaries.
   • Use a RAG pipeline to:
     • Retrieve relevant information from the combined summaries.
     • Generate answers by leveraging the retrieved context.
   • Display the answers within the chat interface.

2. Suggested Questions:

   • Automatically generate and display suggested starter questions to guide the user.
   • Suggested questions should be based on the content in the summaries and should offer common or interesting inquiries the user might have.
   • Display these questions above or near the chat input box, allowing users to click on them to populate the chat box.

3. Backend Implementation:

   • Summarized News Retrieval:
     • Implement a retrieval mechanism that can extract relevant information from the corpus of news summaries.
   • Generation of Responses:
     • Use an (open source) LLM to answer user queries based on the retrieved context.
   • Suggested Questions Generation:
     • Leverage NLP techniques to automatically generate a set of suggested questions from the summaries.

4. Frontend (Streamlit):

   • Chat Box UI:
     • Implement a chat interface where the user can type queries and receive answers in real time.
     • Handle the interaction flow between asking questions and displaying answers.
   • Display Suggested Questions:
     • Show a list of generated suggested questions to the user as a starter guide.
     • Allow users to click on a suggested question to automatically populate the query field.

---

Files to Create/Change:

• src/api/routes/chat.py

  • Defines the routes for the chat interface, including question answering and retrieving suggestions.

• src/api/models/rag.py

  • The RAG pipeline, including the retrieval and generation logic for answering questions.

• src/api/schemas/chat.py

  • Pydantic schemas for validating user queries and responses.

• src/api/dependencies/rag.py

  • Dependencies for managing the retrieval and generation processes.

---

Folder Structure to Follow:

src/
├── api/
│   ├── main.py
│   ├── routes/
│   │   └── chat.py
│   ├── models/
│   │   └── rag.py
│   ├── schemas/
│   │   └── chat.py
│   ├── dependencies/
│   │   └── rag.py
│   └── core/
│       ├── config.py
│       └── security.py
├── pipeline/
│   └── question_generator.py

---

Checklist:

• [ ] Chat box interface created in Streamlit.
• [ ] RAG pipeline implemented for retrieving and generating answers from news summaries.
• [ ] Suggested questions are generated from the summaries and displayed to the user.
• [ ] Proper error handling and logging added.
• [ ] Folder structure implemented as outlined.

---

Considerations:

• Ensure the user experience is smooth with low latency on the query/response interaction.
• Follow RESTful API design principles and use FastAPI’s capabilities to handle requests efficiently.
• Consider scalability and caching for frequently queried news summaries.

[Devasy23]

@say-het do you want to perform this task?