Desktop: profile switches and Bot opens no longer spawn one backend per profile (#102498, salvage #102512)

[teknium1]

Switching profiles or opening a stale Bot tile no longer starts one local backend per configured profile: read-only GET /api/sessions* requests are served by the already-running primary backend with ?profile=, so the pool slots stay free for bots that are actually running (#102498, salvage of #102512 by @fangliquanflq).

Closes #102498. Supersedes #102512 (cherry-picked to preserve authorship).

What changes

• apps/desktop/electron/connection-config.ts::localPrimaryRequestScope — GET /api/sessions and every subpath (/{id}, /{id}/messages, /{id}/latest-descendant, /search) is a profile-scoped local-primary route. Writes (PATCH/DELETE/POST) stay on the per-profile pool so their process-level side effects are unchanged.
• website/docs/user-guide/bot-mode.md — new Warm Bot Backends section: what the setting bounds, defaults (3 backends, 10 min idle, 30 s slot wait), that the post-reap exited (1) line is cleanup not a crash, and which operations take a slot. Requested in the Discord fleet thread.

Root cause

resolveStoredSession (use-session-actions/utils.ts) probes a stored session id across every profile when the owner is unknown, one GET /api/sessions/{id} per profile. The routing table classified the sessions family as pool-only, so each probe went through ensureBackend(profile) → spawnPoolBackend. With 9+ profiles one click saturated the pool and every later open hit timed out waiting for a free local slot. The Python handlers (hermes_cli/web_routers/sessions.py::get_session_detail / get_session_messages / get_sessions) already accept profile and open that profile's state.db read-only, exactly like the sidebar aggregate the primary already serves.

Complements #107969 (dc90a75), which made background tile reconciles passive: that stopped reconciles spawning, this stops ownership probes spawning. The registry dispatcher (dispatchRegistryApiRequest) is untouched: an explicit connectionId pin means a registry-owned backend, and its forced-local path already reuses ensureBackend for the delegating route.

Validation

Check	Result
vitest --project electron electron/connection-config.test.ts	120/120 (2 route-table cases + 1 resolveProfileApiRequest case from the salvaged commit)
tsc -p tsconfig.electron.json --noEmit	clean
eslint on touched files	0 errors
Cherry-pick onto origin/main	clean, fangliquanflq <fangliquan@qq.com> authorship preserved
Live multi-profile repro	not run here (no 9-profile Windows install); mechanism traced line-level against the reporter's desktop.log pattern

Related

• fix(desktop): profile-switch slot-cap regression (slot-cap half of #102163) #102463 — displaces the stalest running backend to free a slot; closed, it inverts the pool's "never kill a live session for the cap" contract and would surface as failed bot turns for exactly this fleet.
• fix(desktop): make backend pool slot timeouts actionable #105264 (actionable slot timeouts), fix(desktop): raise default warm backend pool cap from 3 to 8 #102828 (default cap), fix: [Bug]: Bot relay timers spawn-storm profile backends against a full local pool — 30s retry loop with no backoff starves the slot pool and degrades the whole UI (typing focus loss) (#102913) #103525 (relay timer storms) — open, adjacent, not folded in.

Infographic

[github-actions]

૮ >ﻌ< ა ci review

_{ran on 20dfd8c — docs(bot-mode): explain Warm Bot Backends, idle reaping, and}

⚠️ Warnings

OSV vulnerability scan · View job

80 known vulnerabilities found in pinned dependencies.

• CVE-2026-83608 in package-lock.json
• CVE-2026-83608 in package-lock.json
• CVE-2026-84375 in package-lock.json
• CVE-2026-84375 in website/package-lock.json
• CVE-2026-67213 in website/package-lock.json
• CVE-2026-85062 in package-lock.json
• CVE-2026-85062 in website/package-lock.json
• CVE-2026-83609 in package-lock.json
• CVE-2026-82417 in scripts/whatsapp-bridge/package-lock.json
• CVE-2026-82417 in website/package-lock.json
• CVE-2026-84369 in website/package-lock.json
• CVE-2026-83605 in package-lock.json
• CVE-2026-83605 in package-lock.json
• CVE-2026-75931 in package-lock.json
• CVE-2026-75931 in website/package-lock.json
• CVE-2026-83610 in package-lock.json
• CVE-2026-83610 in package-lock.json
• CVE-2026-83611 in package-lock.json
• CVE-2026-83611 in package-lock.json
• CVE-2026-71554 in uv.lock

How to fix:

Review the findings in the Security tab. Update the affected dependencies if a patched version is available.

---

debug info

CI timings

CI timings · View report · View job

Wall time 5m28s vs 6m9s (-11.1%). 9 job(s) slower, 6 faster, 1 unchanged.

• JS & TS checks / JS & TS checks: -55.0s
• Check no case-colliding filenames / check-case-collisions: +42.0s
• Python tests / Run tests: -40.0s
• OS-specific tests / Windows-only tests: -38.0s
• Docs Site / docs-site-checks: +37.0s