[codex] Add Nano Banana image models

README.md

@@ -140,26 +140,39 @@ Generated files are written through the `AssetStore` protocol to the
 references are stored in Postgres. An S3-compatible store is needed before deploying
 multiple application instances.
 
-## OpenAI Image Playground
+## Image Playground
 
-To enable real GPT Image 2 generation while leaving Etsy and Printful mocked:
+To enable real image generation while leaving Etsy and Printful mocked:
 
 ```text
 INTEGRATIONS_ENABLED=true
 ETSY_MODE=mock
 PRINTFUL_MODE=mock
 OPENAI_MODE=real
 OPENAI_API_KEY=...
+GEMINI_API_KEY=...
+IMAGE_GENERATION_MODEL=gpt-image-2
 OPENAI_IMAGE_MODEL=gpt-image-2
 OPENAI_IMAGE_QUALITY=medium
 OPENAI_IMAGE_WIDTH=1024
 OPENAI_IMAGE_HEIGHT=1024
 OPENAI_IMAGE_MAX_COST_USD=0.50
 ```
 
-The Design Studio stores every prompt and outcome in Postgres. A request is rejected
-before generation when its configured output-cost estimate exceeds the per-image
-limit. Dollar amounts are estimates because the generation response reports token
+The Design Studio can generate with OpenAI `gpt-image-2` or Google Gemini image
+models: `gemini-3.1-flash-image`, `gemini-3-pro-image`, and
+`gemini-2.5-flash-image`. `GOOGLE_API_KEY` is also accepted and takes precedence
+over `GEMINI_API_KEY`. Gemini sizes are limited to documented API sizes and are
+rejected before generation when unsupported. For now, Gemini REST generation is
+limited to 1024x1024 because the live API rejects the optional sizing config
+shown in the current docs.
+
+Create a Gemini key from Google AI Studio:
+https://aistudio.google.com/app/apikey
+
+The Design Studio stores every prompt and outcome in Postgres. A request is
+rejected before generation when its configured output-cost estimate exceeds the
+per-image limit. Dollar amounts are estimates because provider responses report
 usage rather than the final billed amount.
 
 ## Trend Research And Reports

src/ecommerce_agent/api/posting_routes.py

@@ -2,7 +2,7 @@
 from typing import Annotated, Any, cast
 
 import httpx
-from fastapi import APIRouter, Depends, Request, status
+from fastapi import APIRouter, Depends, HTTPException, Request, status
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.orm import selectinload
@@ -36,6 +36,7 @@
 from ecommerce_agent.pipeline.errors import ConflictError, NotFoundError
 from ecommerce_agent.services import ServiceContainer
 from ecommerce_agent.services.etsy import EtsyOutcomeUnknown
+from ecommerce_agent.services.image_models import gemini_image_size, provider_for_model
 from ecommerce_agent.services.printful import PrintfulError
 
 router = APIRouter(prefix="/posting", tags=["posting"])
@@ -291,6 +292,11 @@ async def generate_lifestyle_image(
             "The Printful mockup could not be downloaded for AI image generation."
         ) from error
     content_type = response.headers.get("content-type", "image/jpeg").split(";", 1)[0]
+    settings = request.app.state.settings
+    model = body.model or settings.image_generation_model
+    width = body.width or settings.openai_image_width
+    height = body.height or settings.openai_image_height
+    _validate_image_model(model, width=width, height=height, settings=settings)
     generated = await services.image.generate(
         ImageGenerationRequest(
             prompt=(
@@ -299,8 +305,9 @@ async def generate_lifestyle_image(
                 f"possible. {body.prompt}"
             ),
             product_concept_id=f"posting-{item.id}",
-            width=request.app.state.settings.openai_image_width,
-            height=request.app.state.settings.openai_image_height,
+            model=model,
+            width=width,
+            height=height,
             references=(
                 ImageReference(
                     filename="printful-reference.jpg",
@@ -320,14 +327,42 @@ async def generate_lifestyle_image(
         approved=False,
         ai_drift_acknowledged=False,
         alt_text=item.image_alt_text,
-        metadata_={"warning": "AI image may drift from the exact product design."},
+        metadata_={
+            "warning": "AI image may drift from the exact product design.",
+            "model": generated.model,
+            "provider": generated.provider,
+            "generation_metadata": generated.metadata,
+        },
     )
     session.add(image)
     await session.commit()
     await session.refresh(image)
     return _image_view(image)
 
