Add Forge embeddings provider

[JCorners68]

Adds a named forge embeddings provider. Forge (by Voxell) is an OpenAI-compatible embeddings API, so this mirrors the existing Voyage/Scaleway/Tensorix providers rather than introducing any custom logic — the OpenAI base class's _embed() handles the embedding path.

Forge is embeddings-only (no chat/autocomplete), so this follows the embeddings-provider precedent set by #3288 (Scaleway). No edits were made to core/llm/autodetect.ts (chat-only maps; Voyage/Scaleway are absent there too).

Provider details

• OpenAI-compatible, base URL https://api.voxell.ai/v1/
• Authorization: Bearer <key>
• Models: forge-turbo (1024d), forge-pro (2560d, default), forge-ultra-4k (4096d); OpenAI aliases also accepted; dimensions supported

Changes

• core/llm/llms/Forge.ts — class Forge extends OpenAI (mirrors Tensorix.ts)
• core/llm/llms/index.ts — import + register in LLMClasses
• packages/openai-adapters/src/index.ts + types.ts — openAICompatible(...) case + provider literal
• core/control-plane/schema.ts — forge in modelDescriptionSchema and embeddingsProviderSchema
• packages/config-types/src/index.ts — forge in both provider enums (matching Scaleway)
• gui/src/pages/AddNewModel/configs/providers.ts — forge block (mirrors tensorix)
• docs/customize/model-providers/more/forge.mdx + docs/docs.json nav
• packages/llm-info/src/providers/forge.ts + registered in index.ts
• packages/openai-adapters/src/test/main.test.ts — config test (apiBase/apiKey), mirroring the Inception case

Notes

• extensions/vscode/config_schema.json is left untouched. There's no generator that writes it from these schemas in the tree (the prepackage scripts only copy it), and hand-editing it is out of scope — so it's intentionally not modified here.

How tested

• packages/openai-adapters: npm run build (tsc) clean; new Forge config test passes (vitest run).
• packages/llm-info and packages/config-types: npm run build (tsc) clean.
• core: the changed files introduce no new type errors (baseline vs. branch tsc --noEmit error counts are identical; remaining errors are pre-existing workspace-link resolution unrelated to this change).

---

Summary by cubic

Adds a new OpenAI-compatible embeddings provider: Forge (Voxell). Lets users configure forge-turbo, forge-pro (default), and forge-ultra-4k for embeddings via the existing OpenAI path.

• New Features
  • New Forge provider extending OpenAI with defaults: apiBase https://api.voxell.ai/v1/, model forge-pro, batch size 128.
  • Registered in the LLM registry and packages/openai-adapters (constructLlmApi) as an OpenAI-compatible endpoint.
  • Added forge to embeddings/model schemas in core/control-plane and packages/config-types.
  • New GUI provider card with API key input; added docs page and nav entry.
  • packages/llm-info now lists Forge and its models.
  • Added config tests in packages/openai-adapters.
  • Embeddings-only; not added to chat autodetect maps.

^{Written for commit 811618d. Summary will update on new commits.}

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[cubic-dev-ai]

No issues found across 12 files

_{Re-trigger cubic}