Podcast Ingestion — n8n Workflow Flow

Podcast Ingestion — n8n Workflow Flow A workflow diagram generated by Archify. EX / Transcripts Already Present 02 / Discover & Register 03 / Transcribe & Vectorize Scan & check Register & transcribe Vectorize & persist Early Exit · nothing to transcribe · Transcripts Already Present › Register & transcribe Early Exit nothing to transcribe GitHub List Files · scan repo for .mp3 · Discover & Register › Scan & check · GitHub GitHub List Files scan repo for .mp3 Function Node · check .txt / .srt missing · Discover & Register › Scan & check Function Node check .txt / .srt missing Airtable Register · lookup / create episode · Discover & Register › Register & transcribe · Airtable Airtable Register lookup / create episode Airtable Update · mark Transcribed + Vectorized · Discover & Register › Vectorize & persist · Airtable Airtable Update mark Transcribed + Vectorized Whisper ASR · HTTP nodes → TXT + SRT · Transcribe & Vectorize › Register & transcribe Whisper ASR HTTP nodes → TXT + SRT GitHub Commit · push transcripts · Transcribe & Vectorize › Register & transcribe · GitHub GitHub Commit push transcripts Ollama Embeddings · Llama3.1:8b · Transcribe & Vectorize › Vectorize & persist · Ollama Ollama Embeddings Llama3.1:8b Supabase Insert · transcript + metadata + vectors · Transcribe & Vectorize › Vectorize & persist · Supabase Supabase Insert transcript + metadata + vectors transcripts exist missing queued Legend Agent logic Context / trace Cloud service External system

Ingestion Guarantees

  • • Only episodes missing .txt / .srt transcripts run Whisper ASR
  • • Transcripts are committed back to the repo before vectorization

Vector Store Path

  • • Ollama Llama3.1:8b embeddings persist in Supabase
  • • Airtable records the final Transcribed + Vectorized state