 
+def _validate_image_model(model: str, *, width: int, height: int, settings: Any) -> None:
+    provider_name = provider_for_model(model, openai_model=settings.openai_image_model)
+    if provider_name is None:
+        raise HTTPException(status_code=422, detail=f"Unsupported image model: {model}.")
+    if provider_name == "gemini" and gemini_image_size(model, width, height) is None:
+        raise HTTPException(
+            status_code=422,
+            detail=(
+                f"{model} does not support {width}x{height}. "
+                "Choose one of the documented Gemini image sizes."
+            ),
+        )
+    if settings.openai_mode == "real":
+        if provider_name == "openai" and settings.openai_api_key is None:
+            raise HTTPException(
+                status_code=422, detail="OpenAI image generation is not configured."
+            )
+        if provider_name == "gemini" and not (settings.gemini_api_key or settings.google_api_key):
+            raise HTTPException(
+                status_code=422, detail="Gemini image generation is not configured."
+            )
+
+
 @router.patch("/images/{image_id}")
 async def update_listing_image(
     image_id: uuid.UUID,

src/ecommerce_agent/api/routes.py

@@ -15,6 +15,7 @@
     DecisionRequest,
     ImageGenerationCreateRequest,
     ImageGenerationUpdateRequest,
+    ProductWorkflowRequest,
     PublishRequest,
     RunResponse,
 )
@@ -44,6 +45,12 @@
 from ecommerce_agent.pipeline.events import log_event
 from ecommerce_agent.pipeline.transitions import advance, require_status
 from ecommerce_agent.services import ServiceContainer
+from ecommerce_agent.services.image_models import (
+    gemini_image_size,
+    image_model_options,
+    is_gemini_image_model,
+    provider_for_model,
+)
 from ecommerce_agent.services.openai_images import (
     SUPPORTED_OUTPUT_FORMATS,
     OpenAIImageError,
@@ -56,7 +63,7 @@
 ServicesDep = Annotated[ServiceContainer, Depends(get_services)]
 MAX_REFERENCE_BYTES = 10 * 1024 * 1024
 MAX_REFERENCE_PIXELS = 25_000_000
-MAX_IMAGE_PIXELS = 4_194_304
+MAX_IMAGE_PIXELS = 18_000_000
 REFERENCE_CONTENT_TYPES = {
     "PNG": "image/png",
     "JPEG": "image/jpeg",
@@ -68,6 +75,7 @@
 class ParsedImageRequest:
     prompt: str
     title: str
+    model: str | None
     width: int | None
     height: int | None
     output_format: str
@@ -80,6 +88,37 @@ async def health() -> dict[str, str]:
     return {"status": "ok"}
 
 
+@router.get("/image-models")
+async def list_image_models(request: Request) -> list[dict[str, Any]]:
+    settings = request.app.state.settings
+    mock_mode = settings.openai_mode != "real"
+    options = image_model_options(
+        openai_model=settings.openai_image_model,
+        openai_available=mock_mode or settings.openai_api_key is not None,
+        gemini_available=mock_mode
+        or settings.gemini_api_key is not None
+        or settings.google_api_key is not None,
+    )
+    return [
+        {
+            "model": option.model,
+            "label": option.label,
+            "provider": option.provider,
+            "available": option.available,
+            "sizes": [
+                {
+                    "width": size.width,
+                    "height": size.height,
+                    "aspect_ratio": size.aspect_ratio,
+                    "image_size": size.image_size,
+                }
+                for size in option.sizes
+            ],
+        }
+        for option in options
+    ]
+
+
 @router.post("/image-generations", status_code=status.HTTP_201_CREATED)
 async def create_image_generation(
     request: Request,
@@ -124,6 +163,7 @@ async def edit_image_generation(
     parsed = ParsedImageRequest(
         prompt=parsed.prompt,
         title=parent.title,
+        model=parsed.model,
         width=parsed.width,
         height=parsed.height,
         output_format=parsed.output_format,
@@ -148,25 +188,30 @@ async def _execute_image_generation(
     services: ServiceContainer,
 ) -> dict[str, Any]:
     settings = request.app.state.settings
+    model = parsed.model or settings.image_generation_model
     width = parsed.width or settings.openai_image_width
     height = parsed.height or settings.openai_image_height
+    _validate_image_model(model, width=width, height=height, settings=settings)
     is_real = settings.openai_mode == "real"
+    provider_name = provider_for_model(model, openai_model=settings.openai_image_model)
     try:
-        output_estimate = (
-            estimate_image_output_cost(
-                model=settings.openai_image_model,
+        if not is_real:
+            output_estimate = Decimal("0")
+            reference_reserve = Decimal("0")
+        elif provider_name == "gemini":
+            size = gemini_image_size(model, width, height)
+            output_estimate = size.estimated_cost_usd if size is not None else Decimal("0")
+            reference_reserve = Decimal("0")
+        else:
+            output_estimate = estimate_image_output_cost(
+                model=model,
                 quality=settings.openai_image_quality,
                 width=width,
                 height=height,
             )
-            if is_real
-            else Decimal("0")
-        )
-        reference_reserve = (
-            Decimal(str(settings.openai_image_reference_cost_reserve_usd)) * len(parsed.references)
-            if is_real
-            else Decimal("0")
-        )
+            reference_reserve = Decimal(
+                str(settings.openai_image_reference_cost_reserve_usd)
+            ) * len(parsed.references)
         estimated_cost = output_estimate + reference_reserve
     except OpenAIImageError:
         estimated_cost = Decimal("0")
@@ -178,9 +223,9 @@ async def _execute_image_generation(
         status=ImageGenerationStatus.PENDING,
         width=width,
         height=height,
-        provider="openai" if is_real else "mock",
-        model=settings.openai_image_model if is_real else "mock-image-v1",
-        quality=settings.openai_image_quality,
+        provider=provider_name if is_real and provider_name is not None else "mock",
+        model=model if is_real else "mock-image-v1",
+        quality="standard" if provider_name == "gemini" else settings.openai_image_quality,
         output_format=parsed.output_format,
         background_removal_friendly=parsed.background_removal_friendly,
         estimated_cost_usd=estimated_cost,
@@ -193,6 +238,7 @@ async def _execute_image_generation(
             ImageGenerationRequest(
                 prompt=generation.prompt,
                 product_concept_id=str(generation.id),
+                model=model,
                 width=generation.width,
                 height=generation.height,
                 references=parsed.references,
@@ -235,6 +281,7 @@ async def _parse_image_request(request: Request, *, max_references: int) -> Pars
     content_type = request.headers.get("content-type", "")
     prompt_value: Any
     title_value: Any
+    model_value: Any
     width_value: Any
     height_value: Any
     output_format_value: Any
@@ -244,14 +291,15 @@ async def _parse_image_request(request: Request, *, max_references: int) -> Pars
         form = await request.form()
         prompt_value = form.get("prompt")
         title_value = form.get("title", "Untitled design")
+        model_value = form.get("model")
         width_value = form.get("width")
         height_value = form.get("height")
         output_format_value = form.get("output_format", "png")
         background_value: Any = form.get("background_removal_friendly", "true")
         uploads = [
             value
             for key, value in form.multi_items()
-            if key == "reference_images" and isinstance(value, UploadFile)
+            if key == "reference_images" and isinstance(value, UploadFile) and value.filename
         ]
         if len(uploads) > max_references:
             raise HTTPException(
@@ -268,6 +316,7 @@ async def _parse_image_request(request: Request, *, max_references: int) -> Pars
             ) from error
         prompt_value = body.prompt
         title_value = body.title
+        model_value = body.model
         width_value = body.width
         height_value = body.height
         output_format_value = body.output_format
@@ -299,6 +348,7 @@ async def _parse_image_request(request: Request, *, max_references: int) -> Pars
     return ParsedImageRequest(
         prompt=prompt,
         title=title,
+        model=str(model_value).strip() if model_value else None,
         width=width,
         height=height,
         output_format=output_format,
@@ -369,11 +419,34 @@ def _parse_dimension(value: Any, label: str) -> int | None:
         dimension = int(value)
     except (TypeError, ValueError) as error:
         raise HTTPException(status_code=422, detail=f"Invalid image {label}.") from error
-    if dimension <= 0 or dimension > 4096:
+    if dimension <= 0 or dimension > 5504:
         raise HTTPException(status_code=422, detail=f"Invalid image {label}.")
     return dimension
 
