Add end-to-end pipeline validation command
Objective
Introduce a full pipeline validation command that runs all stages and verifies deterministic output.
Why this matters
This is the first step toward production-grade reproducibility guarantees.
Parallelization metadata
Track
pipeline
Depends on
Safe to run in parallel with
Merge risk
Files in scope
Primary:
scripts/pipeline/validate.ts
May touch:
Required implementation
1. Add validation runner
Execute full pipeline:
- sync
- transcribe
- diarize
- align
- assign
- merge
2. Determinism check
Compare artifact hashes across runs.
3. Report diffs
Highlight stage-level differences.
Constraints
- read-only validation
- no pipeline mutation
Acceptance criteria
Functional
Determinism
Verification commands
Run validation twice.
Expected result:
- identical artifact hashes
Explicitly out of scope
- fixing nondeterminism issues
- caching improvements
Suggested branch
refactor/s7-pipeline-validation
Suggested commit slug
phase-2-step-8-validation
Add end-to-end pipeline validation command
Objective
Introduce a full pipeline validation command that runs all stages and verifies deterministic output.
Why this matters
This is the first step toward production-grade reproducibility guarantees.
Parallelization metadata
Track
pipelineDepends on
assign-speakerspipeline node #46merge-docpipeline node #47Safe to run in parallel with
Merge risk
Files in scope
Primary:
scripts/pipeline/validate.tsMay touch:
runner.tsRequired implementation
1. Add validation runner
Execute full pipeline:
2. Determinism check
Compare artifact hashes across runs.
3. Report diffs
Highlight stage-level differences.
Constraints
Acceptance criteria
Functional
Determinism
Verification commands
Run validation twice.
Expected result:
Explicitly out of scope
Suggested branch
refactor/s7-pipeline-validationSuggested commit slug
phase-2-step-8-validation