Refactor models

examples/fastapi-vite/backend/agent.py

@@ -10,7 +10,7 @@
 
 import ai
 
-MODEL = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+MODEL = ai.ai_gateway("anthropic/claude-sonnet-4")
 
 
 @ai.tool
@@ -42,7 +42,9 @@ async def graph(context: ai.Context) -> AsyncGenerator[ai.Message]:
         if not tool_calls:
             return
 
-        results = await asyncio.gather(*(_execute_with_approval(tc) for tc in tool_calls))
+        results = await asyncio.gather(
+            *(_execute_with_approval(tc) for tc in tool_calls)
+        )
         yield ai.tool_message(*results)
 
 

examples/multiagent-textual/server.py

@@ -58,7 +58,7 @@ class Approval(pydantic.BaseModel):
 # Model
 # ---------------------------------------------------------------------------
 
-MODEL = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+MODEL = ai.ai_gateway("anthropic/claude-sonnet-4")
 
 
 # ---------------------------------------------------------------------------

examples/samples/agent_custom_loop.py

@@ -19,7 +19,7 @@ async def get_population(city: str) -> int:
 
 
 async def main() -> None:
-    model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+    model = ai.ai_gateway("anthropic/claude-sonnet-4")
 
     tools = [get_weather, get_population]
     my_agent = ai.agent(tools=tools)

examples/samples/agent_hooks.py

@@ -26,7 +26,7 @@ async def contact_mothership(query: str) -> str:
 
 
 async def main() -> None:
-    model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+    model = ai.ai_gateway("anthropic/claude-sonnet-4")
 
     my_agent = ai.agent(tools=[contact_mothership])
 

examples/samples/agent_hooks_serverless.py

@@ -31,7 +31,7 @@ async def delete_file(path: str) -> str:
 
 
 async def main() -> None:
-    model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+    model = ai.ai_gateway("anthropic/claude-sonnet-4")
 
     my_agent = ai.agent(tools=[delete_file])
 

examples/samples/check_connection.py

@@ -1,13 +1,19 @@
-"""Check connection — verify credentials and model availability for all providers."""
+"""Check connection and list models — verify credentials and model availability."""
 
 import asyncio
 
 import ai
 
 MODELS = [
-    ai.model("ai-gateway", "anthropic/claude-sonnet-4"),
-    ai.model("anthropic", "claude-sonnet-4-20250514"),
-    ai.model("openai", "gpt-5.4-mini"),
+    ai.ai_gateway("anthropic/claude-sonnet-4"),
+    ai.anthropic("claude-sonnet-4-20250514"),
+    ai.openai("gpt-5.4-mini"),
+]
+
+PROVIDERS = [
+    ("ai_gateway", ai.ai_gateway),
+    ("anthropic", ai.anthropic),
+    ("openai", ai.openai),
 ]
 
 
@@ -20,9 +26,23 @@ async def _check(model: ai.Model) -> None:
     print(f"  {status}  {model.provider}/{model.id}")
 
 
+async def _list_models(name: str, provider: object) -> None:
+    try:
+        ids: list[str] = await provider.list()  # type: ignore[union-attr]
+        print(f"  {name}: {len(ids)} models")
+        for mid in ids:
+            print(f"    - {mid}")
+    except Exception as exc:
+        print(f"  {name}: [ERR] {exc}")
+
+
 async def main() -> None:
     print("Checking connections...\n")
     await asyncio.gather(*[_check(m) for m in MODELS])
+
+    print("\nListing models...\n")
+    await asyncio.gather(*[_list_models(n, p) for n, p in PROVIDERS])
+
     print()
 
 

examples/samples/explicit_client.py

@@ -5,21 +5,21 @@
 
 import ai
 
-model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
-
 # Explicit client — useful for custom auth, proxies, or self-hosted gateways.
 client = ai.Client(
     base_url="https://ai-gateway.vercel.sh/v3/ai",
     api_key=os.environ["AI_GATEWAY_API_KEY"],
     headers={"X-Custom-Header": "example"},
 )
 
+model = ai.ai_gateway("anthropic/claude-sonnet-4", client=client)
+
 messages = [ai.user_message("Hello!")]
 
