Recipe 52: Retrieval pipeline evaluation #52

Recipe 52 of 60 · Level 4 (Advanced) · Search & retrieval · Choice + Score

Use case

Combine passage filtering, reranking, and claim verification in a retrieval pipeline and measure each stage's effect on answer support using a fixed labeled corpus.

What to build

The folder recipes/52-retrieval-pipeline-evaluation/ with notebook.ipynb, README.md, and fixtures/, meeting the recipe contract. Start from python tools/new_recipe.py 52 and the template in recipes/_template/.

Build notes

• Three stages: passage filtering, reranking, and claim verification.
• Measure each stage's effect on answer support with an ablation over a fixed labeled corpus.
• Report request counts per configuration.

Decision design

• Choice. Options are a fixed set supplied by Python, including the fallback outcome this use case names. Show the choice, the per-option probabilities, and the confidence for individual examples before aggregating.
• Score. Describe every rubric level in words, in order. Show the score and the distribution over levels. Mapping, weighting, and arithmetic on scores happen in Python.
• Composition. Independent questions share one request. A question that depends on an earlier answer goes in a later request, and the notebook makes that boundary visible.

Level 4 scope

Stateful workflows, provenance, adversarial cases, model calibration, or several integrated decision stages.

Acceptance criteria

• recipes/52-retrieval-pipeline-evaluation/notebook.ipynb runs top to bottom offline, with no key and no network, and its outputs are committed
• Live mode is opt-in (JEV_COOKBOOK_LIVE=1 with TYPESAFE_API_KEY) and uses the same question definitions as the offline run
• Fixtures are synthetic, include the hard cases named in the build notes, carry gold labels, and pass the fixture validator
• Every response fixture records its provenance, and the notebook states which mode it ran in
• Evaluation: Answer-support rate for each ablation configuration.
• State, provenance, or policy rules are enforced in helpers.py and covered by tests that hold for any model answer
• Fixtures include adversarial, conflicting, or stateful cases, and the notebook walks through at least one
• No statement about Jev's quality, latency, or cost appears without recorded live inference; in a synthetic run, metrics are labelled as pipeline checks
• recipes/52-retrieval-pipeline-evaluation/README.md says what the recipe teaches, how to run it, and what was and was not measured live
• The pull request changes only recipes/52-retrieval-pipeline-evaluation/ and CI is green

Dependencies

• Blocked by Foundation: CI executes every notebook offline and validates fixtures #69 (CI executes every notebook offline and validates fixtures)
• Builds on Recipe 17: Passage reranking #17 (Passage reranking), Recipe 20: Claim support classification #20 (Claim support classification), Recipe 30: Citation verification #30 (Citation verification). Merge those first and reuse their patterns; do not copy their fixtures.
• Tracked in Epic: Level 4 · Advanced recipes #78

Sources

Read these before writing the questions.

• S02 · TypeSafe AI: Primitives (Questions): Choice, Noul, Score, independent questions, and dependent follow-up calls.
• S03 · TypeSafe AI: Confidence: Distribution-derived confidence and task-specific thresholds informed by observed outcomes.
• S04 · TypeSafe AI: How to build with TypeSafe: Narrow semantic judgments integrated with workflows, deterministic logic, and side effects controlled by code.
• S06 · TypeSafe AI: Cookbooks: Worked patterns for extraction, citation checking, skill selection, hierarchical classification, and feature discovery.

Out of scope

• Changes to shared code, the catalog, or the root README. If the shared package is missing something, open a foundation issue and link it here.
• Real side effects of any kind. Actions are simulated.