README.md

Hugging Face Documentation Question Answering System

A multi-interface Q&A system that uses Hugging Face's LLM and Retrieval Augmented Generation (RAG) to deliver answers based on Hugging Face documentation. Operable as an API, Discord bot, or Gradio app, it also provides links to the documentation used to formulate each answer.

Example

Table of Contents

• Setting up
• Running
  • Gradio
  • API
  • Discord Bot
• Indexes
• Development instructions

Setting up

To execute any of the available interfaces, specify the required parameters in the .env file based on the .env.example located in the config/ directory. Alternatively, you can set these as environment variables:

• QUESTION_ANSWERING_MODEL_ID - (str) A string that specifies either the model ID from the Hugging Face Hub or the directory containing the model weights
• EMBEDDING_MODEL_ID - (str) embedding model ID from the Hugging Face Hub. We recommend using the hkunlp/instructor-large
• INDEX_REPO_ID - (str) Repository ID from the Hugging Face Hub where the index is stored. List of the most actual indexes can be found in this section: Indexes
• PROMPT_TEMPLATE_NAME - (str) Name of the model prompt template used for question answering, templates are stored in the config/api/prompt_templates directory
• USE_DOCS_FOR_CONTEXT - (bool) Use retrieved documents as a context for a given query
• NUM_RELEVANT_DOCS - (int) Number of documents used for the previous feature
• ADD_SOURCES_TO_RESPONSE - (bool) Include sources of the retrieved documents used as a context for a given query
• USE_MESSAGES_IN_CONTEXT - (bool) Use chat history for conversational experience
• DEBUG - (bool) Provides additional logging

Install the necessary dependencies from the requirements file:

pip install -r requirements.txt

Running

Gradio

After completing all steps as described in the Setting up section, run:

python3 app.py

API Serving

By default, the API is served at http://0.0.0.0:8000. To launch it, complete all the steps outlined in the Setting up section, then execute the following command:

python3 -m api

Discord Bot

To interact with the system as a Discord bot, add additional required environment variables from the Discord bot section of the .env.example file in the config/ directory.

• DISCORD_TOKEN - (str) API key for the bot application
• QA_SERVICE_URL - (str) URL of the API service. We recommend using: http://0.0.0.0:8000
• NUM_LAST_MESSAGES - (int) Number of messages used for context in conversations
• USE_NAMES_IN_CONTEXT - (bool) Include usernames in the conversation context
• ENABLE_COMMANDS - (bool) Allow commands, e.g., channel cleanup
• DEBUG - (bool) Provides additional logging

After completing all steps, run:

python3 -m bot

Indexes List

The following list contains the most current indexes that can be used for the system:

• All Hugging Face repositories over 50 Stars - 512-Character Chunks
• All Hugging Face repositories over 50 Stars - 812-Character Chunks

Development Instructions

We use Python 3.10

To install all necessary Python packages, run the following command:

pip install -r requirements.txt

We use the pipreqsnb to generate the requirements.txt file. To install pipreqsnb, run the following command:

pip install pipreqsnb

To generate the requirements.txt file, run the following command:

pipreqsnb --force .

To run unit tests, you can use the following command:

pytest -o "testpaths=tests" --noconftest