Add Kyma API as a built-in provider

[sonpiaz]

Proposal

Add Kyma API as a built-in provider in Continue.dev.

What is Kyma?

Kyma is an OpenAI-compatible API gateway that gives developers access to 20+ open-source models through a single API key and endpoint. It's designed for coding agents and AI assistants.

Why add Kyma?

• OpenAI-compatible: Standard /v1/chat/completions endpoint — same pattern as OpenRouter, Together, etc.
• 20+ models: Qwen 3.6 Plus, DeepSeek V3/R1, Kimi K2.5, Gemma 4, Llama 3.3, MiniMax M2.5, and more
• Free to start: $0.50 free credits on signup, no credit card required
• Auto-failover: Multi-provider routing with automatic fallback across inference providers
• Prompt caching: 90% discount on cached tokens
• Tool calling: Native function calling support across most models

Implementation Plan

Since Kyma is OpenAI-compatible, the implementation would follow the Mimo/Nous Research pattern:

1. `core/llm/llms/KymaApi.ts` — Extend OpenAI class (~15 lines)
2. Register in `core/llm/llms/index.ts`
3. Add to `core/llm/autodetect.ts`
4. Model entries in `gui/src/pages/AddNewModel/configs/models.ts`
5. Provider config in `gui/src/pages/AddNewModel/configs/providers.ts`
6. Documentation at `docs/customize/model-providers/more/kyma.mdx`
7. Logo asset
8. Tests using `createOpenAISubclassTests`

Happy to submit the PR once this is discussed.

Links

• Website: https://kymaapi.com
• Docs: https://docs.kymaapi.com
• API: `https://kymaapi.com/v1\`
• GitHub: https://github.com/sonpiaz/kyma-api
• Models: https://kymaapi.com/v1/models