11 Commits

Author SHA1 Message Date
Aiden Cline af5c74ff69 fix: align *-latest alias reasoning_options and metadata with current targets
Latest aliases were left on previous-generation controls/pricing after the
underlying models rotated (e.g. gemini-flash-latest still had Gemini 2.5
toggle+budget while mapping to 3.5 Flash effort). Sync RO, cost, and
related metadata to the models these aliases currently resolve to.
2026-07-14 22:15:49 -05:00
Adam 5c281e4feb feat: model descriptions 2026-07-01 14:16:49 -05:00
Aiden Cline 004fa2b9c3 [orcarouter] Refine reasoning effort metadata 2026-06-27 17:23:56 -05:00
Aiden Cline 0ccc0933d7 [orcarouter] Narrow reasoning options metadata 2026-06-27 17:16:38 -05:00
Aiden Cline f6c09475e0 [orcarouter] Fix reasoning options metadata 2026-06-27 11:30:13 -05:00
Aiden Cline 4e85eac00a docs: document provider reasoning request formats 2026-06-26 09:37:21 -05:00
Adam c7e827b320 fix(sync): use zhipuai metadata (#2010) 2026-06-04 18:30:39 -05:00
Adam f6c6f04367 feat(models): add model metadata (#1974)
* feat(models): add model metadata

* feat(models): rename model metadata namespaces
2026-06-03 15:13:54 -05:00
zhenjun.chen 022b1b9946 feat(orcarouter): expand to 80 chat models and add brand logo
Adds 55 additional upstream-mirrored models alongside the existing 25,
covering the full OrcaRouter chat catalog as exposed by
https://www.orcarouter.ai/api/pricing (text-only chat — TTS, embeddings,
video, and image generation are filtered out).

Per-namespace upstream mappings used by [extends]:

  OrcaRouter ns  -> models.dev provider
  ---------------- + ---------------
  openai         -> openai
  anthropic      -> anthropic   (dot version -> dash, e.g. opus-4.7 -> opus-4-7)
  google         -> google
  deepseek       -> deepseek
  qwen           -> alibaba
  grok           -> xai
  kimi           -> moonshotai
  minimax        -> minimax     (minimax-m2.7 -> MiniMax-M2.7)
  z-ai           -> zai

OrcaRouter-specific aliases (dated snapshots like gpt-5-2025-08-07,
search-preview variants, qwen3-vl-* visual variants) are excluded from
v1 because their upstream canonical files do not yet exist in models.dev.

Also adds providers/orcarouter/logo.svg.
2026-05-15 14:48:31 +08:00
zhenjun.chen 699284ce91 chore(orcarouter): drop oversize logo, fall back to models.dev default
The previously committed logo is ~100KB; existing wrapper-provider logos
(openrouter, llmgateway, kilo, aihubmix, ambient) are all 0.3-6KB and use
`currentColor`. Falling back to the default logo per README:

  > If we don't have a provider's logo, a default logo is served instead.

A properly-sized currentColor logo will follow in a separate PR.
2026-05-14 21:37:48 +08:00
zhenjun.chen 7102978cb4 feat: add OrcaRouter provider
OrcaRouter is an OpenAI-compatible meta-router aggregating 150+ LLMs
(OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen, Kimi, MiniMax, ...)
behind a single API key, with a virtual orcarouter/auto smart-routing
entry that picks an upstream per request.

This initial scope covers 26 models (1 AUTO router + 25 upstream mirrors
using [extends]). Pricing computed from https://www.orcarouter.ai/api/pricing
on 2026-05-14: input = model_ratio * $2, output = model_ratio *
completion_ratio * $2 (USD per 1M tokens).

Disclosure: I'm an engineer on the OrcaRouter team.
2026-05-14 20:55:47 +08:00