Skip to content

Add AIgateway provider (84 text & vision models)#1569

Open
Rakesh1002 wants to merge 1 commit intoanomalyco:devfrom
Rakesh1002:add-aigateway-provider
Open

Add AIgateway provider (84 text & vision models)#1569
Rakesh1002 wants to merge 1 commit intoanomalyco:devfrom
Rakesh1002:add-aigateway-provider

Conversation

@Rakesh1002
Copy link
Copy Markdown

Summary

Adds AIgateway (aigateway.sh) as a provider. AIgateway is a universal, OpenAI-compatible inference API that aggregates 100+ AI models across every modality, reached at api.aigateway.sh/v1.

This PR contributes:

  • providers/aigateway/provider.toml — OpenAI-compatible (@ai-sdk/openai-compatible), env = ["AIGATEWAY_API_KEY"], doc: https://aigateway.sh/models
  • providers/aigateway/logo.svg — theme-safe SVG using currentColor
  • 84 model TOMLs across text + vision modalities covering Anthropic (Claude Opus/Sonnet/Haiku 4.x), OpenAI (GPT-5 / 5.4 / 4.1 / o4-mini / gpt-oss), Google (Gemini 3.x, Gemma 3/4), Meta (Llama 3.x / 4), Mistral, xAI (Grok 4 / 4 Fast), Moonshot (Kimi K2.5 / K2.6), DeepSeek, Microsoft Phi-2, Alibaba Qwen 2.5 / 3 / 3.5, IBM Granite, Perplexity Sonar, NVIDIA, MiniMax, AI Singapore SEA-Lion, Zhipu GLM, and more.

Testing

  • All 85 TOML files parse with tomllib locally.
  • Each model entry includes the required name, family, release_date, last_updated, attachment, reasoning, tool_call, structured_output, temperature, open_weights, [cost] (input/output in USD/Mtok), [limit] (context/input/output), [modalities].
  • Generated by a script in the aigateway repo so we can keep this entry fresh as pricing and the catalog evolve.

Notes for reviewers

  • A small subset of open-weight community models (some HuggingFace / older Meta Llama / Qwen 1.5 variants) default release_date to 2024-01-01 where the upstream catalog does not track a specific date. Happy to tighten those in a follow-up if you'd prefer strict dates only.
  • The aigateway catalog also includes audio (STT/TTS), image, video, embedding, translation, rerank, classification and OCR models. Those are not included here because the current models.dev schema targets chat models. If/when the schema grows to cover other modalities, I'll send a follow-up.
  • Follows the same shape as existing aggregator providers on models.dev — see providers/openrouter, providers/helicone, providers/requesty, providers/llmgateway, providers/fastrouter. We register as a provider rather than duplicating canonical entries. Happy to convert everything to [extends] in a follow-up if that's preferred.

Thanks for maintaining this — AIgateway relies on the models.dev JSON feed for our own catalog page and we want to contribute back. 🙏

AIgateway (https://aigateway.sh) is a universal inference API that
aggregates 100+ AI models across every modality behind a single
OpenAI-compatible endpoint. Users reach it at api.aigateway.sh/v1.

This PR adds:
  - providers/aigateway/provider.toml (OpenAI-compatible, @ai-sdk/openai-compatible)
  - providers/aigateway/logo.svg (currentColor, theme-safe SVG)
  - 84 model TOMLs across text + vision modalities (chat + attachment)
    covering Anthropic, OpenAI, Google, Meta, Mistral, xAI, Moonshot,
    DeepSeek, Microsoft, Alibaba Qwen, IBM Granite, Perplexity, NVIDIA,
    MiniMax, and more.

Notes for reviewers:
  - A subset of community HuggingFace models default release_date to
    2024-01-01 where the upstream catalog does not track it. Happy to
    amend via a follow-up if the project prefers to omit them.
  - Audio (STT/TTS), image, video, embedding, translation, and
    rerank models from the aigateway catalog are deliberately NOT
    included because the current models.dev schema targets chat models.
    If/when the schema grows to cover them I'll send a follow-up PR.

Source of truth: web/src/lib/catalog.generated.json in the
aigateway repo. Regeneration is one command
(`node scripts/build-modelsdev-pr.mjs`) so we can keep this entry
fresh as we update pricing and add models.
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.

1 participant