[Inference Provider]: Add Oxlo.ai as an inference provider

[ms-shashank]

Add OxloAI as an Inference Provider

Oxlo.ai is a developer-first AI inference platform with request-based pricing (flat fee per API call, not per token). We serve 36+ open-source models including Qwen 3 32B, Llama 3.3 70B, DeepSeek R1, Mistral, Gemma 3, and Whisper.

What this PR adds

• packages/inference/src/providers/oxloai.ts — Provider implementation for conversational and text-to-image tasks
• Registration in types.ts, consts.ts, and getProviderHelper.ts

API Compatibility

OxloAI is fully OpenAI SDK compatible:

• Chat: POST /v1/chat/completions
• Images: POST /v1/images/generations

Organization

• HF Org: huggingface.co/oxlo-ai
• Website: oxlo.ai

We're submitting this to get the technical integration reviewed. We have 36 models ready to map using the Model Mapping API once approved.

---

Note

Medium Risk
Adds a new external provider integration with custom request/response shaping for text-to-image, which could surface runtime issues if OxloAI’s API behavior differs from assumptions. Changes are otherwise additive and gated behind selecting the new oxloai provider.

Overview
Adds OxloAI as a new inference provider option and wires it into provider selection/typing (INFERENCE_PROVIDERS, PROVIDERS_HUB_ORGS, HARDCODED_MODEL_INFERENCE_MAPPING, and PROVIDERS).

Introduces providers/oxloai.ts implementing conversational via the shared OpenAI-compatible v1/chat/completions route and a text-to-image helper targeting v1/images/generations, including payload mapping (inputs→prompt, response_format) and response decoding for url, base64-to-Blob, and dataUrl outputs with validation/erroring on malformed responses.

^{Reviewed by Cursor Bugbot for commit 0a8de69. Bugbot is set up for automated code reviews on this repo. Configure here.}