Skip to content

feat: support bounded model output modality lists #472

Description

@sjungwon03

Problem

Compatible model discovery rejects the documented output_modalities=text,image query. External clients cannot discover the union of authorized output capabilities in one request.

Expected behavior

Support one to nine distinct exact documented output modalities as a single comma-separated query value on GET /api/v1/models. Match any listed output against captured administrator metadata after model and final-provider IAM. Keep all as a standalone sentinel, preserve omitted-filter behavior, existing cross-filter conjunction, search/order/paging and safe relative continuations. Reject empty/duplicate/unknown/whitespace/case-variant items, mixed all and repeated query keys before catalog reads. Keep input_modalities and supported_parameters singleton until their list semantics are established.

Acceptance

Public and installed OpenRouter/OpenAI SDK regression tests cover union semantics, bounds, paging, fresh Deny, authentication, invalid syntax, required audit/catalog failures and privacy. Whole catalog validation and no inference/secret/limit/usage calls remain enforced. Record an English plan before coding and red/green evidence; run npm run check.

Refs #116; unresolved #7 and complete compatibility remain open. Assisted-by: Codex.

Activity

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

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions