All projects
ExperimentPersonal running analysis tool

Discord Run Analyst

A Discord running analyst designed around tools I already use, free-tier infrastructure, and GitHub Actions to avoid separately billed AI API calls.

My role
Product direction and development
When
2026
Tools and themes
TypeScript, Cloudflare Workers, Cloudflare D1, COROS MCP, Strava API, Discord, GitHub Actions

A running analyst built around constraints

I wanted a useful analysis after a run without adding another paid API subscription or another app to check. So I designed the project around services I already use, free-tier infrastructure, and a few deliberate workarounds to connect them.

Discord is the interface because it is where I already spend my time. A Telegram bot would have meant checking another app just for this one report.

From a run to a report

A Strava activity triggers the Cloudflare Workers backend, which looks for the corresponding run in COROS and gathers the available details, laps, recent training, and calendar context. Cloudflare D1 stores the jobs and reports so the pipeline can recover from interruptions without processing the same run twice.

The backend calculates the figures before analysis. A private GitHub Actions runner is designed to handle the AI analysis without separate, billed AI API calls. The resulting report is checked against the available evidence before it is saved and delivered to Discord.

Working within the limits

The goal is to run the infrastructure within available free allowances and avoid a separate per-call AI bill. The services still have usage limits, so the design accounts for those limits rather than assuming unlimited free operation.

The core pipeline and private runner have been implemented and tested locally. A complete live report through the runner is still pending, so the end-to-end experience remains a work in progress.