 
+def _validate_image_model(model: str, *, width: int, height: int, settings: Any) -> None:
+    provider_name = provider_for_model(model, openai_model=settings.openai_image_model)
+    if provider_name is None:
+        raise HTTPException(status_code=422, detail=f"Unsupported image model: {model}.")
+    if settings.openai_mode == "real":
+        if provider_name == "openai" and settings.openai_api_key is None:
+            raise HTTPException(
+                status_code=422, detail="OpenAI image generation is not configured."
+            )
+        if provider_name == "gemini" and not (settings.gemini_api_key or settings.google_api_key):
+            raise HTTPException(
+                status_code=422, detail="Gemini image generation is not configured."
+            )
+    if is_gemini_image_model(model) and gemini_image_size(model, width, height) is None:
+        raise HTTPException(
+            status_code=422,
+            detail=(
+                f"{model} does not support {width}x{height}. "
+                "Choose one of the documented Gemini image sizes."
+            ),
+        )
+
+
 def _extension(output_format: str) -> str:
     return "jpg" if output_format == "jpeg" else output_format
 
@@ -454,14 +527,33 @@ async def start_research(session: SessionDep, services: ServicesDep) -> RunRespo
     response_model=RunResponse,
     status_code=status.HTTP_201_CREATED,
 )
-async def start_product(product_id: uuid.UUID, session: SessionDep) -> RunResponse:
+async def start_product(
+    product_id: uuid.UUID,
+    request: Request,
+    session: SessionDep,
+    body: ProductWorkflowRequest | None = None,
+) -> RunResponse:
     product = await session.get(Product, product_id)
     if product is None:
         raise NotFoundError("Product concept not found.")
+    image_model = body.image_model if body is not None else None
+    if image_model:
+        settings = request.app.state.settings
+        _validate_image_model(
+            image_model,
+            width=settings.openai_image_width,
+            height=settings.openai_image_height,
+            settings=settings,
+        )
     require_status(product, ProductStatus.CONCEPT)
     advance(product, ProductStatus.GENERATING_DESIGN)
     await log_event(session, "product", product.id, "start", "Product workflow started.")
-    await enqueue(session, "generate_design", product.id)
+    await enqueue(
+        session,
+        "generate_design",
+        product.id,
+        {"image_model": image_model} if image_model else None,
+    )
     await session.commit()
     return RunResponse(id=product.id, subject_type="product", status=product.status.value)