 
 async def main() -> None:
     try:
-        async for msg in await ai.models.stream(model, messages, client=client):
+        async for msg in await ai.models.stream(model, messages):
             if msg.text_delta:
                 print(msg.text_delta, end="", flush=True)
         print()

examples/samples/image_generation.py

@@ -6,7 +6,7 @@
 
 import ai
 
-model = ai.model("ai-gateway", "google/imagen-4.0-generate-001")
+model = ai.ai_gateway("google/imagen-4.0-generate-001")
 
 messages = [
     ai.user_message(

examples/samples/middleware_simple.py

@@ -88,7 +88,7 @@ async def get_population(city: str) -> int:
 
 
 async def main() -> None:
-    model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+    model = ai.ai_gateway("anthropic/claude-sonnet-4")
 
     my_agent = ai.agent(tools=[get_weather, get_population])
 

examples/samples/stream.py

@@ -4,7 +4,7 @@
 
 import ai
 
-model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+model = ai.ai_gateway("anthropic/claude-sonnet-4")
 
 messages = [
     ai.system_message("Be concise."),

examples/samples/video_generation.py

@@ -6,7 +6,7 @@
 
 import ai
 
-model = ai.model("ai-gateway", "google/veo-3.0-generate-001")
+model = ai.ai_gateway("google/veo-3.0-generate-001")
 
 messages = [
     ai.user_message(

examples/temporal-direct/main.py

@@ -96,7 +96,7 @@ class LLMResult:
 @temporalio.activity.defn
 async def llm_call_activity(params: LLMParams) -> LLMResult:
     """Call the LLM, drain the stream, return the final message."""
-    model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+    model = ai.ai_gateway("anthropic/claude-sonnet-4")
     messages = [ai.Message.model_validate(m) for m in params.messages]
     tools = [ai.ToolSchema(return_type=None, **t) for t in params.tool_schemas]
 
@@ -166,7 +166,7 @@ async def run_tool(tc: ai.ToolCallPart) -> ai.ToolResultPart:
 class WeatherWorkflow:
     @temporalio.workflow.run
     async def run(self, user_query: str) -> str:
-        model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+        model = ai.ai_gateway("anthropic/claude-sonnet-4")
         messages: list[ai.Message] = [
             ai.system_message(
                 "Answer questions using the weather and population tools."

examples/temporal-middleware/main.py

@@ -113,7 +113,7 @@ class LLMResult:
 @temporalio.activity.defn
 async def llm_call_activity(params: LLMParams) -> LLMResult:
     """Call the LLM, drain the stream, return the final message."""
-    model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+    model = ai.ai_gateway("anthropic/claude-sonnet-4")
     messages = [ai.Message.model_validate(m) for m in params.messages]
     tools = [ai.ToolSchema(return_type=None, **t) for t in params.tool_schemas]
 
@@ -194,7 +194,7 @@ async def wrap_tool(
 class WeatherWorkflow:
     @temporalio.workflow.run
     async def run(self, user_query: str) -> str:
-        model = ai.model("ai-gateway", "anthropic/claude-sonnet-4")
+        model = ai.ai_gateway("anthropic/claude-sonnet-4")
         messages: list[ai.Message] = [
             ai.system_message(
                 "Answer questions using the weather and population tools."

src/ai/__init__.py

@@ -20,12 +20,14 @@
     Client,
     ImageParams,
     Model,
-    ModelCost,
+    Provider,
     StreamResult,
     VideoParams,
+    ai_gateway,
+    anthropic,
     check_connection,
     generate,
-    model,
+    openai,
     stream,
 )
 
@@ -83,17 +85,20 @@
     "thinking",
     # Models (from models/)
     "Model",
-    "ModelCost",
+    "Provider",
     "ImageParams",
     "VideoParams",
     "Client",
     "StreamResult",
     "StreamResultLike",
     "check_connection",
-    "model",
-    "models",
     "stream",
     "generate",
+    "models",
+    # Provider factories
+    "openai",
+    "anthropic",
+    "ai_gateway",
     # Agents — primary API
     "Agent",
     "agent",

src/ai/middleware.py

@@ -38,7 +38,6 @@
 
 if TYPE_CHECKING:
     from .agents.agent import Tool
-    from .models.core.client import Client
     from .models.core.model import Model
 
 
@@ -50,7 +49,6 @@ class ModelContext:
     messages: list[messages_.Message]
     tools: Sequence[tools_.ToolLike] | None
     output_type: type[pydantic.BaseModel] | None
-    client: Client | None
     kwargs: dict[str, Any]
 
     def __post_init__(self) -> None:
@@ -67,7 +65,6 @@ class GenerateContext:
     model: Model
     messages: list[messages_.Message]
     params: Any
-    client: Client | None = None
 
     def __post_init__(self) -> None:
         object.__setattr__(self, "messages", list(self.messages))