feat: add OpenAI as provider for text and image generation
- Add openai_text.py: text generation via OpenAI chat completions API (gpt-4o, gpt-4o-mini, gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, o3-mini) - Add openai_image.py: image generation via OpenAI images API (gpt-image-1 with reference image support, dall-e-3, dall-e-2) - Refactor builder provider dispatch from TargetType to model-name index to support multiple providers per target type - Fix circular import between config.py and providers/__init__.py using TYPE_CHECKING guard - Fix stale default model assertions in tests - Add openai>=1.0.0 dependency
This commit is contained in:
parent
d0dac5b1bf
commit
870023865d
GPG key ID:
A793722F9933C1A5
9 changed files with 571 additions and 58 deletions
|
|
@ -4,10 +4,13 @@ from __future__ import annotations
|
|||
|
||||
import abc
|
||||
from pathlib import Path
|
||||
from typing import TYPE_CHECKING
|
||||
|
||||
from bulkgen.config import TargetConfig
|
||||
from bulkgen.providers.models import ModelInfo
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from bulkgen.config import TargetConfig
|
||||
|
||||
|
||||
class Provider(abc.ABC):
|
||||
"""Abstract base for generation providers."""
|
||||
|
|
|
|||
194
bulkgen/providers/openai_image.py
Normal file
194
bulkgen/providers/openai_image.py
Normal file
|
|
@ -0,0 +1,194 @@
|
|||
"""OpenAI image generation provider."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
from pathlib import Path
|
||||
from typing import Literal, override
|
||||
|
||||
import httpx
|
||||
from openai import AsyncOpenAI
|
||||
from openai.types.images_response import ImagesResponse
|
||||
|
||||
from bulkgen.config import TargetConfig
|
||||
from bulkgen.providers import Provider
|
||||
from bulkgen.providers.models import Capability, ModelInfo
|
||||
|
||||
_SIZE = Literal[
|
||||
"auto",
|
||||
"1024x1024",
|
||||
"1024x1536",
|
||||
"1536x1024",
|
||||
"1024x1792",
|
||||
"1792x1024",
|
||||
"256x256",
|
||||
"512x512",
|
||||
]
|
||||
|
||||
_VALID_SIZES: frozenset[str] = frozenset(
|
||||
{
|
||||
"auto",
|
||||
"1024x1024",
|
||||
"1024x1536",
|
||||
"1536x1024",
|
||||
"1024x1792",
|
||||
"1792x1024",
|
||||
"256x256",
|
||||
"512x512",
|
||||
}
|
||||
)
|
||||
|
||||
|
||||
def _build_size(width: int | None, height: int | None) -> _SIZE | None:
|
||||
"""Convert width/height to an OpenAI size string, or *None* for the default."""
|
||||
if width is None and height is None:
|
||||
return None
|
||||
w = width or 1024
|
||||
h = height or 1024
|
||||
size = f"{w}x{h}"
|
||||
if size not in _VALID_SIZES:
|
||||
msg = f"Unsupported OpenAI image size '{size}'. Valid sizes: {', '.join(sorted(_VALID_SIZES))}"
|
||||
raise ValueError(msg)
|
||||
return size # pyright: ignore[reportReturnType]
|
||||
|
||||
|
||||
class OpenAIImageProvider(Provider):
|
||||
"""Generates images via the OpenAI API."""
|
||||
|
||||
_api_key: str
|
||||
|
||||
def __init__(self, api_key: str) -> None:
|
||||
self._api_key = api_key
|
||||
|
||||
@staticmethod
|
||||
@override
|
||||
def get_provided_models() -> list[ModelInfo]:
|
||||
return [
|
||||
ModelInfo(
|
||||
name="gpt-image-1",
|
||||
provider="OpenAI",
|
||||
type="image",
|
||||
capabilities=[
|
||||
Capability.TEXT_TO_IMAGE,
|
||||
Capability.REFERENCE_IMAGES,
|
||||
],
|
||||
),
|
||||
ModelInfo(
|
||||
name="dall-e-3",
|
||||
provider="OpenAI",
|
||||
type="image",
|
||||
capabilities=[Capability.TEXT_TO_IMAGE],
|
||||
),
|
||||
ModelInfo(
|
||||
name="dall-e-2",
|
||||
provider="OpenAI",
|
||||
type="image",
|
||||
capabilities=[Capability.TEXT_TO_IMAGE],
|
||||
),
|
||||
]
|
||||
|
||||
@override
|
||||
async def generate(
|
||||
self,
|
||||
target_name: str,
|
||||
target_config: TargetConfig,
|
||||
resolved_prompt: str,
|
||||
resolved_model: ModelInfo,
|
||||
project_dir: Path,
|
||||
) -> None:
|
||||
output_path = project_dir / target_name
|
||||
size = _build_size(target_config.width, target_config.height)
|
||||
|
||||
async with AsyncOpenAI(api_key=self._api_key) as client:
|
||||
if target_config.reference_images:
|
||||
response = await _generate_edit(
|
||||
client,
|
||||
resolved_prompt,
|
||||
resolved_model.name,
|
||||
target_config.reference_images,
|
||||
project_dir,
|
||||
size,
|
||||
)
|
||||
else:
|
||||
response = await _generate_new(
|
||||
client,
|
||||
resolved_prompt,
|
||||
resolved_model.name,
|
||||
size,
|
||||
)
|
||||
|
||||
image_data = _extract_image_bytes(response, resolved_model.name)
|
||||
_ = output_path.write_bytes(image_data)
|
||||
|
||||
|
||||
async def _generate_new(
|
||||
client: AsyncOpenAI,
|
||||
prompt: str,
|
||||
model: str,
|
||||
size: _SIZE | None,
|
||||
) -> ImagesResponse:
|
||||
"""Generate a new image from a text prompt."""
