The Open-Source Slack AI App

This repository is a ready-to-run basic Slack AI solution you can host yourself and unlock the ability to summarize threads and channels on demand using OpenAI (support for alternative and open source LLMs will be added if there's demand). The official Slack AI product looks great, but with limited access and add-on pricing, I decided to open-source the version I built in September 2023. Learn more about how and why I built an open-source Slack AI.

Once up and running (instructions for the whole process are provided below), all your Slack users will be able to generate to both public and private:

1. Thread summaries - Generate a detailed summary of any Slack thread (powered by GPT-6 Luna by default)
2. Channel overviews - Generate an outline of the channel's purpose based on the extended message history (powered by an ensemble of NLP models and a little GPT-6 Luna to explain the analysis in natural language)
3. Channel summaries since - Generate a detailed summary of a channel's messages since a given point in time (powered by GPT-6 Luna by default). Now with support for custom prompts! e.g. /tldr_since anonymize the summary. Note: this doesn't include threads yet.
4. Full channel summaries (experimental) - Generate a detailed summary of a channel's extended history (powered by GPT-6 Luna by default). Now with support for custom prompts! e.g. /tldr_extended anonymize the summary. Note: this can get very long!

Note: these features were all working well even on GPT-3.5-Turbo (and GPT-4 for channel overviews)

Table of Contents

• Getting Started
  • Prerequisites
  • Installation
  • Usage
  • Customization
• Testing
• Future Enhancements
• Contributing

Getting Started

Follow these instructions to get a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

Ensure you have the following preconfigured or installed on your local development machine:

• Python 3.9.x - 3.11.x
• OpenAI API key
• Slack App & associated API tokens
• Poetry package manager
• ngrok (recommended)

Installation

1. Clone the repository to your local machine.
2. Navigate to the project directory.
3. Install the required Python packages using Poetry:

poetry install

4. Install the dictionary model

poetry run python -m spacy download en_core_web_md

5. Create a .env file in the root directory of the project, and fill it with your API keys and tokens. Use the example.env file as a template.

cp example.env .env && open .env

Slack app configuration

Make a copy of manifest.json and change the request URL to your ngrok or server URL.

Create a new Slack app here and configure it using your manifest.yaml file.

You shouldn't need to make any other changes but you can change the name, description, and other copy related settings.

If you wish to adjust the name of the slash commands, you'll need to modify slack_server.py.

Once configured, retrieve the "Bot User OAuth Token" from the "Install App" page and add it to your .env file as SLACK_BOT_TOKEN.

Then, on the Basic Information page under the App-Level Tokens heading create a token with the scop connections:write and add it to your .env file as SLACK_APP_TOKEN.

Usage

To run the application, run the FastAPI server:

poetry run uvicorn ossai.slack_server:app --reload

You'll then need to expose the server to the internet using ngrok.

Run ngrok with the following command: ngrok http 8000

Then add the ngrok URL to your Slack app's settings.

Customization

The main customization options are:

• Channel Summary: customize the ChatGPT prompt in topic_analysis.py
• Thread Summary: customize the ChatGPT prompt in summarizer.py

Testing

This project uses pytest and pytest-cov to run tests and measure test coverage.

Unit tests

Run the unit tests (no API keys required):

pytest --cov=ossai tests/ -m "not integration"

Integration tests

Integration tests make real OpenAI API calls to verify the end-to-end summarization flow works with the current dependency versions. They require OPENAI_API_KEY to be set in your environment or .env file.

pytest tests/ -m integration

The integration test uses an unambiguous conversation about migrating from MongoDB to PostgreSQL and asserts that the key topics (mongodb, postgresql, migrat*) appear in the summary. This is intentional: it lets us catch any SDK version bump that silently breaks the actual API integration even when unit tests pass.

Both suites run automatically in CI as part of the same test step.

Future Enhancements

• Move to LangChain & LangSmith for extensibility, tracing, & control
• leverage LangSmith's feedback capabilities to capture & learn from user feedback
• Add a /tldr_since command to summarize a channel's messages since a given date
• Add slack app setup details and sample app manifest to README
• Incorporate threaded conversations in channel-level summaries
• Implement evals suite to complement unit tests
• Add support for alternative and open-source LLMs
• Explore workflow for collecting data & fine-tuning models for cost reduction
• Add support for anonymized message summaries
• Leverage prompt tools like Chain of Destiny
• Add support for pulling supporting context from external sources like company knowledge bases
• Explore caching and other performance optimizations
• Explore sentiment analysis

Contributing

I more than welcome contributions! Please read CONTRIBUTING.md for details on how to submit feedback, bugs, feature requests, enhancements, or your own pull requests.

License

This project is licensed under the GPL-3.0 License - see the LICENSE.md file for details.