[WIP] Add M365 integration support to Project Memory MCP server

[Copilot]

• Add M365Ref, SyncSource, SyncState models to src/models.py; update RefsIndexEntry with m365_refs
• Add SYNC_YAML_TEMPLATE to src/templates.py
• Extend src/filesystem.py: M365 ref parser, _sync.yaml protection, sync state methods (get_sync_state, update_sync_state, add_sync_source, list_projects_due_for_sync, list_projects_due_for_synthesis, resolve_m365_ref), extend get_refs_for
• Register 6 new tools + memory://m365 resource in src/server.py; update get_project_context deep mode
• Add generate_m365_guide() to src/guide.py
• Add tests: TestSyncState, TestM365RefParser, TestRefsIndexM365, TestResolveM365Ref, TestListDueForSync, TestM365GuideResource

Original prompt

Build Prompt — M365 Integration Layer (Phase 4a + 4b)

What you are adding

Extend the existing Project Memory MCP server with M365 integration support.
This covers three things:

1. _sync.yaml — new per-project file for M365 source registration and sync watermarks
2. Four new sync state tools on the memory MCP server
3. resolve_m365_ref tool for resolving M365 references from project files
4. Extended _refs-index.json parsing to index M365 reference tokens
5. Updated reference parser to handle the new [tm:], [ol:], [sp:] syntax

Critical design constraint — M365 optional

The memory MCP server must work perfectly without any M365 connection.
M365 is an optional enhancement layer, not a dependency.

Rules:

• No tool may fail or error because M365 is unavailable
• No tool may import or depend on any M365 library
• Tools that would benefit from M365 data return what they have locally
  and include a m365_available: false flag in the result when M365
  is not reachable — they never block or raise
• _sync.yaml files are valid and useful even with no M365 configured
• A project with no _sync.yaml is fully functional — treat missing
  _sync.yaml as "no M365 sources configured", not as an error

---

Part 1 — _sync.yaml schema

Location

projects/{slug}/_sync.yaml

Full schema

# Auto-managed by sync pipeline. Do not edit manually.
# Missing file = no M365 sources configured for this project. Not an error.

last_sync: null                    # ISO 8601 UTC datetime or null if never synced

sources:
  teams:
    - id: "teams-general"          # stable kebab-case slug, used in [tm:id] refs
      label: "General channel"     # human-readable, used in summaries
      channel_id: "19:abc123..."   # M365 Teams channel ID
      last_processed_at: null      # ISO 8601 UTC datetime or null
      last_message_id: null        # Teams epoch timestamp string or null (watermark)
      unprocessed_count: 0         # estimated unprocessed messages since watermark
      enabled: true                # false = skip this source in pipeline runs

  outlook:
    - id: "outlook-internal"
      label: "Internal thread"
      folder_id: "AAMkAGI2..."     # Outlook folder ID
      last_processed_at: null
      last_message_id: null        # Outlook message ID string (watermark)
      unprocessed_count: 0
      enabled: true

  sharepoint:
    - id: "sp-contracts"
      label: "Contracts library"
      site_url: "https://company.sharepoint.com/sites/acme"
      library: "Contracts"         # SharePoint document library name
      last_processed_at: null
      last_modified_etag: null     # SharePoint ETag for change detection (watermark)
      unprocessed_count: 0
      enabled: true

pipeline:
  correspondence_frequency: daily    # daily | weekly | manual
  knowledge_frequency: weekly        # daily | weekly | manual
  last_knowledge_synthesis: null     # YYYY-MM-DD or null
  next_knowledge_synthesis: null     # YYYY-MM-DD or null (computed on update)
Rules
_sync.yaml is auto-managed. Add it to the is_manifest() family — block direct writes via write_file. Only writable via sync state tools.
Missing _sync.yaml is not an error anywhere in the server. All sync state tools return {"result": null, "warnings": ["No _sync.yaml found"]}.
id fields within each source list must be unique within their type (two teams sources cannot share an id, but teams and outlook can).
id values are kebab-case. Enforce on add_sync_source.
enabled: false sources are preserved in the file but skipped by list_projects_due_for_sync.
Part 2 — Reference syntax extension
New token types
Extend the existing reference parser in src/filesystem.py to recognise three new token types in Markdown file bodies:

Code

[tm:source-id]                   Teams channel reference
[tm:source-id/message-id]        Specific Teams message
[ol:source-id]                   Outlook folder/thread reference
[ol:source-id/message-id]        Specific email message
[sp:source-id]                   SharePoint library root
[sp:source-id/path/to/file.pdf]  Specific SharePoint file
Parsing rules
Square bracket syntax [type:...] — distinct from existing [[link]] syntax
source-id is the id field from _sync.yaml sources, not a raw M365 ID
Everything after the first / in the ref is the optional path/message-id
Refs with unregistered source IDs produce a warning (same pattern as unresolved @refs) — file is still written
The parser must not choke on malformed refs — skip and warn, never raise
_refs-index.json extension
Add a new m365_refs array to each file entry alongside existing refs, tags, links:

JSON

{
  "entries": {
    "projects/acme-portal/knowledge/contract-summary.md": {
      "path": "projects/acme-portal/knowledge/contract-summary.md",
      "refs": ["john-doe", "acme-corp"],
      "tags": ["legal"],
      "links": ["acme-portal"]...

</details>



<!-- START COPILOT CODING AGENT SUFFIX -->

Created from [VS Code](https://code.visualstudio.com/docs/copilot/copilot-coding-agent).

<!-- START COPILOT CODING AGENT TIPS -->
---

📱 Kick off Copilot coding agent tasks wherever you are with [GitHub Mobile](https://gh.io/cca-mobile-docs), available on iOS and Android.