# docbot

![license](https://img.shields.io/badge/license-MIT-blue)
![status](https://img.shields.io/badge/status-active-brightgreen)
![python](https://img.shields.io/badge/python-3.9+-blue)
![code style](https://img.shields.io/badge/code%20style-black-black)
![PRs](https://img.shields.io/badge/PRs-welcome-brightgreen)

> AI assistant that answers questions from your documentation

## table of contents

- [install](#install)
- [usage](#usage)
- [api](#api)
- [features](#features)
- [configuration](#configuration)
- [contributing](#contributing)
- [license](#license)

## install

```bash
pip install docbot

usage

from docbot import DocBot

# initialize with your documentation directory
bot = DocBot(docs_path="./docs")

# index your documentation
bot.index()

# ask questions
response = bot.ask("How do I configure authentication?")
print(response.answer)
print(f"Sources: {response.sources}")

cli usage

# index documentation
docbot index ./docs

# start interactive session
docbot chat ./docs

# single question
docbot ask ./docs "What are the API rate limits?"

api

DocBot(docs_path, model="gpt-3.5-turbo", chunk_size=512)

Main class for the documentation chatbot.

Parameters:

• docs_path (str): Path to documentation directory
• model (str): LLM model to use for answers
• chunk_size (int): Size of text chunks for indexing

Methods:

index(file_patterns=["*.md", "*.txt"])

Index documentation files for retrieval.

bot.index(file_patterns=["*.md", "*.rst"])

ask(question, max_sources=3)

Ask a question and get an answer with sources.

response = bot.ask("How do I install the package?", max_sources=5)

Returns: Response object with answer and sources attributes.

clear_index()

Clear the current index.

bot.clear_index()

Response

Object returned by ask() method.

Attributes:

• answer (str): Generated answer to the question
• sources (list): List of source documents used
• confidence (float): Confidence score between 0 and 1

features

• Multiple formats: Supports Markdown, reStructuredText, plain text
• Vector search: Uses embeddings for semantic document retrieval
• Source attribution: Shows which docs were used for each answer
• Local or cloud: Run with local models or OpenAI/Anthropic APIs
• Incremental indexing: Only re-index changed files

configuration

Create a docbot.yaml file in your project root:

model: gpt-4
chunk_size: 512
overlap: 50
embedding_model: text-embedding-ada-002
max_sources: 3
file_patterns:
  - "*.md"
  - "*.rst"
  - "docs/**/*.txt"

Or set environment variables:

export DOCBOT_MODEL=gpt-4
export DOCBOT_CHUNK_SIZE=512
export OPENAI_API_KEY=your_key_here

contributing

prs welcome. open an issue first for big changes.

license

MIT