Built for CI: Handling Logs & Artifacts

Like many hosted AI tools, Colimit indexes your codebase to make finding relevant files more effective (actually, it goes beyond the status quo by semantically indexing your codebase in terms of formal models, but that's another story).

However, unlike most other AI tools, Colimit is specially built for:

1. Handling large log files
2. Handling build artifacts generated by your CI runs
3. Preserving Workflow Run History and attaching them to commits instead of PRs

Large Log Files

Logs are often verbose and can sometimes be impenetrable to humans, requiring a lot of squinting and scrolling to figure out what's going on. For LLM's, logs are often difficult due to context size limitations, either due to hard token limits or soft token limits where effectiveness drops off rapidly.

Colimit's platform blends conventional log pruning, searching, and indexing techniques with LLM-based analysis to be able to handle and understand your verbose CI logs, while still retaining good LLM-based reasoning and debugging capabilities.

Build Artifacts

Another CI-specialized feature of Colimit is that it takes into account uploaded build artifacts when doing its debugging analysis.

For example, you might be running end-to-end Playwright integration tests, which take screenshots of the UI as the tests progress or fail, and save them as build artifacts. Or, you might have extra log files in containers for some of your services, which you also upload as build artifacts at the end of your CI run.

Whenever, you upload such build artifacts, Colimit will automatically consider them in its analysis, in a multimodal way (meaning Colimit understands things like images/screenshots, in addition to text).