Guide for bulk ingestion and batch processing of Plaud recordings into the memory search index. Use when the user wants to index multiple recordings, set up auto-ingestion, check job progress, or do batch operations.
Batch ingestion of Plaud recordings into the memory search index (Cognee with Qdrant vectors + Neo4j graph). Handles transcription, chunking, embedding, and graph construction at scale.
dry_run=true first for any bulk operationsource="cache" (free) over source="all" (costs ElevenLabs credits)since to limit scope when possiblejob_id returned from initial callmemory_ingestMain bulk ingestion tool. Transcribes, chunks, embeds, and indexes recordings into the memory search pipeline.
| Parameter | Type | Default | Description |
|---|
source | "all" | "cache" | "all" | "all": transcribe missing (COSTS CREDITS). "cache": cached only (FREE) |
mode | "once" | "auto" | "once" | "once": single run. "auto": background poller |
action | string | null | Required when mode="auto": "start", "stop", or "status" |
file_ref | string | null | Ingest a single recording only |
since | string | null | Only recordings after date ("7d", "30d", "2026-01-01") |
language | string | "eng" | ElevenLabs 3-letter language code |
dry_run | bool | false | Show cost estimate without processing |
job_id | string | null | Poll status of a running job |
memory_ingest(dry_run=true) # Cost estimate for all recordings
memory_ingest(source="cache") # Index only cached transcripts (FREE)
memory_ingest(source="all") # Transcribe + index everything (COSTS CREDITS)
memory_ingest(file_ref="abc123") # Ingest a single recording
memory_ingest(since="7d") # Only recent recordings
memory_ingest(since="7d", dry_run=true) # Cost estimate for recent only
memory_ingest(job_id="abc-123") # Poll job progress
memory_ingest(mode="auto", action="start") # Start background auto-poller
memory_ingest(mode="auto", action="status") # Check auto-poller status
memory_ingest(mode="auto", action="stop") # Stop auto-poller
memory_ingest(dry_run=true) -- review cost estimatememory_ingest(source="all")job_id: memory_ingest(job_id="<id>")memory_ingest(source="cache") -- no ElevenLabs costjob_idmemory_ingest(since="7d", dry_run=true) -- estimatememory_ingest(since="7d") -- executememory_ingest(file_ref="<id>") -- transcribes + indexes one recordingmemory_ingest(mode="auto", action="start")memory_ingest(mode="auto", action="status")memory_ingest(mode="auto", action="stop")memory_ingest(job_id="<id>") -- returns progress, count, errorssource="all" | source="cache" | |
|---|---|---|
| Cost | ElevenLabs API credits per uncached recording | Free |
| Coverage | All recordings (transcribes missing ones) | Only already-transcribed recordings |
| Speed | Slower (transcription per recording) | Fast (skip transcription step) |
| Use when | User wants everything indexed | User wants to avoid cost |
memory_ingest(source="all") without dry_run=true first -- always estimate costjob_id -- always poll for completion to confirm successjob_id status and verify with memory_search