Skip to content

Tags

Tags

v0.9.0

Toggle v0.9.0's commit message
Grida FX 0.9.0

v0.8.1

Toggle v0.8.1's commit message
Grida FX 0.8.1

v0.8.0

Toggle v0.8.0's commit message
Grida FX 0.8.0

v0.7.1

Toggle v0.7.1's commit message
Grida FX 0.7.1

v0.7.0

Toggle v0.7.0's commit message
Grida FX 0.7.0

v0.6.1

Toggle v0.6.1's commit message
Grida FX 0.6.1

v0.6.0

Toggle v0.6.0's commit message
Grida FX 0.6.0

v0.5.1

Toggle v0.5.1's commit message
Grida FX 0.5.1

v0.5.0

Toggle v0.5.0's commit message
Grida FX 0.5.0

v0.4.0

Toggle v0.4.0's commit message

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature.
Add Grida Gateway (GG) as a provider; release 0.4.0 (#49)

Grida Gateway (`@gg`) serves text, agent turns, images, video, 3D models
and rigs with one key, `GG_API_KEY` (base `GG_BASE_URL`). It is an
ordinary provider: adapters behind the shared transport, no engine change.

- Adapters (spec/providers.md §9.7): `gg-image` (generate, multipart edit),
  `gg-structured` and `gg-tool-loop` (chat completions, with
  `image_detail` and `reasoning_effort`), and long jobs `gg-video`
  (first frame as a multipart part), `gg-mesh` (multiview from local
  views) and `gg-rig` (a free riggability check, then the rig).
- Costs: GG's settled charge, from `x-gg-charged-mills` or the request's
  billing readback, for answers and for failures GG accepted; the
  reservation is never booked. An unsettled charge falls back to the
  whole hold, as for any provider.
- Idempotency: one key per attempt and invocation; a 429 that says
  nothing was admitted and asks for a new key renews it.
- The OpenAI-style chat code OpenRouter used is now the provider-neutral
  `chat` module both share.
- Nine built-in `@gg` routes (26 in all); the guide, README, skill and
  two examples lead with GG; key lists across tools and tests gain the
  GG variables.
- Version 0.4.0 in every manifest (tools/check_versions.py agrees).