memory-gdocs-ingest

name	memory-gdocs-ingest
description	Ingest recently modified Google Docs into the memory knowledge base with email correlation
user_invocable	true

Google Docs Ingest for Memory Knowledge Base

Pull recently modified Google Docs, correlate with related emails, and append enriched entries to the daily log.

Process

1. Load State

Read state from /Users/jesseanglen/Documents/RandomStuff/memory-kb/state/gdocs-state.json. If it doesn't exist, this is the first run — set last_checked to 7 days ago.

2. Search for Recently Modified Docs

Use search_files (Google Drive MCP) with:

query: "modifiedTime > 'YYYY-MM-DDTHH:MM:SSZ' and mimeType = 'application/vnd.google-apps.document'"
pageSize: 20

Important: Timestamps MUST use UTC with Z suffix: 2026-04-10T00:00:00Z

Alternatively, use list_recent_files with orderBy: "lastModified" and filter to documents.

3. Process Each Document

The search tools return content snippets directly. For full content, use read_file_content with the document's id.

For each document:

1. Check if title + modifiedTime differ from gdocs-state.json — skip if unchanged
2. Extract: title, doc ID, last modified, content summary
3. Identify people and companies mentioned in the document (names, company names, email addresses)

4. EMAIL CORRELATION (Critical Enhancement)

For each document that mentions people or companies, search Gmail for related email threads:

Use gmail_search_messages with queries derived from the document content:

• Search by company name: e.g., "spyder construction" or "monetate"
• Search by person name: e.g., "david gallo" or "steve maher"
• Search by email addresses found in the doc

For each matching email thread:

1. Use gmail_read_message to get the most recent message in the thread
2. Extract: subject, participants, key status/decisions, last activity date

This creates a unified view — the document plus all related email communications.

5. Append Enriched Entry to Daily Log

For each doc, append to /Users/jesseanglen/Documents/RandomStuff/memory-kb/daily/YYYY-MM-DD.md:

## Google Doc: [Document Title] @ HH:MM PT
**source:** gdocs
**doc_id:** [document ID]
**last_modified:** [ISO timestamp]
**people:** [Names and emails found in doc]

### Document Summary
- [Key content — decisions, plans, terms, important details]

### Related Email Threads
- **[Subject line]** (last activity: [date]) — [1-line summary of thread status]
  - Participants: [list]
  - Key: [latest decision or action item from thread]
- **[Subject line]** (last activity: [date]) — [1-line summary]
  - Participants: [list]
  - Key: [latest status]

### Action Items
- [ ] [Any outstanding items from doc + emails combined]

Create the daily log with # Daily Log: YYYY-MM-DD header if it doesn't exist.

6. Update State

Write to /Users/jesseanglen/Documents/RandomStuff/memory-kb/state/gdocs-state.json:

{
  "last_checked": "ISO timestamp with Z suffix",
  "total_runs": N,
  "last_run": {
    "timestamp": "ISO",
    "docs_found": N,
    "docs_ingested": N,
    "docs_unchanged": N,
    "email_threads_correlated": N,
    "status": "success"
  },
  "doc_hashes": {
    "doc_id_1": "title:modifiedTime",
    "doc_id_2": "title:modifiedTime"
  }
}

7. Report

"Google Docs ingest complete. Found X modified docs, ingested Y with changes, correlated Z email threads."

MCP Tools Reference

Google Drive:

• search_files — search by query with date filters and mime type
• list_recent_files — list most recently modified files
• read_file_content — read full document content by fileId
• get_file_metadata — get metadata for a specific file

Gmail (for correlation):

• gmail_search_messages — search for emails related to doc content
• gmail_read_message — read full email for thread context

Correlation Strategy

The goal is that when Jesse asks about a deal or document, the compiled knowledge article has BOTH the document content AND the email thread context — so the memory surfaces everything relevant in one place.

What to correlate:

• Proposals/MSAs → find email threads with the client (by company name or contact name)
• Engagement letters → find the originating conversation and any signed/executed versions
• Meeting notes → find follow-up emails from attendees
• Any doc with a person's name → search for emails with that person

What NOT to correlate:

• Internal HR/salary docs → don't search for related emails (privacy)
• Template docs with no specific client → skip correlation
• Docs under 50 words (stubs) → skip entirely

Notes

• Zero Chrome dependency — pure MCP (Google Drive + Gmail)
• Email correlation is what makes this ingest powerful — it creates unified deal context
• Limit to 3 email threads per document to keep daily log entries manageable
• For large docs (10,000+ words), summarize key sections only