[sanil-23]

Summary

• The Memory sources list showed multiple rows per provider (e.g. several "Gmail") even when only one connection was active.
• Filters the sources listing to the live active-connection set and dedupes by connection_id, reusing the Composio scan list_rpc already runs every 5s poll.
• Display-layer only, non-destructive: no row, setting, or ingested memory is removed; an inactive connection's row reappears on its own once it re-activates. Explicit Disconnect remains the only removal path.
• On a failed/unavailable scan the list passes through untouched — a transient outage never hides every source.

Problem

Composio memory sources are keyed by connection_id. Two independent mechanisms produced duplicate rows:

• RC-A (primary): the registry keeps one row per connection and the list returned every row status-blind. Re-auth / token expiry / reconnect mints a new connection_id, so the old (now-inactive) connection's row lingers — "multiple Gmail rows, only one active." The reconciler only ever upserts; nothing prunes a connection that dropped out of the active set.
• RC-B (secondary): list_rpc reconciles on every poll and upsert_composio_source is a non-atomic read-check-write, so two overlapping polls can double-insert the same connection_id.

Solution

• reconcile::ensure_composio_sources now returns the active connection_id set it scans (Some(set) on success, None when the scan couldn't run — config / network / auth failure).
• list_rpc feeds that set to a new pure helper filter_to_active_composio_sources:
  • Composio sources kept only if their connection_id is in the active set (hides RC-A stragglers), first occurrence per id (collapses RC-B duplicates).
  • Non-Composio sources (folder / git / …) always shown.
  • None ⇒ pass-through — never hide everything on a transient scan failure.
• No change to removal semantics or to time-to-appear for new connections: the live scan still runs inline in list_rpc, so a just-connected integration still shows within one poll. The filter only subtracts inactive/duplicate rows — no added network cost.

Submission Checklist

• Tests added or updated (happy path + failure / edge case) — 6 unit tests over the pure filter + 2 JSON-RPC E2E tests (active-filter path and scan-unavailable fallback) driving a mock Composio /connected_accounts endpoint.
• Diff coverage ≥ 80% — every changed production line is exercised except one (reconcile.rs return None on a config-load failure, impractical to unit-test); ~16/17 measured changed lines ≈ 94%.
• Coverage matrix updated — N/A: bug fix to existing memory-sources list behaviour; no feature row added/removed/renamed.
• All affected feature IDs listed under ## Related — N/A: behaviour-only fix.
• No new external network dependencies — mock Composio endpoint used in-test per mock policy.
• Manual smoke checklist — N/A: does not touch a release-cut surface.
• N/A: Linked issue closed — no tracking issue filed for this report.

Impact

• Desktop (Memory → Sources registry). The only UI affected is MemorySourcesRegistry.tsx, driven solely by memory_sources_list; the status_list poll only supplies per-row stats and never adds rows. Other Composio surfaces (integration grids, Add Source picker) read composio_list_connections and are intentionally out of scope.
• No migration, no schema change, no security impact. Reversible: worst case a stale row shows briefly until the next good scan.

Related

• Closes: N/A (no tracking issue filed)
• Follow-up PR(s)/TODOs: eliminating the non-atomic upsert_composio_source write race at the source (RC-B residue) is deliberately out of scope — the dedupe hides it; growth is slow.

---

AI Authored PR Metadata

Linear Issue

• Key: N/A
• URL: N/A

Commit & Branch

• Branch: fix/memory-sources-hide-inactive-connections
• Commit SHA: 32f640de6bd09f5dfefacd9d0c78b4a7bc891c69

Validation Run

• N/A: pnpm --filter openhuman-app format:check not run — no app/ changes.
• N/A: pnpm typecheck not run — no TS changes.
• Focused tests: cargo test -p openhuman --lib memory_sources::rpc::filter_tests (6 passed); cargo test --test json_rpc_e2e json_rpc_memory_sources_list (2 passed).
• Rust fmt/check: cargo fmt --check clean; cargo check clean.
• N/A: Tauri fmt/check not run — no shell changes.

Behavior Changes

• Intended behavior change: inactive/duplicate Composio rows no longer appear in the sources list.
• User-visible effect: one row per active connection; stale rows vanish (and reappear if the connection re-activates).

Parity Contract

• Legacy behavior preserved: explicit Disconnect removal unchanged; new-connection time-to-appear unchanged (live scan stays inline).
• Guard/fallback/dispatch parity checks: scan-failure (None) falls back to showing all sources — verified by none_active_set_shows_all (unit) and json_rpc_memory_sources_list_shows_all_when_scan_unavailable (E2E).

Duplicate / Superseded PR Handling

• Duplicate PR(s): N/A
• Canonical PR: this
• Resolution: N/A

Summary by CodeRabbit

• Bug Fixes

  • Listings now show only active Composio connections and consolidate duplicates when live scan is available.
  • Inactive/stale Composio rows are hidden when scan succeeds; when the live scan is unavailable, all sources are preserved (no hiding).
  • Composio rows lacking connection IDs are retained instead of dropped.

• Tests

  • Added E2E and unit tests covering filtering, deduplication, reactivation, fallback, and orphan-row behavior.

[coderabbitai]

Looking for one thing? Review this PR in Change Stack to search files, summaries, diffs, and code without losing your place.

📝 Walkthrough

Walkthrough

Composio memory sources are now filtered based on active connection scans. ensure_composio_sources returns an Option<HashSet<String>> containing live connection IDs or None when the scan is unavailable. list_rpc uses this set to filter listings: hiding inactive Composio sources, deduplicating by active connection ID, and preserving all sources when unavailable. E2E tests verify both filtering behavior and graceful handling when the Composio endpoint is unreachable.

Changes

Composio Memory Sources Active Connection Filtering

Layer / File(s)	Summary
Active connection set return contract src/openhuman/memory_sources/reconcile.rs	ensure_composio_sources signature changes to return Option<HashSet<String>>: on successful scan it returns Some(...) containing live connection IDs; on config-load or scan failure it returns None instead of exiting, explicitly signaling the active set is unavailable to callers.
Listing-layer filtering and deduplication src/openhuman/memory_sources/rpc.rs	list_rpc calls ensure_composio_sources and filters registry results using a new pure helper: hides inactive Composio rows, collapses duplicates by active connection_id, keeps non-Composio sources, retains Composio rows missing connection_id, and preserves all rows when the active-connection set is None. Debug logging reports scan availability and filtered counts. Unit tests cover filtering semantics, deduplication, reactivation, non-Composio pass-through, unavailable-scan behavior, and orphan connection handling.
End-to-end integration tests tests/json_rpc_e2e.rs	Adds mock Composio v3 router returning ACTIVE Gmail connection, helpers to write Composio direct-mode config and add memory sources over JSON-RPC, and response normalization. Two E2E tests: (1) verifies filtering collapses active Composio duplicates into one row while hiding inactive ones and preserving folder sources; (2) verifies all sources are shown when the Composio endpoint is unreachable (scan unavailable).

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Suggested labels

rust-core, memory

Poem

🐰 I hopped through code to clear the stew,
Active links kept, stale ones withdrew.
Duplicate tails fold into one,
And when the scan sleeps, the full list runs.
A rabbit’s tidy hop—now the sources are true. 🌱

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: hiding inactive Composio connections in the sources list display layer, which matches the primary objective of the PR.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}