ML Chatbot

An ML-powered chatbot that answers technical questions using semantic retrieval from PostgreSQL and context-aware generation with LLaMA 3.1, with Scikit-learn supporting the ML workflow.

Students and learners need quick answers to ML concepts without digging through scattered notes. The project needed a retrieval-backed chatbot that could surface relevant material before generating a response.

I built Flask APIs for document indexing and retrieval, stored content in PostgreSQL, and used semantic search with LLaMA 3.1 to produce context-aware answers. Scikit-learn supports ML-related processing in the pipeline.

Documents are indexed and stored in PostgreSQL. User queries trigger retrieval of relevant content, which is passed to LLaMA 3.1 for generation. Flask exposes the chat and retrieval endpoints.

• Information retrieval for ML concept queries
• PostgreSQL-backed document indexing
• Context-aware responses with LLaMA 3.1
• Flask APIs for chat and retrieval
• Scikit-learn in the ML workflow

• Indexing technical content

  ML concepts needed structured storage and retrieval so the chatbot could surface relevant explanations.

• Retrieval before generation

  Responses had to use retrieved PostgreSQL content as context for LLaMA 3.1 rather than generating blindly.

• API design for chat flows

  Flask endpoints needed to handle indexing, retrieval, and response generation in a maintainable way.

• PostgreSQL for document storage

  Indexed content lives in PostgreSQL for reliable retrieval during chat sessions.

• LLaMA 3.1 for context-aware answers

  Generation uses retrieved context to keep responses tied to stored material.

• Flask service APIs

  Flask exposes indexing, retrieval, and chat endpoints behind a single application.

• Retrieval-backed chat works well for bounded technical domains.
• Document indexing quality affects answer relevance.
• Flask and PostgreSQL provide a solid base for IR-style chatbots.

• Broader ML topic coverage
• Source references in responses
• Improved retrieval ranking