transcript-processinglisted

# Transcript Processing (v3) ## Transcript sources the user has four active transcript sources. The plugin treats them uniformly. | Source | Drive folder name | Folder ID | Format | Notes | | --- | --- | --- | --- | --- | | Plaud.ai (current) | `_meetings` | `1myldke5cEDWhBqssE0ZvZXRZiaZ2F02v` | Markdown files named `YYYY-MM-DD — Title.md` | Primary source as of ~Feb 2026 | | Plaud.ai (summaries) | `Meetings` | `15K9x_-sK_DPZ1Ch_gd3ItfNm8zvBhVpa` | Contains `Daily Summaries` and `2026/` subfolders | Secondary Plaud outputs | | Spellar AI | `Spellar Meetings` | `1Q1-tmJnRe_WkxrPsQVpHM-jDGncrbh6b` | Google Docs grouped by date-bracketed subfolders | Legacy; use until cutover date | | Zoom Cloud | `YYYY-MM-DD HH.mm.ss the user's Zoom Meeting` | varies per meeting | VTT + M4A + summary | Only process if a transcript file is present | The inventory doc at `docs/transcript-inventory.md` keeps the current folder IDs. ## Processing flow ``` 1. Discover → list new files in each source since last run watermark [USES: Zapier MCP] 2. Fetch → export/download to local cache [USES: Direct OAuth MCP] 3. Parse → detect format (Plaud md / Spellar doc / Zoom vtt) 4. Classify → meeting type (1:1, team, customer, community, internal review) 5. Route → Obsidian path by meeting type 6. Write → structured note (see template) 7. Extract → action items, decisions, risks, entities 8. Stage → Things for tasks, wiki inbox for entities, draft for Jira 9. Record → update watermark +