Add Pipecat Evals framework documentation#889
Open
aconchillo wants to merge 18 commits into
Open
Conversation
Document the new built-in evals framework with a top-level Evals group in the Pipecat tab: - Overview: why evals matter, eval transport + harness + judge architecture, text vs audio modes, requirements - Quickstart: run an existing agent with -t eval and a first scenario - Writing Scenarios: full YAML reference (turns, expectations, judge, audio mode, includes, interruptions, vision) - Eval Suites: manifests, pipecat eval suite, run output, CI - Using the Library: EvalScenario/EvalSession/EvalSuite Python API - Agent Self-Improvement: closing the loop with AI coding assistants Also add a CLI reference page for pipecat eval run/suite, and update the Fundamentals evaluations overview to feature the built-in framework.
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Workflows to automatically generate PRs for you. |
markbackman
approved these changes
Jun 11, 2026
markbackman
left a comment
markbackman
left a comment
Contributor
There was a problem hiding this comment.
LGTM! Just a few small changes. Nice docs 👏
| A scenario is a YAML file describing a scripted conversation and the behavior you expect. Save this as `scenarios/capital_question.yaml`: | ||
|
|
||
| <Tabs> | ||
| <Tab title="Ollama judge (default)"> |
Contributor
There was a problem hiding this comment.
When using the default, is ollama, gemma2:9b the default? Do you not need to specify it?
Contributor
Author
There was a problem hiding this comment.
Yes ollama with gemma2:9b is the default. I'll add a comment
markbackman
reviewed
Jun 11, 2026
Drop the tail page and its nav entry, repoint the old /cli/tail redirect to the CLI overview, and replace the overview's tail references with the eval command (which was missing from that page).
…ith text and audio
Move the third-party platform pages (Bluejay, Cekura, Coval) under a Third-party Platforms subgroup in the Evals tab group, absorb the old evaluations overview's production-evaluation content into the Evals overview, and add redirects for the old URLs.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Documents the new built-in Pipecat Evals framework with a top-level Evals group in the Pipecat tab:
pipecat/evals/overview): why evals matter, the eval transport + harness + judge architecture, text vs audio modes, and requirements.pipecat/evals/quickstart): start an existing agent with-t eval, write a first scenario, and runpipecat eval run.pipecat/evals/scenarios): the full YAML reference: turns, expectations (eval,text_contains,within_ms, function calls), interruptions withsend_after, audio mode (user:/judge:blocks),!include,reset:, and vision turns.pipecat/evals/suites): manifests,pipecat eval suite, run output layout, and CI usage.pipecat/evals/library): thepipecat.evalsPython API (EvalScenario,EvalSession,EvalResult,EvalManifest,EvalSuite), building scenarios in code, and injecting a custom judge/speech/transcriber.pipecat/evals/agent-self-improvement): closing the loop with AI coding assistants (edit, run evals, read results, iterate).Also:
pipecat eval run/pipecat eval suite(api-reference/cli/eval).TranscriptionFramelocal-testing workaround) while keeping the third-party platform content.pipecat tailCLI reference page (the command is going away): nav entry dropped, the old/cli/tailredirect repointed to the CLI overview, and the CLI overview now referencespipecat evalinstead.All content verified against the framework source in
pipecat-ai/pipecat(src/pipecat/evals/, runner integration, and the release-evals scenarios).mint broken-linkspasses.