refactor(supermemory): session-level ingest + kebab aliases (salvaged from #32487)

[alt-glitch]

Salvages #32487 (by @MaheshtheDev), rebased onto current main and brought to green CI. Closes #32487.

What's kept (the genuine refactor)

• sync_turn now buffers cleaned turns instead of firing a per-turn add_memory thread. The full session is ingested once at session end / switch / shutdown via the conversations endpoint.
• ingest_conversation() accepts and forwards functional document metadata (type, session_id, message_count, partial) for retrieval/profile building.
• Kebab-case tool aliases (supermemory-save / -search / -forget / -profile) registered alongside the snake_case names.
• README + docs (EN / zh-Hans) updated for the simplified session model.

What's removed vs the original PR — no telemetry

Per project policy (we don't tag Hermes traffic for any vendor's analytics), all source/vendor attribution was stripped:

• ❌ default_headers={"x-sm-source": "hermes"} on the SDK client
• ❌ x-sm-source: hermes header on the raw conversations POST
• ❌ sm_source: hermes metadata injection in _merge_metadata
• ❌ sm_capture_mode analytics tags (session / tool)

Note: retaindb (x-sdk-runtime: hermes-plugin) and openviking (X-OpenViking-Agent: hermes) ship the same class of header on main. Left for a separate cleanup so this PR stays focused.

Fixes that made it salvageable

• CI check-attribution: added maheshthedev@gmail.com → MaheshtheDev to scripts/release.py AUTHOR_MAP.
• CI test (4): updated tests/plugins/memory/test_supermemory_provider.py for the buffer-and-flush model (FakeClient.ingest_conversation now accepts metadata; rewrote the sync/shutdown tests). 30/30 pass locally.
• Stale-base regression: the original PR branched before the atomic_json_write(mode=0o600) security hardening for supermemory.json and silently reverted it back to plain write_text. Restored the 0o600 atomic write.
• Removed a latent NameError (session_id referenced with no such param in on_session_end).

Test plan

• pytest tests/plugins/memory/test_supermemory_provider.py → 30 passed
• ruff check on changed files → clean
• check-attribution logic reproduced locally → PASS

Co-authored-by: @MaheshtheDev

[github-actions]

🔎 Lint report: salvage/supermemory-refactor-no-telemetry vs origin/main

ruff

Total: 0 on HEAD, 0 on base (➖ 0)

🆕 New issues: none

✅ Fixed issues: none

Unchanged: 0 pre-existing issues carried over.

ty (type checker)

Total: 9781 on HEAD, 9782 on base (✅ -1)

🆕 New issues: none

✅ Fixed issues: none

Unchanged: 5074 pre-existing issues carried over.

Diagnostics are surfaced as warnings — this check never fails the build.

[Dhravya]

Hey! Can you please suggest a good way to handle this? Right now, the intention for sm_source is not for telemetry, Supermemory enabled cross-agent context as well, and users can navigate those things through our UI on the app - as in the image

[alt-glitch]

Thanks @Dhravya, that's the missing context — the Spaces UI makes it clear sm_source is a functional routing key (so users can filter/bulk-manage Hermes memories per source agent), not telemetry. I'd mis-read it as analytics. Reversed it:

• ✅ Restored x-sm-source: hermes on both the SDK client (default_headers) and the raw conversations POST
• ✅ Restored metadata.sm_source: hermes on document writes
• 📝 README now documents it explicitly as Spaces routing (not telemetry), and there's a unit test pinning the behavior so it can't get silently dropped again

One thing I left out pending your confirmation: sm_capture_mode (session/tool). From the screenshot, Spaces look routed purely by sm_source — so sm_capture_mode reads as analytics-only and I kept it dropped. If it actually drives something user-facing (a filter/grouping in the app), say the word and I'll add it straight back.

Everything else in the salvage (session-level ingest refactor, kebab tool aliases, the restored 0o600 atomic config write, test updates) is unchanged and CI is green.

[MaheshtheDev]

@alt-glitch , Thanks for restoring sm_source. we are okay to leave sm_capture_mode

[alt-glitch]

Perfect, thanks @MaheshtheDev — sm_capture_mode stays out then, which is exactly the current state of the branch. Nothing further to change on that front.

So to recap where this landed:

• ✅ x-sm-source: hermes + metadata.sm_source: hermes restored (Spaces routing)
• ✅ sm_capture_mode left dropped (your call)
• ✅ session-level ingest refactor + kebab tool aliases
• ✅ restored the 0o600 atomic config write the original branch had reverted
• ✅ provider tests updated + CI green

Appreciate you both jumping in to clarify the intent. Should be good to review/merge whenever.