feat(pi-ai): refresh MiniMax-M3 video input and adaptive thinking

[octo-patch]

Reason: MiniMax-M3 accepts video input and supports adaptive and disabled thinking modes, but the generated catalog capped its input at text and image and left it on the budget-based thinking default.

What changed

• Add video to the Model.input modality union and to the catalog validation schema, so a model can declare video input support.
• Update the model generator so MiniMax-M3 is emitted with ["text", "image", "video"] input and with forceAdaptiveThinking enabled to match its adaptive and disabled thinking modes.
• Refresh the generated catalog (models.generated.ts and models.generated.json) for MiniMax-M3 on both the global and China providers.
• Update the MiniMax-M3 regression test to assert the video input modality and the thinking compatibility flag.

MiniMax-M2.7 is intentionally left unchanged, as its parameters already match.

Checks

• pnpm --filter @gsd/pi-ai exec vitest --run test/generated-models.test.ts — 14 passed (includes the catalog mirror and validation tests).
• pnpm --filter @gsd/pi-ai exec tsc -p tsconfig.json --noEmit — clean.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[github-actions]

🔴 PR Risk Report — CRITICAL

Files changed	6
Systems affected	2
Overall risk	🔴 CRITICAL

Affected Systems

Risk	System
🔴 critical	Agent Core
🟠 high	AI Providers

File Breakdown

Risk	File	Systems
🔴	packages/pi-ai/src/types.ts	Agent Core, AI Providers
⚪	packages/pi-ai/scripts/generate-models.ts	(unclassified)
⚪	packages/pi-ai/src/model-catalog.ts	(unclassified)
⚪	packages/pi-ai/src/models.generated.json	(unclassified)
⚪	packages/pi-ai/src/models.generated.ts	(unclassified)
⚪	packages/pi-ai/test/generated-models.test.ts	(unclassified)

⚠️ 🔴 Critical risk — the following systems require verification before merge:

• 🔴 Agent Core: check for race conditions and loop exit paths in the agent loop
• 🟠 AI Providers: confirm provider API contract and error handling are intact

⛔ This PR should not be merged without executing this follow-up prompt.

Ask your coding agent to verify before submitting:

Review this PR for risks in: Agent Core, AI Providers. Verify:

1. check for race conditions and loop exit paths in the agent loop
2. confirm provider API contract and error handling are intact

Before modifying any code, assess the scope of this fix:

- Identify the root cause, not just the reported symptom.
- Search the codebase for other call sites, similar patterns, or duplicated logic that may share the same bug.
- List affected tests, documentation, and any downstream consumers that depend on the current behavior.
- Flag any changes that extend beyond the immediate file or function.

Report findings first. Then propose a fix scoped to the actual root cause, and wait for confirmation before applying changes outside the originally reported location.

💡 Have a Codex subscription? Get an independent second opinion: codex review --adversarial

[jeremymcs]

@octo-patch please create a matching issue for PR