Make your computer a home for your AI partner. Avibe is the local-first Agent OS: Vibey, its built-in AI partner, works out of the box, and Claude Code, Codex & OpenCode run unchanged beside it — reach them from your browser, phone, or any chat app.
Python agentagent-osaiai-agentsai-partnerchatopsclaudeclaude-codecodexdevtoolsdiscord-botlark-botllmlocal-firstlong-term-memoryopencodeslack-botvibe-codingvibeywechat
feat(model-hub): offer gpt-image-2 on a Codex subscription (#2810)
Call White48d2ef1c896e
diff --git a/docs/plans/vibey-hosted-tools.md b/docs/plans/vibey-hosted-tools.mdindex abc2fba37..4948ee9f1 100644--- a/docs/plans/vibey-hosted-tools.md+++ b/docs/plans/vibey-hosted-tools.md@@ -634,11 +634,16 @@ Unchanged from v1 §7.3 (owner decision, 2026-10-13 19:14), with the producer noempty, `xai` or `grok` prefix and then routes the bare id, which `force-model-prefix` never registers (unchangedin v8.0.28). That 400 is the engine's own refusal, recognized as local, never learned. xAI ships with no imagemodels until the engine or the xAI rendering changes.- - **A Codex (ChatGPT) subscription is not offered images.** The engine runs it through the same direct Image API+ - **A Codex (ChatGPT) subscription** (`gpt-image-2`): the engine runs it through the same direct Image APIfunction against `https://chatgpt.com/backend-api/codex`, a host no stub can stand in for (an OAuth auth file- takes no base URL), so no hermetic E2E verifies it. The overlay's `kinds: ["api_key"]` keeps the subscription- out. Codex CLI itself calls `chatgpt.com/backend-api/codex/images/generations`, so the path very likely works;- one real call at a release E2E, with the owner's approval, can turn it on.+ takes no base URL), so no hermetic E2E verifies it. One real call with the owner's approval verified it on+ 2026-10-18. The call went through engine v8.0.23 with a Codex subscription credential, and its body was the+ one the gateway sends. It returned one `b64_json` image with HTTP 200, plus `usage` with 515 output image+ tokens, and no refusal. Its cost reads as `included`, as for every monthly Source. The subscription path+ ignored `output_format`, `output_compression` and `size`: it answered with a PNG at 1254×1254 and `quality: low`.+ The result reader takes any supported image type, and the size an answer reports is the size recorded. So+ `gpt-image-2` lists both credential kinds. The overlay's other image models stay `api_key` only until a call+ verifies them.- **Route verification is the implementation's job, not the user's.** A model whose route is not verified shipswithout an overlay entry, or without the credential kind it was not verified on, so it is never offered and theUI never needs an "unverified" tag.diff --git a/tests/test_model_hub_hosted_tools_resolution.py b/tests/test_model_hub_hosted_tools_resolution.pyindex d8f7b1628..8d150044f 100644--- a/tests/test_model_hub_hosted_tools_resolution.py+++ b/tests/test_model_hub_hosted_tools_resolution.py@@ -72,7 +72,7 @@ def _image(model_id: str, description: str, *, price: float | None = None, kindsIMAGE_MODELS = {- # Verified on API keys only: a subscription that lists the model is not offered it.+ # An image model verified on API keys only: a subscription that lists it is not offered it."openai": (_image("gpt-image-2", "OpenAI's image model", kinds=("api_key",)),),"xai": (_image("grok-imagine-image", "Grok Imagine", price=0.02),),}diff --git a/vibe/data/backend_models.json b/vibe/data/backend_models.jsonindex 5be21b1e8..130fc9669 100644--- a/vibe/data/backend_models.json+++ b/vibe/data/backend_models.json@@ -88,7 +88,7 @@"description": "GPT Image 2: precise prompts, legible text, edits","params": ["images", "n", "size"],"size_forms": ["pixels"],- "kinds": ["api_key"],+ "kinds": ["api_key", "subscription"],"request": {"output_format": "jpeg", "output_compression": 90}}]