fix api types (#116)

samuelcolvin · web-flow · commit b017559fbcbb · 2025-11-04T13:52:21.000-08:00
diff --git a/deploy/example.config.ts b/deploy/example.config.ts
@@ -37,7 +37,7 @@ export const config: Config<ProviderKeys> = {
   },
   // providers
   providers: {
-    // you would use this provider by using the model id `gateway:openai-chat/gpt-5` in Pydantic AI
+    // you would use this provider by using the model id `gateway/openai-chat:gpt-5` in Pydantic AI
     a: {
       // providerId decides on the logic used to process the request and response
       providerId: 'openai',
diff --git a/examples/ex_openai.py b/examples/ex_openai.py
@@ -6,9 +6,9 @@
 logfire.instrument_httpx(capture_all=True)
 
 client = OpenAI(
-    api_key='VOE4JMpVGr71RgvEEidPCXd4ov42L24ODw9q5RI7uYc',
+    api_key='_api_key_',
     base_url='http://localhost:8787/openai',
-    # base_url='https://pydantic-ai-gateway.pydantic.workers.dev/openai',
+    # base_url='https://gateway.pydantic.dev/proxy/openai',
 )
 
 response = client.responses.create(
diff --git a/examples/pai_anthropic.py b/examples/pai_anthropic.py
@@ -26,7 +26,7 @@ def validate_dob(cls, v: date) -> date:
 
 
 person_agent = Agent(
-    'gateway:anthropic/claude-sonnet-4-0',
+    'gateway/anthropic:claude-sonnet-4-0',
     output_type=Person,
     instructions='Extract information about the person',
     model_settings=AnthropicModelSettings(max_tokens=1024),
diff --git a/examples/pai_anthropic_builtin.py b/examples/pai_anthropic_builtin.py
@@ -8,7 +8,7 @@
 
 
 person_agent = Agent(
-    'gateway:anthropic/claude-sonnet-4-0',
+    'gateway/anthropic:claude-sonnet-4-0',
     instructions='Use the code execution tool to execute the code',
     model_settings=AnthropicModelSettings(max_tokens=1024),
     builtin_tools=[CodeExecutionTool()],
diff --git a/examples/pai_anthropic_otel.py b/examples/pai_anthropic_otel.py
@@ -15,7 +15,7 @@
 
 
 agent = Agent(
-    'gateway:anthropic/claude-sonnet-4-0',
+    'gateway/anthropic:claude-sonnet-4-0',
     instructions='Extract information about the image.',
     model_settings=AnthropicModelSettings(
         anthropic_thinking={'type': 'enabled', 'budget_tokens': 1024},
diff --git a/examples/pai_anthropic_stream.py b/examples/pai_anthropic_stream.py
@@ -8,7 +8,7 @@
 logfire.instrument_httpx(capture_all=True)
 
 person_agent = Agent(
-    'gateway:anthropic/claude-sonnet-4-0',
+    'gateway/anthropic:claude-sonnet-4-0',
     instructions='You are a helpful assistant.',
     model_settings={'max_tokens': 1024},
     retries=0,
diff --git a/examples/pai_anthropic_vertex.py b/examples/pai_anthropic_vertex.py
@@ -34,7 +34,7 @@ def validate_dob(cls, v: date) -> date:
 client = AsyncAnthropicVertex(
     base_url='http://localhost:8787/google-vertex', access_token=api_key, region='unknown', project_id='unknown'
 )
-provider = AnthropicProvider(anthropic_client=client)  # type: ignore[reportUnknownArgumentType]
+provider = AnthropicProvider(anthropic_client=client)
 model = AnthropicModel('claude-sonnet-4', provider=provider)
 
 person_agent = Agent(
diff --git a/examples/pai_google.py b/examples/pai_google.py
@@ -25,7 +25,7 @@ def validate_dob(cls, v: date) -> date:
 
 
 person_agent = Agent(
-    'gateway:google-vertex/gemini-2.5-flash',
+    'gateway/google-vertex:gemini-2.5-flash',
     output_type=Person,
     instructions='Extract information about the person',
     model_settings=GoogleModelSettings(stop_sequences=['potato'], temperature=0.5, top_p=0.9),
diff --git a/examples/pai_google_stream.py b/examples/pai_google_stream.py
@@ -8,7 +8,7 @@
 logfire.instrument_httpx(capture_all=True)
 
 person_agent = Agent(
-    'gateway:google-vertex/gemini-2.5-flash',
+    'gateway/google-vertex:gemini-2.5-flash',
     # 'google-vertex:gemini-2.5-flash',
     instructions='You are a helpful assistant.',
     model_settings={'max_tokens': 1024},
diff --git a/examples/pai_groq.py b/examples/pai_groq.py
@@ -26,7 +26,7 @@ def validate_dob(cls, v: date) -> date:
 
 
 person_agent = Agent(
-    'gateway:groq/openai/gpt-oss-120b',
+    'gateway/groq:openai/gpt-oss-120b',
     output_type=Person,
     instructions='Extract information about the person',
     model_settings=GroqModelSettings(stop_sequences=['potato'], temperature=0.5, top_p=0.9),
diff --git a/examples/pai_openai.py b/examples/pai_openai.py
@@ -2,11 +2,12 @@
 
 import logfire
 from pydantic import BaseModel, field_validator
-from pydantic_ai import Agent
+from pydantic_ai import Agent, __version__
 
 logfire.configure(service_name='testing')
 logfire.instrument_pydantic_ai()
 logfire.instrument_httpx(capture_all=True)
+print('pydantic-ai version:', __version__)
 
 
 class Person(BaseModel, use_attribute_docstrings=True):
@@ -25,7 +26,7 @@ def validate_dob(cls, v: date) -> date:
 
 
 person_agent = Agent(
-    'gateway:openai/gpt-4.1-mini',
+    'gateway/openai-chat:gpt-4.1-mini',
     output_type=Person,
     instructions='Extract information about the person',
 )
diff --git a/examples/pai_openai_responses.py b/examples/pai_openai_responses.py
@@ -26,7 +26,7 @@ def validate_dob(cls, v: date) -> date:
 
 
 person_agent = Agent(
-    'gateway:openai-responses/gpt-4.1',
+    'gateway/openai-responses:gpt-4.1',
     output_type=Person,
     instructions='Extract information about the person',
     model_settings=OpenAIResponsesModelSettings(stop_sequences=['potato'], temperature=0.5, top_p=0.9),
diff --git a/examples/pai_openai_stream.py b/examples/pai_openai_stream.py
@@ -26,7 +26,7 @@ def validate_dob(cls, v: date) -> date:
 
 
 person_agent = Agent(
-    'gateway:openai/gpt-5',
+    'gateway/openai:gpt-5',
     instructions='You are a helpful assistant.',
     model_settings={'max_tokens': 1024},
     retries=2,
diff --git a/examples/pyproject.toml b/examples/pyproject.toml
@@ -6,7 +6,7 @@ dependencies = [
     "devtools>=0.12.2",
     "logfire[httpx]>=4.3.3",
     "opentelemetry-instrumentation-botocore>=0.57b0",
-    "pydantic-ai",
+    "pydantic-ai>=1.10.0",
     "types-boto3[bedrock-runtime]",
     "mypy-boto3-bedrock-runtime",
 ]
diff --git a/gateway/src/gateway.ts b/gateway/src/gateway.ts
@@ -18,7 +18,16 @@ export async function gateway(
   if (!apiTypeMatch) {
     return textResponse(404, 'Path not found')
   }
-  const [, apiType, restOfPath] = apiTypeMatch as unknown as [string, string, string]
+  let [, apiType, restOfPath] = apiTypeMatch as unknown as [string, string, string]
+
+  // support for other common names for openai api types
+  if (apiType === 'openai' || apiType === 'openai-chat') {
+    apiType = 'chat'
+  } else if (apiType === 'openai-responses') {
+    apiType = 'responses'
+  } else if (apiType === 'google-vertex') {
+    apiType = 'gemini'
+  }
 
   if (!guardAPIType(apiType)) {
     return textResponse(400, `Invalid API type '${apiType}', should be one of ${apiTypesArray.join(', ')}`)
@@ -43,7 +52,7 @@ export async function gateway(
   }
 
   // sort providers on priority, highest first
-  providerProxies.sort((a, b) => (b.priority ?? 1) - (a.priority ?? 1))
+  providerProxies.sort((a, b) => (b.priority ?? 0) - (a.priority ?? 0))
 
   const providerProxy = providerProxies[0]
   if (!providerProxy) {
diff --git a/gateway/src/providers/google/index.ts b/gateway/src/providers/google/index.ts
@@ -108,6 +108,6 @@ function regionFromUrl(url: string): null | string {
     return 'global'
   }
   // The group includes regions with hyphen like "europe-west4"
-  const match = url.match(/^https:\/\/([^-]+)-aiplatform\.googleapis\.com$/)
+  const match = url.match(/^https:\/\/(.+?)-aiplatform\.googleapis\.com$/)
   return match?.[1] ?? null
 }
diff --git a/gateway/test/providers/openai.spec.ts b/gateway/test/providers/openai.spec.ts
@@ -103,7 +103,7 @@ describe('openai', () => {
   test('openai responses', async ({ gateway }) => {
     const { fetch, otelBatch } = gateway
 
-    const client = new OpenAI({ apiKey: 'healthy', baseURL: 'https://example.com/chat', fetch })
+    const client = new OpenAI({ apiKey: 'healthy', baseURL: 'https://example.com/responses', fetch })
 
     const completion = await client.responses.create({
       model: 'gpt-5',
@@ -118,7 +118,7 @@ describe('openai', () => {
   test('openai responses with builtin tools', async ({ gateway }) => {
     const { fetch, otelBatch } = gateway
 
-    const client = new OpenAI({ apiKey: 'healthy', baseURL: 'https://example.com/chat', fetch })
+    const client = new OpenAI({ apiKey: 'healthy', baseURL: 'https://example.com/responses', fetch })
 
     const completion = await client.responses.create({
       model: 'gpt-5',
@@ -170,4 +170,42 @@ describe('openai', () => {
     expect(otelBatch, 'otelBatch length not 1').toHaveLength(1)
     expect(JSON.parse(otelBatch[0]!).resourceSpans?.[0].scopeSpans?.[0].spans?.[0]?.attributes).toMatchSnapshot('span')
   })
+
+  test('openai chat legacy name', async ({ gateway }) => {
+    const { fetch, otelBatch } = gateway
+
+    const client = new OpenAI({ apiKey: 'healthy', baseURL: 'https://example.com/openai', fetch })
+
+    const completion = await client.chat.completions.create({
+      model: 'gpt-5',
+      messages: [
+        { role: 'developer', content: 'You are a helpful assistant.' },
+        { role: 'user', content: 'What is the capital of France?' },
+      ],
+      max_completion_tokens: 1024,
+    })
+
+    expect(completion).toMatchSnapshot('llm')
+    expect(otelBatch, 'otelBatch length not 1').toHaveLength(1)
+    expect(JSON.parse(otelBatch[0]!).resourceSpans?.[0].scopeSpans?.[0].spans?.[0]?.attributes).toMatchSnapshot('span')
+  })
+
+  test('openai responses legacy name', async ({ gateway }) => {
+    const { fetch, otelBatch } = gateway
+
+    const client = new OpenAI({ apiKey: 'healthy', baseURL: 'https://example.com/openai-responses', fetch })
+
+    const completion = await client.chat.completions.create({
+      model: 'gpt-5',
+      messages: [
+        { role: 'developer', content: 'You are a helpful assistant.' },
+        { role: 'user', content: 'What is the capital of France?' },
+      ],
+      max_completion_tokens: 1024,
+    })
+
+    expect(completion).toMatchSnapshot('llm')
+    expect(otelBatch, 'otelBatch length not 1').toHaveLength(1)
+    expect(JSON.parse(otelBatch[0]!).resourceSpans?.[0].scopeSpans?.[0].spans?.[0]?.attributes).toMatchSnapshot('span')
+  })
 })
diff --git a/gateway/test/providers/openai.spec.ts.snap b/gateway/test/providers/openai.spec.ts.snap
diff --git a/gateway/test/worker.ts b/gateway/test/worker.ts
diff --git a/pyproject.toml b/pyproject.toml
diff --git a/uv.lock b/uv.lock

Original file line number	Diff line number	Diff line change
`@@ -34,7 +34,7 @@ def validate_dob(cls, v: date) -> date:`
`34`	`34`	`client = AsyncAnthropicVertex(`
`35`	`35`	`base_url='http://localhost:8787/google-vertex', access_token=api_key, region='unknown', project_id='unknown'`
`36`	`36`	`)`
`37`		`-provider = AnthropicProvider(anthropic_client=client) # type: ignore[reportUnknownArgumentType]`
	`37`	`+provider = AnthropicProvider(anthropic_client=client)`
`38`	`38`	`model = AnthropicModel('claude-sonnet-4', provider=provider)`
`39`	`39`
`40`	`40`	`person_agent = Agent(`