Neura Market
DataChad logo

DataChad

Free

A streamlit app that lets you chat with any data source. Supporting both OpenAI and local mode with GPT4All. ![GitHub Repo stars](https://img.shields.io/github/stars/gustavz/DataChad?style=social)

FreeFree tier
Inputs: text, file, urlOutputs: text
Type
Open Source
Visit Website GitHub (40 stars)

About DataChad

DataChad is an open-source Streamlit application that allows users to ask questions about any data source. It leverages embeddings, vector databases, large language models (LLMs), and LangChain to process uploaded files, URLs, or paths. Users can create Knowledge Bases from multiple files and Smart FAQs from curated Q&As. The app supports both OpenAI and HuggingFace embeddings, stores vectors in Activeloop's database hub, and uses LangChain with a customizable LLM (default GPT-3.5-turbo). Chat history is cached locally for conversational interactions. DataChad is licensed under Apache-2.0 and is available on GitHub.

Key Features

Upload any file(s), path, or URL to create Knowledge Bases
Create Smart FAQs from curated numbered Q&As
Uses OpenAI or HuggingFace embeddings to embed text chunks
Stores embeddings as vector datasets in Activeloop's database hub
LangChain integrating customizable LLM (default gpt-3.5-turbo) with vector stores
Chat history cached locally for ChatGPT-like conversations
Open source under Apache-2.0 license
Supports both cloud (OpenAI) and local mode with GPT4All

Pros & Cons

Pros
  • User-friendly Streamlit interface for easy interaction
  • Supports multiple file types and sources (files, URLs, paths)
  • Flexible embedding and LLM selection (OpenAI or HuggingFace)
  • Open source and customizable with Apache-2.0 license
  • Local mode available for offline use with GPT4All
Cons
  • Requires Python 3.10 to run
  • Currently limited to Activeloop's database hub for vector storage
  • Advanced options require manual configuration in constants.py
  • Local mode and some features are still under development (V1/V2 legacy)
  • No built-in user authentication or multi-user support yet

Best For

Question answering over uploaded documentsChatting with PDFs and other file formatsBuilding conversational knowledge basesExploring and querying data from URLs or local pathsCreating intelligent FAQ assistants from curated content

FAQ

What models does DataChad support?
By default, DataChad uses OpenAI's GPT-3.5-turbo for the LLM and OpenAI or HuggingFace embeddings. It also supports local mode with GPT4All for offline use.
How does DataChad handle data privacy?
DataChad offers a local mode using GPT4All that runs entirely on your machine, providing offline data processing. Cloud mode uses OpenAI services, so data is sent to OpenAI's servers.
What file formats can I upload?
The app supports any file type that can be loaded and split into text chunks. It handles common formats like PDF, text, and others through its loaders.