QVAC-18827 feat[bc|api]: add unloadModel autoClose option, default-off on Bare

[opaninakuffo]

🎯 What problem does this PR solve?

• unloadModel calls close() unconditionally when no models/providers remain, which terminates long-lived Bare worker hosts on every routine unload.
• Bare consumers expecting their process to outlive a model unload have to skip unloadModel or work around the auto-close.

📝 How does it solve it?

• Adds an optional autoClose: boolean param to unloadModel.
• Default flips by runtime via which-runtime.isBare: true on Node/Electron (preserves current behavior), false on Bare (worker survives unload).
• Explicit autoClose: true | false always wins.
• Client-only — the wire request schema (unloadModelRequestSchema) is unchanged, no server changes needed.
• API reference page updated to describe the new runtime defaults.

🧪 How was it tested?

Two-runtime integration smoke against examples/quickstart.ts (load → completion → unload):

	Bun (Node-like)	Bare
Wall time	23.4s (self-exit)	75s (gtimeout cap)
Exit code	0	124
Token stream emitted	yes	yes
Model … unloaded server log	yes	yes
🧹 No models or providers active, automatically closing RPC connection…	1	0

Last row is the verdict: auto-close branch fires on Bun, skipped on Bare. eslint, tsc --noEmit, and bun run build all clean.

💥 Breaking Changes

Bare consumers no longer have their host process closed by unloadModel. Node/Electron behavior is unchanged.

BEFORE: (Bare)

import { unloadModel } from "@qvac/sdk";

await unloadModel({ modelId });
// RPC connection closed → Bare worker host terminated. Long-lived workers
// had to avoid unloadModel or work around the auto-close.

AFTER: (Bare)

import { unloadModel } from "@qvac/sdk";

await unloadModel({ modelId });
// Connection stays open, worker survives. Opt in to closing explicitly:
await unloadModel({ modelId, autoClose: true });

🔌 API Changes

await unloadModel({ modelId });

await unloadModel({ modelId, autoClose: true });

await unloadModel({ modelId, autoClose: false });

[github-actions]

QVAC E2E — android — running

View run
Config: suite=smoke · filter=(none) · exclude=(none)

[github-actions]

QVAC E2E — windows — ✅ all tests passed (91/91, 491s)

Config: suite=smoke · filter=(none) · exclude=(none)
View run · Artifacts: reports

[github-actions]

QVAC E2E — ios — running

View run
Config: suite=smoke · filter=(none) · exclude=(none)

[github-actions]

QVAC E2E — linux — ✅ all tests passed (91/91, 492s)

Config: suite=smoke · filter=(none) · exclude=(none)
View run · Artifacts: reports

[github-actions]

QVAC E2E — macos — ✅ all tests passed (91/91, 339s)

Config: suite=smoke · filter=(none) · exclude=(none)
View run · Artifacts: reports

[opaninakuffo]

review

[github-actions]

Tier-based Approval Status

**PR Tier:** TIER1

**Current Status:** ✅ APPROVED

**Requirements:**
- 1 Team Member approval ✅ (1/1)
- 1 Team Lead OR Management approval ✅ (1/1)



---
*This comment is automatically updated when reviews change.*

[opaninakuffo]

review