Every eval framework your team runs.One wire. One ledger.

DeepEval is the pytest-shaped eval suite for LLM apps — metrics like answer relevancy, faithfulness, and hallucination run as unit tests. Plumbline captures DeepEval runs as signed records so a passing or failing test in CI lands on the ledger alongside the metric output your team already trusts.

RAGAS scores retrieval-augmented generation on faithfulness, context precision, and answer relevance. Capture each RAGAS run as a signed Plumbline record so the same context-window + retriever-config your RAGAS report cites becomes a replayable artifact on the ledger.

OpenAI Evals runs JSON-defined tasks against an OpenAI-compatible completion endpoint. Register a Plumbline final_summarize step in your eval template and every Eval invocation in CI pushes a signed (input, output, score) record to the ledger alongside the oaieval summary output.

Braintrust's Eval() / braintrust SDK runs end-to-end eval loops with built-in scoring. Wrap your scorer so each braintrust run forwards (input, output, score) to Plumbline's /api/eval-records — same wire you can already see on /docs/sdk#Braintrust — so the run shows up on /dashboard/runs and your Braintrust project simultaneously.

Same Braintrust eval loop on the Node side — braintrust.eval() with custom scorers. Forward each scorer's (input, output, score) tuple to Plumbline's /api/eval-records so the same TS run shows up on /dashboard/runs and your Braintrust project simultaneously.

The Vercel AI SDK's experimental eval harness scores streams from generateText / streamText against test data. Capture each scored run as a signed Plumbline record from a Node init file so the stream token counts, model ids, and scorer outputs are all chained into the ledger.

No SDK needed for a custom in-house scorer — POST the same EvalRecordCreate envelope to /api/eval-records with your team's tooling of choice. The wire is what every other tile in this directory already posts; the framework integration is whatever wraps it.