feat(providers): add Kie provider for Claude, GPT, Grok, and Kimi models

[waleedlatif1]

Summary

• Add Kie (kie.ai) as a BYOK-only model provider with 22 chat models under a kie/ prefix: Claude (Opus 5.5 → Haiku 4.5), GPT 6 / 5.6 / 5.5, Grok 4.6 / 4.5 / 4.3, and Kimi K3
• Claude routes through the shared Anthropic core against Kie's Messages proxy (bearer auth); GPT/Grok/Kimi route through the shared Responses core, one endpoint per model family
• Pricing is Kie's own per-model rate; capabilities mirror the upstream entries, limited to what Kie documents (e.g. reasoning effort low–xhigh)
• Kie answers most failures with HTTP 200 + { code, msg }; createKieFetch rewrites those into real HTTP errors so bad keys fail fast, rate limits retry, and failures never surface as empty completions
• Kie defaults stream to true, so the Responses and Anthropic cores now send stream: false explicitly on non-streaming calls (a no-op for OpenAI/Anthropic)
• ResponsesProviderConfig.capabilityModel lets a reseller's wire slug differ from the catalog id used for reasoning-effort lookup
• Conversation history protocol can now resolve per model (getProviderHistoryProtocol), since Kie serves both Anthropic and Responses wire formats under one provider
• BYOK wiring (settings row, contract enum, key resolution), icon, landing catalog, usage icon, tokenization, regenerated agent streaming docs

Type of Change

• New feature

Testing

• Live against the Kie API: Grok 4.6/4.5/4.3 and Kimi K3 verified for plain, streaming, reasoning effort, strict JSON output, non-streaming and streaming tool loops, and memory tool replay (conversation-smoke.test.ts responses case); computed cost matches Kie's billed credits
• All 22 routes × {plain, stream} return clean errors for an invalid key (previously a crash on Claude and a silent empty success on GPT/Grok)
• Claude and GPT backends on Kie were intermittently unavailable upstream during testing; one Claude response confirmed the Messages wire shape
• providers/kie/transport.test.ts (reverting the guard turns 6 of 10 red), history-adapters.test.ts, providers/memory/BYOK suites (2,204 tests)
• bun run lint, check:audits, docs-manifest:check, block registry check, type-check

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Sep 28, 2026 10:34pm UTC

[cubic-dev-ai]

1 issue found across 24 files

Confidence score: 4/5

• In apps/sim/providers/models.ts, accepting arbitrary kie/... IDs as custom models can let unsupported non-Claude IDs pass validation and then fail at execution; restrict the custom namespace to Claude IDs.

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="apps/sim/providers/models.ts">

<violation number="1" location="apps/sim/providers/models.ts:3706">
P2: This pattern accepts arbitrary `kie/...` IDs as custom models, but non-Claude IDs must match the fixed Responses endpoint map and otherwise fail at execution. Restrict the custom namespace to Claude IDs; the 11 catalogued GPT/Grok/Kimi IDs still resolve through exact catalog matches.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Adds a new LLM provider integration for Kie.

The PR appears safe to merge; no outstanding findings or accepted new issues remain.

Summary

The PR adds a BYOK-only Kie model provider, routing Claude through the Anthropic Messages core and GPT, Grok, and Kimi through the Responses core. It also adds model catalog and settings support, error-envelope handling, and per-model conversation-history protocol selection.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Request[Agent model request] --> Kie[Kie provider]
  Kie -->|Claude| Anthropic[Anthropic Messages core]
  Kie -->|GPT, Grok, Kimi| Responses[Responses core]
  Anthropic --> Fetch[Kie fetch wrapper]
  Responses --> Fetch
  Fetch --> API[Kie API]

Loading

_{Reviews (4) · Last reviewed commit: "fix(providers): parse untyped Kie JSON r..."}

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

All reported issues were addressed across 25 files

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

All reported issues were addressed across 26 files

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 26 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Re-trigger cubic}