refactor!: Reduce dependencies of core, remove image from URL (#4919)

* Reduce dependencies of core

* Update lock

* add package to test deps

* Update lock

* aio console dep

* Update lock

python/packages/autogen-agentchat/pyproject.toml

@@ -16,6 +16,7 @@ classifiers = [
 ]
 dependencies = [
     "autogen-core==0.4.0.dev13",
+    "aioconsole>=0.8.1"
 ]
 
 [tool.ruff]

python/packages/autogen-core/pyproject.toml

@@ -15,24 +15,20 @@ classifiers = [
     "Operating System :: OS Independent",
 ]
 dependencies = [
-    "openai>=1.3",
     "pillow>=11.0.0",
-    "aioconsole>=0.8.1",
-    "aiohttp>=3.10.10",
-    "typing-extensions",
+    "typing-extensions>=4.0.0",
     "pydantic<3.0.0,>=2.10.0",
     "protobuf~=4.25.1",
-    "tiktoken>=0.8.0",
-    "opentelemetry-api~=1.27.0",
-    "asyncio_atexit",
+    "opentelemetry-api>=1.27.0",
     "jsonref~=1.1.0",
 ]
 
 
 [dependency-groups]
 dev = [
-    "autogen_test_utils",
     "aiofiles",
+    "asyncio_atexit",
+    "autogen_test_utils",
     "azure-identity",
     "chess",
     "colorama",
@@ -46,6 +42,7 @@ dev = [
     "llama-index",
     "markdownify",
     "nbqa",
+    "opentelemetry-sdk>=1.27.0",
     "pip",
     "polars",
     "python-dotenv",
@@ -55,12 +52,11 @@ dev = [
     "textual-imageview",
     "textual",
     "types-aiofiles",
+    "types-docker",
     "types-pillow",
     "types-protobuf",
     "types-requests",
-    "types-docker",
     "wikipedia",
-    "opentelemetry-sdk>=1.27.0",
 
     # Documentation
     "myst-nb==1.1.2",

python/packages/autogen-core/src/autogen_core/_image.py

@@ -4,17 +4,41 @@
 import re
 from io import BytesIO
 from pathlib import Path
-from typing import Any, cast
+from typing import Any, Dict, cast
 
-import aiohttp
-from openai.types.chat import ChatCompletionContentPartImageParam
 from PIL import Image as PILImage
 from pydantic import GetCoreSchemaHandler, ValidationInfo
 from pydantic_core import core_schema
 from typing_extensions import Literal
 
 
 class Image:
+    """Represents an image.
+
+
+    Example:
+
+        Loading an image from a URL:
+
+        .. code-block:: python
+
+            from autogen_core import Image
+            from PIL import Image as PILImage
+            import aiohttp
+            import asyncio
+
+
+            async def from_url(url: str) -> Image:
+                async with aiohttp.ClientSession() as session:
+                    async with session.get(url) as response:
+                        content = await response.read()
+                        return Image.from_pil(PILImage.open(content))
+
+
+            image = asyncio.run(from_url("https://example.com/image"))
+
+    """
+
     def __init__(self, image: PILImage.Image):
         self.image: PILImage.Image = image.convert("RGB")
 
@@ -31,13 +55,6 @@ def from_uri(cls, uri: str) -> Image:
         base64_data = re.sub(r"data:image/(?:png|jpeg);base64,", "", uri)
         return cls.from_base64(base64_data)
 
-    @classmethod
-    async def from_url(cls, url: str) -> Image:
-        async with aiohttp.ClientSession() as session:
-            async with session.get(url) as response:
-                content = await response.read()
-                return cls(PILImage.open(content))
-
     @classmethod
     def from_base64(cls, base64_str: str) -> Image:
         return cls(PILImage.open(BytesIO(base64.b64decode(base64_str))))
@@ -60,7 +77,9 @@ def _repr_html_(self) -> str:
     def data_uri(self) -> str:
         return _convert_base64_to_data_uri(self.to_base64())
 
-    def to_openai_format(self, detail: Literal["auto", "low", "high"] = "auto") -> ChatCompletionContentPartImageParam:
+    # Returns openai.types.chat.ChatCompletionContentPartImageParam, which is a TypedDict
+    # We don't use the explicit type annotation so that we can avoid a dependency on the OpenAI Python SDK in this package.
+    def to_openai_format(self, detail: Literal["auto", "low", "high"] = "auto") -> Dict[str, Any]:
         return {"type": "image_url", "image_url": {"url": self.data_uri, "detail": detail}}
 
     @classmethod

python/packages/autogen-ext/src/autogen_ext/models/openai/_openai_client.py

@@ -51,6 +51,7 @@
 from openai.types.chat import (
     ChatCompletion,
     ChatCompletionAssistantMessageParam,
+    ChatCompletionContentPartImageParam,
     ChatCompletionContentPartParam,
     ChatCompletionContentPartTextParam,
     ChatCompletionMessageParam,
@@ -154,7 +155,7 @@ def user_message_to_oai(message: UserMessage) -> ChatCompletionUserMessageParam:
             elif isinstance(part, Image):
                 # TODO: support url based images
                 # TODO: support specifying details
-                parts.append(part.to_openai_format())
+                parts.append(cast(ChatCompletionContentPartImageParam, part.to_openai_format()))
             else:
                 raise ValueError(f"Unknown content type: {part}")
         return ChatCompletionUserMessageParam(