Skip to content

[OA-021] Add an OpenAI image-vision adapter #21

Open

Description

@johnnycoderbot-cloud

Planned implementation slice OA-021. No application behavior is implemented by this issue.

Phase: later · Dependencies: OA-005, OA-006, OA-008

Full roadmap detail · Contracts

Requirements

OA-MOD-004, OA-DAT-003, OA-DAT-004, OA-DAT-005, OA-AI-003, OA-AI-006

Target files

  • packages/providers/src/openai/inference.ts
  • packages/providers/src/openai/capabilities.ts
  • tests/providers/openai.spec.ts
  • docs/integrations.md

Acceptance criteria

  • Select a currently documented image-capable API/model and record exact supported content shape; image analysis is not described as continuous video understanding.
  • Use the shared observation schema, source evidence, context budget, frame masks, and explicit cloud destination consent.
  • API billing and credentials are separate from ChatGPT application subscriptions; profile selection never uploads frames by itself.
  • Bound spend/retries/concurrency and handle refusals/unsupported images/timeout; no unverified realtime video/voice behavior is required.

Verification

  • Official documented schema fixture review and shared mock provider conformance suite.
  • Consent-denial, redaction, cancel/late-result, rate-limit and budget tests; opt-in synthetic live image check.

Gate and estimate

L3: current API/model capability, privacy/billing choice, and consent/budget gates pass.

TBD after the prerequisite capability/authorization spike; excluded from initial work-unit budget. Split the implementation into 30-60 minute work units after evidence.

Handoff

Record changed files, actual commands/results, mock versus live evidence, residual limits, and review verdict. Do not connect personal feeds or accounts in CI.

Bounded work units

  • Record chosen capability, permissions, and current primary-source evidence (30-60 minute discovery unit; implementation estimate TBD)
  • Produce synthetic contract/negative-test fixture and a bounded implementation breakdown (30-60 minute planning unit; implementation estimate TBD)
  • Review gate evidence and decide pass, defer, or changes required before external effects (30-60 minute review unit; implementation estimate TBD)

Linked dependencies

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    integration-gateExternal or native capability requires reproducible evidencelaterOutside initial alpha scopeplannedExecution-ready planned work; not implemented

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions