Omni: A Workplace AI Assistant That Actually Makes Sense
The Problem
Workplace apps are great—until you can’t find the file, email, or Jira ticket you need. You know it’s there, but where? Somewhere in the 47 tabs you have open? Searching across tools is a nightmare, and switching contexts to copy-paste between them kills your flow. Oh, and let’s throw in some enterprise-grade data security concerns for good measure. Fun times.
What This Does
omni is an AI-powered assistant and search platform that connects your disorganized mess of workplace apps—Google Drive, Slack, Confluence, Jira, and more—into a unified search and automation tool. Think of it as your own private GPT-powered assistant, but with boundaries.
The heavy lifting is done by a mix of Rust (for search and indexing), Python (for AI/LLM orchestration), and SvelteKit (for the web UI). Instead of relying on a patchwork of databases, it uses Postgres for everything—BM25 full-text search, pgvector for semantic search, and general app data. One database to rule them all. The connectors, which live in connectors/, are modular containers, so you can mix Python, Rust, and other dependencies without breaking anything. For example, connectors/atlassian/src has all the logic to sync data from Jira and Confluence.
The AI agent? It’s not just chatGPT slapped on top of your data. It can execute code in a sandboxed Docker container (docker/), with strict security limits like read-only filesystems and no internet access. This means it can analyze your data, run Python scripts, and even execute bash commands, all without compromising your infrastructure.
Real-World Use
Say you’re on Slack, and someone asks, “Where’s the latest sales deck?” You fire up Omni’s chat UI, and it seamlessly searches Google Drive, Jira, and Slack messages. It finds the deck in Drive, confirms the version number in Jira, and pastes the link back into Slack. No tab-hopping, no guessing.
Need to analyze some sales data? Upload a CSV file, and the AI agent can spin up a Python script inside a sandboxed container to generate insights. It’s like having an intern who doesn’t ask questions or need coffee breaks.
The Bottom Line
omni solves a problem every enterprise has but few solve well: making sense of scattered data without creating security headaches. The modular connector system is smart, and using Postgres for everything is a bold but practical move. That said, the setup is not for dabblers—if you don’t speak Docker Compose or Terraform, you’re going to have a bad time. But if you’re building a self-hosted, secure, and scalable AI assistant for your org, this is worth a shot.