Add `diarize` pipeline node

[natashaannn]

Add diarize pipeline node

Objective

Migrate speaker diarization into the DAG as a first-class pipeline node.

---

Why this matters

After sync and transcribe, diarization is the next critical stage for realistic transcript pipeline execution.

---

Parallelization metadata

Track

pipeline

Depends on

• Add transcribe pipeline node #36

Safe to run in parallel with

• type contract issues
• remotion utility issues

Merge risk

• medium

---

Files in scope

Primary:

• scripts/pipeline/nodes/diarize.ts

Avoid touching:

• Python diarization internals

---

Required implementation

1. Implement node wrapper

Expose diarization as a PipelineNode.

2. Define inputs/outputs

Use content-addressed artifact references.

3. Delegate to current implementation

No ML logic rewrite.

---

Constraints

• wrapper only
• no pyannote changes

---

Handoff contract

Runner should support:

sync -> transcribe -> diarize

---

Acceptance criteria

Functional

• diarize executes through the runner

Integration checks

• expected diarization artifact is produced

---

Verification commands

Run a three-stage pipeline.

Expected result:

• diarize executes after transcribe

---

Explicitly out of scope

• align node
• assign-speakers node

---

Suggested branch

refactor/s6-diarize-node

Suggested commit slug

phase-2-step-2-diarize-node