Skill: speaker-diarize

Version: 1.0.0

Split transcript into per-speaker segments.

Useful for multi-speaker meetings and interviews.

name: speaker-diarize version: 1.0.0 description: Split transcript into per-speaker segments with speaker identification

Credentials

credentials:

• name: OPENAI_API_KEY required: false description: OpenAI API key

Skill inputs

inputs:

• name: transcript type: string required: true description: Raw transcript text
• name: speaker_hints type: array required: false description: Optional list of known speaker names

Skill outputs

outputs:

• name: segments type: array description: "Array of {speaker, text, start_time?, end_time?}"
• name: speakers type: array description: List of identified speakers
• name: metadata type: object description: Processing metadata

Backend options

backends:

• name: openai type: remote requires: [OPENAI_API_KEY]
• name: ollama type: local models: [llama3.2, mistral]

Expanso components used

components: inputs: - stdin - http_server processors: - mapping - openai_chat_completion - log outputs: - stdout - sync_response