[FEATURE] Complete AINative managed chat-completion integration (partially integrated; 3 disconnects block user access) + IDE-plan launch-readiness audit

[urbantech]

Summary

Gap analysis (code-verified) of whether the AINative managed chat-completion API is integrated into the IDE. Verdict: PARTIALLY INTEGRATED (~60-70%). All infrastructure exists but is disconnected at 3 specific glue points, so no IDE user can actually use AINative-billed inference today — even though the provider, auth service, SSE streaming, and usage tracking are all implemented.

This matters for product/billing: the IDE plan tier is a separate, unlaunched product line (family=ide_cli, 0 active subscribers, placeholder Stripe price IDs, metered by completions/chat NOT the credit pool). It cannot be sold until (a) this integration is completed and (b) its Stripe pricing + metering pass their own audit. This issue tracks the integration gap; a separate audit is needed before the IDE plan goes on sale.

What EXISTS (built, verified)

• ✅ AINativeCloudProvider class — full impl: JWT auth, SSE streaming, 401 auto-refresh, reconnect. Real endpoint https://api.ainative.studio — src/vs/workbench/contrib/ainative/electron-main/llmMessage/providers/ainativeCloudProvider.ts:52-65
• ✅ AINativeCloudAuthService — JWT auth, encrypted token storage, 5-min-buffer refresh
• ✅ ManagedChatAPIService — types for credits_consumed/credits_remaining
• ✅ usageTrackingService — trackManagedUsage(), getCreditsStatus(), quota monitoring (wired into chatThreadService)
• ✅ SSE streaming implementation (10 test cases)
• ✅ ainativeCloud registered as a ProviderName with 4 models defined

What is MISSING (the 3 disconnects)

#	Gap	File:Line	Effect
1	sendAINativeCloudChat is a stub returning a hard-coded error ("not yet available... TASK-006"), wired as the provider's sendChat	sendLLMMessage.impl.ts:850-857, :951	Dispatcher never calls the real provider
2	displayInfoOfProviderName() throws for ainativeCloud (no case)	ainativeSettingsTypes.ts:113	Settings UI crashes → user can't select the provider
3	subTextMdOfProviderName() throws for ainativeCloud (no case)	ainativeSettingsTypes.ts:135	Same UI crash on help text

Secondary: no SDK client method for the chat-completions call yet (only /v1/auth/*); model registry still uses mock data (not live /v1/models); usage tracking is wired but inert because the API never returns credit data.

⚠️ Endpoint correctness note (backend integration)

ainativeCloudProvider.ts:54 targets /v1/chat/completions. Our billed managed path is /api/v1/managed/chat/completions (ManagedChatService — the gated, credit-metered handler). Wiring this up must point at the managed endpoint so credit deduction + the single-credit-pool billing (cost×2, chat_credit_service) actually fire. The bare /chat/completions path is the legacy handler and would bypass metering.

Tasks to close the integration

• Implement sendAINativeCloudChat to call AINativeCloudProvider.sendChatCompletion() (inject IAINativeCloudAuthService) — sendLLMMessage.impl.ts:850
• Add ainativeCloud case to displayInfoOfProviderName() — ainativeSettingsTypes.ts:113
• Add ainativeCloud case to subTextMdOfProviderName() — ainativeSettingsTypes.ts:135
• Point the provider at the managed billed endpoint (/api/v1/managed/chat/completions), not /v1/chat/completions
• Add SDK client sendChatCompletion() method
• Replace model-registry mock data with live /v1/models
• Verify credits_consumed/credits_remaining surface in the usage tracker end-to-end
• Before launching the IDE plan: replace placeholder Stripe price IDs + audit IDE-tier metering (currently completions/chat, not credits) against the single-credit-pool model

Estimated effort

~2-3 weeks (1 backend + 1 frontend). Architecture is solid; this is glue + endpoint correctness + a launch-readiness billing audit.