Table of Contents

Kiri Chat

Welcome to Kiri Chat — a chatbot system that uses your documentation as a knowledge base, powered by Retrieval-Augmented Generation (RAG).

Kiri Chat lets you have natural conversations with an AI assistant about your documentation. It uses a Large Language Model (LLM) to answer questions, but only draws information from the documentation you've indexed — ensuring accurate, contextual responses.

The system is built with:

  • docfx — static documentation site generator
  • Qdrant — high-performance vector database for semantic search
  • Ollama — local LLM and embedding models
  • FastAPI — lightweight chat API backend

Why Kiri Chat?

Traditional documentation search relies on keyword matching, which often misses the intent behind your question. Kiri Chat uses semantic search to understand what you mean, then generates answers grounded in your actual documentation — with source links back to the exact sections used.

Key Features

  • Documentation-powered — answers are generated only from your indexed documentation
  • Source attribution — every answer links to the documentation sections used
  • Markdown-aware — responses preserve formatting, code blocks, and lists
  • Embeddable chat widget — floating <chat-button> web component for any page
  • Header-based chunking — document structure is preserved for better context
  • Local processing — runs on your machine using Ollama for embeddings and LLM inference

How It Works

  1. Index — markdown files are chunked by header, embedded with nomic-embed-text, and stored in Qdrant
  2. Search — user questions are embedded and matched against the vector database
  3. Generate — the LLM (gemma:2b) generates answers using only the retrieved documentation context
  4. Cite — source links are returned so you can verify or explore further

Project Overview

Kiri Chat
├── docfx-site/          # Documentation site + chat widget
│   ├── docs/            # Markdown documentation
│   ├── chat-button.js   # Embeddable chat web component
│   └── chat-api/        # FastAPI RAG backend
│       ├── main.py      # /chat endpoint
│       └── index_docs.py # Indexing script
├── Qdrant (Docker)      # Vector database
└── Ollama (local)       # Embeddings + LLM

Ready to get started? Head over to Getting Started to set up Kiri Chat on your machine.

Edit this page