Kb-Slackbot

A Slack-integrated assistant that enables natural language querying of internal documentation such as runbooks, ADRs, and wikis. It uses OpenAI embeddings with a Chroma vector store for semantic search and returns summarized, context-aware answers directly in Slack. Falls back to GPT-based general knowledge when no relevant information is found.

Features

• 🔍 Semantic Search: Uses OpenAI embeddings to semantically match questions to internal documentation
• 🧠 Contextual Summarization: Returns only the most relevant excerpts, not full wiki pages
• 💬 Slack Integration: Ask questions directly in Slack via mentions
• 📄 Source Linking: Provides links to the original documentation sources
• 🧩 Fallback to GPT: Uses GPT for DevOps/infra questions when no internal match is found
• 🚫 Reduces Tribal Knowledge Dependency: Engineers don't need to know specific document titles or locations

Example Use Cases

• "How do I block an IP in AWS WAF?"
• "Why is the Go audit indexer lagging?"
• "How do I manually reindex the org?"

Architecture Overview

Slack → Slack Bolt App (Python) → 
↓
Extract Question → Embed via OpenAI → Query Chroma Vector Store → 
↓                                         ↓
Match Found?                           No Match
↓                                          ↓
Summarize + Cite Source        →      Use GPT to Answer
↓
Respond in Slack Thread

Setup Instructions

1. Clone the Repository

git clone https://github.com/reshmachitre/Kb-Slackbot.git
cd slack-knowledge-assistant

2. Create and Activate a Virtual Environment

python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Environment Variables

Create a .env file in the project root:

SLACK_BOT_TOKEN=xoxb-your-slack-bot-token
SLACK_APP_TOKEN=xapp-your-slack-app-token
OPENAI_API_KEY=sk-...

5. Populate Chroma Vector Store

You must embed and store your internal documents (runbooks, ADRs, etc.) in ChromaDB using OpenAI embeddings. This typically happens in a separate ingestion script.

Ensure your chroma_store/ contains the indexed vectors.

Note: If needed, we can provide a script to do this.

6. Run the Bot

python app.py

The bot will start and listen for mentions via Socket Mode.

Slack Usage

Mention the bot in a Slack channel or DM:

@knowledge-bot How do I reindex the catalog?

It will respond with:

• A summary of the best-matching documentation
• Relevant source links
• A fallback GPT response if no internal docs matched

Document Source Mapping

The bot uses a static map of source files to URLs (e.g., Confluence pages). Update the SOURCE_LINKS dictionary in app.py as you add more documents.

Dependencies

• Slack Bolt (Python)
• OpenAI Python SDK
• ChromaDB
• python-dotenv

Install them with:

pip install -r requirements.txt

Troubleshooting

• Missing ChromaDB or Documents: Ensure your vector store is properly populated and path is correct
• Bot Not Responding: Check your Slack app permissions, tokens, and that it's running in Socket Mode
• Filtered Error: Ensure variable scopes are consistent if modifying the get_context() logic

Future Improvements

• Add streaming responses to Slack
• Allow ingestion via Slack DM/upload
• Admin UI for document ingestion and monitoring
• Add confidence scores in response metadata