Released a DDEV Add-on for Langfuse

Shipped ddev-langfuse today, a DDEV add-on that gives you a full local Langfuse instance for LLM observability, tracing, and prompt management.

If you haven’t run into it yet: Langfuse is an open-source tool for watching what your AI features actually do at runtime. Every prompt, every tool call, every model response gets logged as a trace, so when an agent gives a weird answer you can open the exact request that produced it instead of guessing. It also handles prompt versioning and evals, which matters once you have more than one AI feature to keep track of.

Run ddev add-on get abhisekmazumdar/ddev-langfuse and it adds Langfuse’s whole self-hosted stack (web, worker, Postgres, ClickHouse, Redis, MinIO) straight onto your project’s own Docker network. Secrets get generated with openssl rand on first install, and an org, project, and API keypair get bootstrapped headlessly, so there’s no .env to hand-edit and no signup wizard to click through. Credentials print at the end of ddev add-on get, or any time via ddev langfuse-credentials.

Built with the Drupal AI ecosystem in mind: drupal/ai, drupal/ai_agents, drupal/langfuse. It’s not Drupal-specific though. Anything sending traces to a Langfuse-compatible endpoint can point at http://langfuse-web:3000 from inside the same DDEV project.

A stable release is available to install. If you’re working on AI features in Drupal and want traces without standing up a separate Langfuse deployment, give it a try. Reach out on LinkedIn or Drupal Slack (username abhisekmazumdar) if you hit anything.