Add `assign-speakers` pipeline node

[natashaannn]

Add assign-speakers pipeline node

Objective

Assign speaker labels to diarized + aligned transcript segments.

---

Why this matters

This is the first stage where the pipeline begins producing human-usable structured transcript data, not just ML outputs.

---

Parallelization metadata

Track

pipeline

Depends on

• Add diarize pipeline node #40
• Add align pipeline node #41

Safe to run in parallel with

• merge-doc node
• pipeline validation

Merge risk

• medium

---

Files in scope

Primary:

• scripts/pipeline/nodes/assign-speakers.ts

Avoid touching:

• diarization model logic

---

Required implementation

1. Implement node wrapper

Expose as PipelineNode.

2. Input contract

Must consume:

• diarized segments
• aligned transcript

3. Output contract

Produces:

• speaker-assigned transcript JSON

---

Constraints

• assignment logic reuse only
• no ML changes

---

Acceptance criteria

Functional

• speaker labels assigned correctly

Integration checks

• output artifact exists in DAG

---

Verification commands

Run pipeline through assign-speakers.

Expected result:

• transcript contains speaker labels

---

Explicitly out of scope

• merge-doc
• final transcript formatting

---

Suggested branch

refactor/s7-assign-speakers

Suggested commit slug

phase-2-step-6-assign-speakers