Provider hardening pack: Groq, Cerebras, Together, DeepInfra, Hyperbolic, SiliconFlow

[Hmbown]

Shape. Six new provider entries after #2 lands. Each is one struct + base URL + env var + a row in the model registry:

• Groq: https://api.groq.com/openai/v1, GROQ_API_KEY
• Cerebras: https://api.cerebras.ai/v1, CEREBRAS_API_KEY
• Together: https://api.together.xyz/v1, TOGETHER_API_KEY
• DeepInfra: https://api.deepinfra.com/v1/openai, DEEPINFRA_API_KEY
• Hyperbolic: https://api.hyperbolic.xyz/v1, HYPERBOLIC_API_KEY
• SiliconFlow: https://api.siliconflow.cn/v1, SILICONFLOW_API_KEY (closes Add SiliconFlow as a built-in provider #1710)

Each entry gets at least one model row in the registry pointing to a sensible house default. All inherit the OpenAI-completions wire format from the trait.

Depends on: #2075.

---

Existing issues this resolves on PR merge

Closes:

• Closes Add SiliconFlow as a built-in provider #1710

Partially addresses:

• Partially addresses Zenmux as primary provider for Deepseek-V4-Pro & Flash #1330
• Partially addresses Feature: Support ZenMux / OpenAI-compatible providers with reasoning+cache parsing (like openrouter) #1978

Related:

• Related: Support OpenCode Go/Zen please, it provides DeepSeek-V4 as well. #1481
• Related: help for additional models (MiniMax model is more affordable for students, can use support this model?) #1310
• Related: Support hermes gateway inject #747

When opening a PR for this issue, the PR body must include the Closes #N lines for every issue above that this PR fully resolves, so they auto-close on merge.

Slate item 14/18. Tracking doc: .private/issue-slate.md (maintainer-only).