Micro Slack Assistant

This is a placeholder, further documentation will be more detailed.

Getting Started

See .env.example and project/api to get an overview of what this service does. Currently available tools are mostly implemented according to LangChain examples

Goals

• As a Developer, I'd like to be able to deploy a Slack assistant that can be installed in any workspace

• As a User, I'd like to message the bot directly or mention it on a channel, and have an answer to my query

• As a User, I'd like my bot to have access to web search and internal documentation, and use these resources to answer when appropriate

TODOs

Most important first:

Security

• Proper handling of Slack security (verification token, webhook url challenge)
• Implement Token Auth for management endpoints

Deployment

• Allow dynamic setting of tool docstrings (override source defaults)
• Check viability of deploying SQLite or change vector store
• Deploy on Docker with GitHub Actions

Readability

• Check LangChain usage for anti-patterns and refactoring opportunities
• Proper project documentation

Performance

• Metrics on used tokens, test lighter models for a few tasks, general cost reduction
• Viability of LLM inference caching by query semantics

Possible Features

• Message history and expanded context for answers
• Multiple Retrievers (Library docs, community references, RSS Feeds)
• Slash commands (/tldr, /remindme, /research)