|
||||
if size is not None:
|
||||
return await client.images.generate(
|
||||
prompt=prompt,
|
||||
model=model,
|
||||
n=1,
|
||||
response_format="b64_json",
|
||||
size=size,
|
||||
)
|
||||
return await client.images.generate(
|
||||
prompt=prompt,
|
||||
model=model,
|
||||
n=1,
|
||||
response_format="b64_json",
|
||||
)
|
||||
|
||||
|
||||
async def _generate_edit(
|
||||
client: AsyncOpenAI,
|
||||
prompt: str,
|
||||
model: str,
|
||||
reference_images: list[str],
|
||||
project_dir: Path,
|
||||
size: _SIZE | None,
|
||||
) -> ImagesResponse:
|
||||
"""Generate an image using a reference image via the edits endpoint."""
|
||||
ref_path = project_dir / reference_images[0]
|
||||
image_bytes = ref_path.read_bytes()
|
||||
|
||||
if size is not None:
|
||||
return await client.images.edit(
|
||||
image=image_bytes,
|
||||
prompt=prompt,
|
||||
model=model,
|
||||
n=1,
|
||||
response_format="b64_json",
|
||||
size=size, # pyright: ignore[reportArgumentType]
|
||||
)
|
||||
return await client.images.edit(
|
||||
image=image_bytes,
|
||||
prompt=prompt,
|
||||
model=model,
|
||||
n=1,
|
||||
response_format="b64_json",
|
||||
)
|
||||
|
||||
|
||||
def _extract_image_bytes(response: ImagesResponse, model: str) -> bytes:
|
||||
"""Extract image bytes from an OpenAI images response."""
|
||||
if not response.data:
|
||||
msg = f"OpenAI {model} returned no images"
|
||||
raise RuntimeError(msg)
|
||||
|
||||
image = response.data[0]
|
||||
|
||||
if image.b64_json is not None:
|
||||
return base64.b64decode(image.b64_json)
|
||||
|
||||
if image.url is not None:
|
||||
resp = httpx.get(image.url)
|
||||
_ = resp.raise_for_status()
|
||||
return resp.content
|
||||
|
||||
msg = f"OpenAI {model} returned neither b64_json nor url"
|
||||
raise RuntimeError(msg)
|
||||
169
bulkgen/providers/openai_text.py
Normal file
169
bulkgen/providers/openai_text.py
Normal file
|
|
@ -0,0 +1,169 @@
|
|||
"""OpenAI text generation provider."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import mimetypes
|
||||
from pathlib import Path
|
||||
from typing import override
|
||||
|
||||
from openai import AsyncOpenAI
|
||||
from openai.types.chat import (
|
||||
ChatCompletionContentPartImageParam,
|
||||
ChatCompletionContentPartParam,
|
||||
ChatCompletionContentPartTextParam,
|
||||
ChatCompletionUserMessageParam,
|
||||
)
|
||||
|
||||
from bulkgen.config import IMAGE_EXTENSIONS, TargetConfig
|
||||
from bulkgen.providers import Provider
|
||||
from bulkgen.providers.models import Capability, ModelInfo
|
||||
|
||||
|
||||
def _image_to_data_url(path: Path) -> str:
|
||||
"""Read an image file and return a ``data:`` URL with base64-encoded content."""
|
||||
mime = mimetypes.guess_type(path.name)[0] or "image/png"
|
||||
b64 = base64.b64encode(path.read_bytes()).decode("ascii")
|
||||
return f"data:{mime};base64,{b64}"
|
||||
|
||||
|
||||
class OpenAITextProvider(Provider):
|
||||
"""Generates text via the OpenAI API."""
