feat(vancine): add Vancine provider catalog

[fx247562340]

Summary

Add Vancine as a new OpenAI-compatible provider catalog entry for three publicly listed models:

• kimi-k3
• glm-5.2
• MiniMax-M3

Vancine is an independent third-party API aggregation platform. It is not an official service of Moonshot AI, Zhipu/Z.AI, or MiniMax.

Provider

Field	Value
ID	vancine
Name	Vancine
npm	@ai-sdk/openai-compatible
API	https://vancine.com/v1
Env	VANCINE_API_KEY
Docs	https://vancine.com/docs

Models and public default-group pricing

Prices are in USD per one million tokens and reflect Vancine public default-group introductory pricing with group_ratio.default = 1.

Model ID	Base model	Input	Output	Cache read	Context
kimi-k3	moonshotai/kimi-k3	2.95	14.76	0.30	1,048,576
glm-5.2	zhipuai/glm-5.2	1.10	3.840677966102	0.275	1,000,000
MiniMax-M3	minimax/MiniMax-M3	0.62	2.48	0.12	1,000,000

Prices are derived from the public pricing catalog:

• input = model_ratio × 2 × group_ratio.default
• output = input × completion_ratio
• cache_read = input × cache_ratio

All three models use reasoning_options = [] because no free, publicly verified Vancine per-model reasoning-control syntax is documented for these endpoints.

Logo

The provider logo uses the current Vancine W mark with a square viewBox, no fixed dimensions, and currentColor fills for light and dark themes.

Public sources

• Documentation: https://vancine.com/docs
• Pricing API: https://vancine.com/api/pricing
• Model pages:
  • https://vancine.com/pricing/kimi-k3
  • https://vancine.com/pricing/glm-5.2
  • https://vancine.com/pricing/MiniMax-M3
• OpenAI-compatible API base: https://vancine.com/v1

Unauthenticated GET /v1/models returns 401 as expected and was used only to confirm that the endpoint exists. No real API keys or paid model calls were used.

Identity disclosure

Vancine is a third-party aggregator providing OpenAI-compatible access. Model names and trademarks belong to their respective upstream providers. This contribution does not claim official affiliation with Moonshot AI, Zhipu/Z.AI, or MiniMax.

Validation

• bun validate — pass
• Generated Vancine catalog contains exactly the three target models
• xmllint --noout providers/vancine/logo.svg — pass
• git diff --check — pass
• Public documentation and pricing links — pass
• Secret scan — no findings
• bun test — 93 pass, 1 known upstream baseline failure

Known upstream baseline failure

The existing test catalog generation > repository open-weight model metadata includes weights links reports missing weights links for multiple pre-existing models/* entries on origin/dev, including moonshotai/kimi-k3.

This contribution does not modify models/, packages/, or the affected test. The contribution scope is limited to providers/vancine/**.

Explicit non-actions

• No paid model or API calls
• No real Vancine API key usage
• No deployment
• No upstream-cost, zero-profit, at-cost, or official-affiliation claims

🤖 Generated with Claude Code

[github-actions]

No actionable findings.