Knuckles-Team Voice Message Transcriber Workflow
CONCEPT:KG-2.12
Detects new local voice notes or audio files, triggers Whisper transcriptions, and prepares structured text outputs for conversational agent consumption.
Steps
Step 0: Audio Transcriber Mcp
Agent: intake-agent
Tools: graph_query, nc_files
Transcribe an audio file or capture live microphone input using the transcribe_audio tool.
Expected: raw_transcript
Step 1: User Interaction
Agent: processor-agent
Tools: graph_analyze, document_tools
Present the raw transcript to the user. Ask if they want to translate it, format it, or route it directly to another specialist agent for text response.
Expected: structured_transcript, target_agent
Step 2: KG Persistence [depends_on: user-interaction]
Agent: processor-agent
Tools: graph_write
Persist workflow results as nodes and edges in the Knowledge Graph. Create appropriate typed nodes with metadata and link to existing domain entities.
Output
- Voice Message Transcriber results persisted in KG
- Structured report (MD/PDF)
- Audit trail with timestamps and agent attributions
Human Oversight Required
✅ Critical decisions require human review and approval.
Execution
Run this workflow as a dependency-ordered DAG. Steps with no unmet depends_on run in parallel; dependents run after their prerequisites complete.
- Run first (in parallel): Step 0 — Audio Transcriber Mcp; Step 1 — User Interaction
- After level 0: Step 2 — KG Persistence
Execution: If graph-os is reachable, offload the whole DAG via graph_orchestrate action=execute_workflow (or the kg-delegate skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet depends_on in parallel, then their dependents.