@onenomad/cortex-adapter-loom

Loom recordings + transcripts source adapter.

Note on Loom API access. Loom's workspace/REST API requires a Business or Enterprise plan, and the exact endpoint shapes are evolving. This adapter is structured so its HTTP client is the only surface that depends on Loom's specifics — swap it when Loom's API changes without touching the adapter/transform layer.

Config in config/cortex.yaml:

adapters:
  loom:
    package: "@onenomad/cortex-adapter-loom"
    enabled: true
    schedule: "*/15 * * * *"
    config:
      workspace: "yourcompany"             # Loom workspace slug
      folders: []                          # folder ids to scope, empty = all
      pageSize: 50
      maxRecordingsPerRun: 0               # 0 = unlimited
      folderToProject:                     # rule-based classifier
        fld_abc: engineering
      defaultProject: ""
      skipWithoutTranscript: true          # ignore recordings w/o transcript

Required secrets (.env):

• LOOM_API_KEY — workspace API key from Loom admin

Feeds into @onenomad/cortex-pipeline-meeting (the 3-pass extraction). Each recording's transcript becomes a meeting-shaped ClassifiedItem whose run through the pipeline produces a brief plus per-decision and per-action-item memories.