Add onomeo provider

[Owaid0723666]

Adds onomeo, an OpenAI-compatible gateway. I operate onomeo; this PR comes from its operator.

• API: https://onomeo.com/v1 (/chat/completions, streaming supported), @ai-sdk/openai-compatible
• Auth: Bearer key created at https://onomeo.com/dashboard. Env var ONOMEO_API_KEY, the same name the site's own OpenCode guide uses: https://onomeo.com/integrations/opencode
• Docs: https://onomeo.com/docs
• Model IDs are exactly what the API accepts; the live list is at https://onomeo.com/api/info (models) and each has a page at https://onomeo.com/models/<id>.

No [cost] tables

onomeo has no per-token USD price. Each reply deducts usage.total_tokens from a credit balance, counted from the visible text sent and received (reasoning is not charged), and premium models deduct more per call. Credits come from a free daily check-in (1,200 a day, up to 3,500 on a 7-day streak); an optional $5 membership and a $5 top-up add credits. There is no honest USD/MTok figure to publish, so the model files omit [cost], as the credit and subscription plans already in the repo do (xiaomi-token-plan-*, stepfun-step-plan). The provider.toml header records this with sources.

Limits: 12 requests/min per key, 60 requests per 5 hours per account. The docs FAQ describes the service as a free beta with no uptime promise.

Models (7), all base_model with overrides only

id	base_model	overrides	basis
gemini-3.1-flash-lite	google/gemini-3.1-flash-lite	effort minimal/low/medium/high; input text, image	onomeo accepts images only as PNG/JPEG/WebP/GIF data URLs (https://onomeo.com/docs#images); effort from the Google lab entry and OpenAI-compatible peers
gemini-3-flash-preview	google/gemini-3-flash-preview	same	same
@cf/openai/gpt-oss-120b	openai/gpt-oss-120b	effort low/medium/high; context 128,000, output 16,384	Cloudflare Workers AI model id; values from providers/cloudflare-workers-ai/models/@cf/openai/gpt-oss-120b.toml
@cf/meta/llama-4-scout-17b-16e-instruct	meta/llama-4-scout-17b-instruct	context 131,000	Cloudflare peer entry for the same id
openai/gpt-oss-20b	openai/gpt-oss-20b	effort low/medium/high	lab entry and peers
codestral-latest	mistral/codestral-latest	none
mistral-Nemo-Instruct-2407	mistral/mistral-nemo	none

Not verified for this PR

onomeo does not publish which upstream serves each model. The reasoning_options follow the lab and same-surface peer entries, per the relay policy in AGENTS.md; each effort value was not re-tested through onomeo. tool_call is inherited from the lab entries and was not tested separately.

This first PR leaves out models whose serving surface I could not pin down from the ID (DeepSeek V4 Flash, GLM-5.2, Nemotron 3 Super and others) and the :free IDs.

Logo

onomeo's pixel-cat icon (https://onomeo.com/icon.svg) with each original color layer set to currentColor plus a fill-opacity. The path data is unchanged; no width/height; square viewBox="0 -2 40 40".

Validation

generate() (the function bun validate runs) passes locally for providers/onomeo against the repo's models/. A full bun validate could not run on my Windows checkout, because the repo's symlinked TOML files and filenames containing : do not check out on Windows; CI runs it on Linux.

🤖 Generated with Claude Code

[github-actions]

No actionable findings.