|
||||
|
||||
_api_key: str
|
||||
|
||||
def __init__(self, api_key: str) -> None:
|
||||
self._api_key = api_key
|
||||
|
||||
@staticmethod
|
||||
@override
|
||||
def get_provided_models() -> list[ModelInfo]:
|
||||
return [
|
||||
ModelInfo(
|
||||
name="gpt-4o",
|
||||
provider="OpenAI",
|
||||
type="text",
|
||||
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
|
||||
),
|
||||
ModelInfo(
|
||||
name="gpt-4o-mini",
|
||||
provider="OpenAI",
|
||||
type="text",
|
||||
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
|
||||
),
|
||||
ModelInfo(
|
||||
name="gpt-4.1",
|
||||
provider="OpenAI",
|
||||
type="text",
|
||||
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
|
||||
),
|
||||
ModelInfo(
|
||||
name="gpt-4.1-mini",
|
||||
provider="OpenAI",
|
||||
type="text",
|
||||
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
|
||||
),
|
||||
ModelInfo(
|
||||
name="gpt-4.1-nano",
|
||||
provider="OpenAI",
|
||||
type="text",
|
||||
capabilities=[Capability.TEXT_GENERATION, Capability.VISION],
|
||||
),
|
||||
ModelInfo(
|
||||
name="o3-mini",
|
||||
provider="OpenAI",
|
||||
type="text",
|
||||
capabilities=[Capability.TEXT_GENERATION],
|
||||
),
|
||||
]
|
||||
|
||||
@override
|
||||
async def generate(
|
||||
self,
|
||||
target_name: str,
|
||||
target_config: TargetConfig,
|
||||
resolved_prompt: str,
|
||||
resolved_model: ModelInfo,
|
||||
project_dir: Path,
|
||||
) -> None:
|
||||
output_path = project_dir / target_name
|
||||
|
||||
all_input_names = list(target_config.inputs) + list(
|
||||
target_config.reference_images
|
||||
)
|
||||
|
||||
has_images = any(
|
||||
(project_dir / name).suffix.lower() in IMAGE_EXTENSIONS
|
||||
for name in all_input_names
|
||||
)
|
||||
|
||||
if has_images:
|
||||
message = _build_multimodal_message(
|
||||
resolved_prompt, all_input_names, project_dir
|
||||
)
|
||||
else:
|
||||
message = _build_text_message(resolved_prompt, all_input_names, project_dir)
|
||||
|
||||
async with AsyncOpenAI(api_key=self._api_key) as client:
|
||||
response = await client.chat.completions.create(
|
||||
model=resolved_model.name,
|
||||
messages=[message],
|
||||
)
|
||||
|
||||
if not response.choices:
|
||||
msg = f"OpenAI API returned no choices for target '{target_name}'"
|
||||
raise RuntimeError(msg)
|
||||
|
||||
content = response.choices[0].message.content
|
||||
if content is None:
|
||||
msg = f"OpenAI API returned empty content for target '{target_name}'"
|
||||
raise RuntimeError(msg)
|
||||
|
||||
_ = output_path.write_text(content)
|
||||
|
||||
|
||||
def _build_text_message(
|
||||
prompt: str,
|
||||
input_names: list[str],
|
||||
project_dir: Path,
|
||||
) -> ChatCompletionUserMessageParam:
|
||||
"""Build a plain-text message (no images)."""
|
||||
parts: list[str] = [prompt]
|
||||
for name in input_names:
|
||||
file_content = (project_dir / name).read_text()
|
||||
parts.append(f"\n--- Contents of {name} ---\n{file_content}")
|
||||
return {"role": "user", "content": "\n".join(parts)}
|
||||
|
||||
|
||||
def _build_multimodal_message(
|
||||
prompt: str,
|
||||
input_names: list[str],
|
||||
project_dir: Path,
|
||||
) -> ChatCompletionUserMessageParam:
|
||||
"""Build a multimodal message with text and image parts."""
|
||||
parts: list[ChatCompletionContentPartParam] = [
|
||||
ChatCompletionContentPartTextParam(type="text", text=prompt),
|
||||
]
|
||||
|
||||
for name in input_names:
|
||||
input_path = project_dir / name
|
||||
suffix = input_path.suffix.lower()
|
||||
|
||||
if suffix in IMAGE_EXTENSIONS:
|
||||
data_url = _image_to_data_url(input_path)
|
||||
parts.append(
|
||||
ChatCompletionContentPartImageParam(
|
||||
type="image_url",
|
||||
image_url={"url": data_url},
|
||||
)
|
||||
)
|
||||
else:
|
||||
file_content = input_path.read_text()
|
||||
parts.append(
|
||||
ChatCompletionContentPartTextParam(
|
||||
type="text",
|
||||
text=f"\n--- Contents of {name} ---\n{file_content}",
|
||||
)
|
||||
)
|
||||
|
||||
return {"role": "user", "content": parts}
|
||||
|
|
@ -9,8 +9,12 @@ def get_all_models() -> list[ModelInfo]:
|
|||
"""Return the merged list of models from all providers."""
|
||||
from bulkgen.providers.blackforest import BlackForestProvider
|
||||
from bulkgen.providers.mistral import MistralProvider
|
||||
from bulkgen.providers.openai_image import OpenAIImageProvider
|
||||
from bulkgen.providers.openai_text import OpenAITextProvider
|
||||
|
||||
return (
|
||||
MistralProvider.get_provided_models()
|
||||
+ BlackForestProvider.get_provided_models()
|
||||
+ OpenAITextProvider.get_provided_models()
|
||||
+ OpenAIImageProvider.get_provided_models()
|
||||
)
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue