Collmbo

A Slack bot that lets you chat with 100+ LLMs via LiteLLM. Pronounced the same as "Colombo".

Quick Start

Collmbo supports multiple LLMs, but let's begin with OpenAI's gpt-5.5 model for a quick setup.

1. Create a Slack App

Create a Slack app and obtain the required tokens:

• App-level token (xapp-1-...)
• Bot token (xoxb-...)

2. Create a .env File

Save your credentials in a .env file:

SLACK_APP_TOKEN=xapp-1-...
SLACK_BOT_TOKEN=xoxb-...
LLM_MODEL=gpt-5.5
OPENAI_API_KEY=sk-...

3. Run Collmbo Container

Start the bot using Docker:

docker run -it --env-file .env ghcr.io/iwamot/collmbo:latest

Note

For versioned releases, you can specify a tag like x.x.x. For more details, please check the list of available tags.

4. Say Hello!

Mention the bot in Slack and start chatting:

@Collmbo hello!

Collmbo should respond in channels, threads, and DMs. The conversation history it sees depends on where you talk to it:

• Channel: only the message that mentions the bot
• Thread: all messages in the thread
• DM (outside a thread): the most recent 100 messages from the last 24 hours. Older DM messages are not remembered, so a conversation from the previous day starts fresh.

Want to Use a Different LLM?

First, pick your favorite LLM from LiteLLM supported providers.

To use it, update the relevant environment variables in your .env file and restart the container.

Here are some examples:

Azure OpenAI (gpt-5.5)

SLACK_APP_TOKEN=xapp-1-...
SLACK_BOT_TOKEN=xoxb-...
LLM_MODEL=azure/<your_deployment_name>

AZURE_API_KEY=...
AZURE_API_BASE=...
AZURE_API_VERSION=...

Amazon Bedrock (Claude Opus 4.8)

SLACK_APP_TOKEN=...
SLACK_BOT_TOKEN=...
LLM_MODEL=bedrock/global.anthropic.claude-opus-4-8

# You can specify a Bedrock region if it's different from your default AWS region
AWS_REGION_NAME=us-west-2

# You can use your access key for authentication, but IAM roles are recommended
AWS_ACCESS_KEY_ID=...
AWS_SECRET_ACCESS_KEY=...

Deployment

Collmbo does not serve endpoints and can run in any environment with internet access.

Features

• MCP Tools - Extends functionality with MCP server.
• Tools (Function Calling) - Extends functionality with function calling.
• Vector Store - Retrieves context from a knowledge base (RAG).
• Custom Callbacks - Hooks into requests and responses for custom processing.
• Redaction - Masks sensitive information before sending requests.
• Slack-friendly Formatting - Formats messages for better readability in Slack.
• Image Input - Enables AI models to analyze uploaded images.
• PDF Input - Enables AI models to analyze uploaded PDFs.
• Prompt Caching - May help reduce costs by reusing context (Anthropic Claude, Amazon Bedrock).

Configuration

Collmbo runs with default settings, but you can customize its behavior by setting optional environment variables.

Contributing

Contributions are welcome! Please see our Contributing Guide for details.

All commits must be signed with DCO (git commit -s).

Related Projects

• seratch/ChatGPT-in-Slack - The original project by @seratch.

License

The code in this repository is licensed under the MIT License.

The Collmbo icon (assets/icon.png) is licensed under CC BY-NC-SA 4.0. For example, you may use it as a Slack profile icon.