Skip to content

feat(providers): add Adorbis AI - sovereign OpenAI-compatible gateway - #3823

Open
adorbistech wants to merge 8 commits into
anomalyco:devfrom
adorbistech:add-adorbis-provider
Open

feat(providers): add Adorbis AI - sovereign OpenAI-compatible gateway#3823
adorbistech wants to merge 8 commits into
anomalyco:devfrom
adorbistech:add-adorbis-provider

Conversation

@adorbistech

@adorbistech adorbistech commented Jul 28, 2026

Copy link
Copy Markdown

Adds Adorbis AI provider with 5 department models (quick, code, writing, auto, reason).

Adorbis AI is a sovereign AI control plane routing requests across 40+ models from OpenAI, Anthropic, Google Gemini, DeepSeek, and NVIDIA through a single OpenAI-compatible API key.

Model IDs verified via live API (GET /v1/models): adorbis/quick, adorbis/code, adorbis/writing, adorbis/auto, adorbis/reason. Files placed under models/adorbis/ subfolder to match generated IDs.

Pricing: Department routing is variable passthrough — each request routes to the best available model dynamically. No fixed USD/MTok rate applies. [cost] section omitted, consistent with other variable routers in this repo (e.g. openrouter/auto).

Release date: Adorbis AI entered public availability June 2025. last_updated reflects current API state July 2026.

Fixes model IDs from previous PR #2650.

satoshibro and others added 3 commits June 17, 2026 21:32
Adorbis AI is a sovereign OpenAI-compatible AI gateway that routes
requests across Gemini 2.5 Flash, Claude, DeepSeek and local Ollama
under a single API key.

- api: https://api.adorbistech.com/v1
- npm: @ai-sdk/openai-compatible
- env: ADORBIS_API_KEY
- models: adorbis/auto, adorbis/writing, adorbis/code, adorbis/quick
- free tier: 1000 credits at ai.adorbistech.com
Replace reasoning/research with reason to match live /v1/models IDs
(adorbis/quick, code, writing, auto, reason) and add [cost] sections.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

Action items

  • [high] [violation] providers/adorbis/models/reason.toml:7 - Check: Reasoning models must declare reasoning_options. Why: reasoning = true with no reasoning_options fails the schema refine and the contribution checklist. Callers cannot tell whether Adorbis exposes any reasoning control on this OpenAI-compatible surface. Action: Add provider-authored reasoning_options. Use reasoning_options = [] unless Adorbis documents a verified control (for example effort or budget); if a control exists, declare it and note the exact request syntax in an adjacent comment.
  • [low] [possible mistake] providers/adorbis/models/auto.toml:10 - Check: Data-changing PRs should cite first-party sources mapped to claims. Why: All five models introduce [cost] and context/output limits, but the PR body only links general docs and does not map pricing or limit values to a pricing page, model catalog, or API reference. Action: In the PR body, cite direct Adorbis sources for each material claim (USD per-million pricing and context/output limits) and state what each link supports.

@github-actions

Copy link
Copy Markdown
Contributor

Action items

  • [high] [violation] providers/adorbis/models/quick.toml:1 - Check: Model IDs come from the path under providers/<id>/models/ (slashes require subfolders). Why: The PR states live /v1/models IDs are adorbis/quick, adorbis/code, adorbis/writing, adorbis/auto, and adorbis/reason, but these files resolve to quick, code, writing, auto, and reason. Clients looking up the API IDs will miss the catalog entries. Action: Move each model to providers/adorbis/models/adorbis/<name>.toml (e.g. providers/adorbis/models/adorbis/quick.toml) so the injected IDs match the provider API.
  • [high] [violation] providers/adorbis/models/reason.toml:8 - Check: Models with reasoning = true must declare reasoning_options. Why: reason sets reasoning = true with no reasoning_options, which fails the schema refine and omits the required control surface for this OpenAI-compatible gateway model. Action: Add reasoning_options for controls Adorbis actually exposes; use reasoning_options = [] if reasoning is always on with no verified caller control (and optionally a short adjacent API-syntax comment if a control is documented).
  • [low] [possible mistake] providers/adorbis/models/auto.toml:4 - Check: release_date should reflect first public availability. Why: All five models use release_date = "2024-01-01" after earlier commits used 2026-01-01, with no source for either date—looks like a placeholder. Action: Set each release_date to the real first-public date (or the best documented approximation) and cite that source in the PR body.
  • [low] [possible mistake] providers/adorbis/models/auto.toml:12 - Check: Data-changing PRs should cite first-party pricing/docs mapped to claims. Why: Costs (0.10/0.30, 0.30/1.20, 0.50/2.00) and limits are introduced without a pricing or model-spec citation tied to those values, so reviewers cannot validate them from the PR alone. Action: In the PR body, link the Adorbis pricing/docs pages and state which costs and limits each source supports.

@github-actions

Copy link
Copy Markdown
Contributor

Action items

  • [high] [violation] providers/adorbis/models/auto.toml:1 - Check: Model IDs must match the provider API ID via the path under models/ (slashes → nested folders). Why: The PR states live /v1/models IDs are adorbis/quick, adorbis/code, adorbis/writing, adorbis/auto, and adorbis/reason, but these files resolve to bare IDs (auto, code, quick, reason, writing). Clients using the catalog would call the wrong model names. Action: Move files to providers/adorbis/models/adorbis/<id>.toml (e.g. .../adorbis/auto.toml) so generated IDs match the live API, or correct the claimed API IDs if the bare names are actually what the endpoint expects.
  • [low] [possible mistake] providers/adorbis/models/auto.toml:3 - Check: release_date should be the first public release date. Why: All five models use 2024-01-01, but earlier commits in this PR used 2026-01-01; the 2024 date looks like a template placeholder rather than a verified launch date. Action: Set each model’s release_date (and last_updated if needed) to the real public dates, or cite evidence for 2024-01-01.
  • [low] [violation] providers/adorbis/models/auto.toml:12 - Check: Data-changing PRs should cite first-party sources mapped to claims. Why: [cost] and limits were added for all five department models, but the PR body only links general docs/API URLs and does not map pricing or context/output limits to a pricing page or model reference. Action: Add PR-body citations (pricing page and/or model docs) stating which URL supports each cost and limit value.

@github-actions

Copy link
Copy Markdown
Contributor

Action items

  • [medium] [possible mistake] providers/adorbis/models/adorbis/auto.toml:14 - Check: Cost values must be USD per million tokens, with non-USD source conversion documented, and fixed rates only when billing is fixed. Why: Every department TOML publishes fixed [cost] rates (e.g. auto/code/writing 0.30/1.20, quick 0.10/0.30, reason 0.50/2.00), but the PR body says pricing is in AC (1 AC = \.001, unit incomplete) and that “Department routing costs vary by underlying model tier.” That contradicts fixed catalog costs and leaves the USD conversion unverifiable; variable passthrough routers in this repo often omit cost (e.g. OpenRouter auto). Action: Confirm whether each department ID has a fixed USD/MTok rate or passthrough/variable billing. If fixed, cite the pricing page claim for each rate, convert from AC with an explicit USD rate/date in a top-of-file comment, and state that in the PR body. If variable, omit [cost] (or use another accurate representation) instead of publishing fixed numbers.

@github-actions

Copy link
Copy Markdown
Contributor

No actionable findings.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants