// the find
middlewarehq/middleware
✨ Open-source DORA metrics platform for engineering teams ✨
Middleware is a self-hosted DORA metrics dashboard that pulls data from GitHub/GitLab PRs, CI/CD workflows, and incident tools to give engineering managers deployment frequency, lead time, change failure rate, and MTTR charts. It targets engineering leaders who want DORA visibility without sending their git history to a SaaS vendor. The Docker-based quick start is genuinely fast — one command and you have a working instance.
1. PR-first DORA calculation is smart: you can get meaningful metrics from just your git data without wiring up a separate deployment pipeline, which removes the biggest onboarding blocker for most teams. 2. The architecture is clean for a self-hosted tool — Flask analytics server, Flask sync server, and Next.js frontend are decoupled enough that you can debug each independently; the per-service log tailing commands in the README reflect real operational thinking. 3. The `make_new_setting.py` codegen script is a nice touch — it means contributors don't need to learn the full settings system architecture to add a field, which actually matters for OSS contribution velocity. 4. SQL migrations in `database-docker/db/migrations` are plain SQL files with timestamps, not wrapped in an ORM — you can read exactly what changed in the schema without decoding abstractions.
1. 16GB RAM minimum for local dev is a real barrier; it means the Dockerfile is doing too much inside a single container (Postgres, Redis, two Flask servers, Next.js all together), which will frustrate anyone on a standard dev machine. 2. The roadmap section in the README literally says 'Coming Soon!' with no content — for a tool where teams are evaluating whether to commit to it, the absence of any direction on integrations (PagerDuty, Jira, Linear) is a gap that will cost signups. 3. MTTR depends on incident data, and the only supported incident sources appear to be git-based incidents and a few third-party services — if your team uses something not on that list, you get a blank chart with no clear path to extend it without reading deep into the ETL layer. 4. The AI feature (`dora_ai.py`, `ai_analytics_service.py`) is present but completely undocumented — no mention in the README of what it does, what model it calls, or what it costs, which means you can't evaluate whether to enable it.