Slack GPT-Assistant is a Slack bot that answers your questions by combining a local PDF-based knowledge base with OpenAI’s GPT-4o-mini.
It listens for /ask commands, immediately acknowledges receipt, then performs RAG-style lookups in your PDF docs (via Redis cache) and on-demand queries to the OpenAI API.
-
PDF Knowledge Base
Upload any PDFs (manuals, guides, specifications) and let the bot search and extract relevant context. -
GPT-4o-mini Integration
Generates coherent, human-like answers and follow-up explanations when PDF context alone is insufficient. -
Redis Caching
Stores recent Q&A to save tokens and deliver instant responses on repeated questions. -
Asynchronous Workflow
Immediate “Request received” in Slack, with detailed answer delivered shortly after—no UI hangs. -
Full Docker Packaging
Single container includes your app, Redis, and all dependencies—deploy anywhere in minutes.
- Python 3.11+
- Docker
- ngrok (for local development/testing)
- Redis (local or hosted)
- Slack App credentials:
- Bot Token (
SLACK_BOT_TOKEN) - Signing Secret (
SLACK_SIGNING_SECRET)
- Bot Token (
- OpenAI API Key (
OPENAI_API_KEY)
Copy .env.example to .env and fill in:
SLACK_BOT_TOKEN=…
SLACK_SIGNING_SECRET=…
OPENAI_API_KEY=…
REDIS_HOST=…
REDIS_PORT=…
REDIS_PASSWORD=…
PDF_DATA_DIR=./data/company_docs
MAX_CONTEXT_TOKENS=7000 # optional override
## Setup Instructions
1. **Clone the Repository:**
```bash
git clone <repository_url>
cd slack-gpt-assistant- Configure Environment Variables:
- Copy the template file:
cp .env.example .env
- Edit
.envto provide your Slack, OpenAI, and Redis credentials, and specify the path to your PDF documents.
- Copy the template file:
- Prepare PDF Documents:
Place all required PDF documents into the directory specified by
PDF_DATA_DIR.
Build the Docker image with the following command:
docker build -t slack-gpt-assistant .Run the Docker container:
docker run -p 3000:3000 --env-file .env -v $(pwd)/data:/app/data slack-gpt-assistantExplanation:
-p 3000:3000: Maps the container's port 3000 to the host.--env-file .env: Loads environment variables.-v $(pwd)/data:/app/data: Mounts your localdatadirectory (containing PDFs) into the container.
- Expose Local Server:
Use ngrok to expose your local server:
ngrok http 3000
- Configure Request URL:
In your Slack App settings, set the Request URL for slash commands (e.g.,
/askand/help) to the ngrok URL (e.g.,https://<ngrok-id>.ngrok.io/slack/events).
- /ask : Ask the bot a question. The bot acknowledges immediately, then provides an answer asynchronously based on a search of PDF documents and OpenAI API results.
- /help: Displays help information outlining how to use the bot.
- Slack Command Handling: Responds to
/askand/helpcommands in Slack. - PDF Knowledge Base: Extracts text from PDFs using PyMuPDF for contextual responses.
- Redis Caching: Caches Q&A pairs to improve performance.
- OpenAI Integration: Uses gpt-4o-mini to generate responses.
- Asynchronous Processing: Provides immediate command acknowledgment with asynchronous detailed responses.
- Logging & Error Handling: Logs significant events and gracefully handles errors.
For a high-level overview of the project's architecture, refer to the Architecture Document. The app is built as a monolithic Docker container with dedicated services for handling Slack events, PDF data extraction, OpenAI API interactions, and Redis caching.
For local development, you can run a Redis server using Docker. For example, execute the following command:
docker run --name slack-redis -p 6379:6379 -d redis:6.2-alpineThen, update your .env file with the following settings:
REDIS_HOST=localhostREDIS_PORT=6379- (Optional) Set
REDIS_PASSWORDif your Redis instance requires authentication.
Install development dependencies:
pip install -r requirements-dev.txtFormat, lint, and test your code:
black .
isort .
flake8 .
mypy .
pytesttiktoken(for token counting, used in KnowledgeBaseService context selection)