fix(ai): route Cloudflare AI Gateway Claude and OpenAI through native passthroughs (#52222)

This commit is contained in:
Aiden Cline
2026-09-30 14:01:56 -05:00
committed by GitHub
parent 7878744505
commit b020f1018b
8 changed files with 317 additions and 34 deletions
@@ -42,14 +42,24 @@ export type Settings = ProviderPackage.Settings &
export const baseURL = (input: GatewayURL) => {
if (input.baseURL) return input.baseURL
if (!input.accountId)
throw new ProviderConfigurationError({
provider: id,
message: "CloudflareAIGateway.configure requires accountId unless baseURL is supplied",
})
return `https://api.cloudflare.com/client/v4/accounts/${encodeURIComponent(input.accountId)}/ai/v1`
return `https://api.cloudflare.com/client/v4/accounts/${encodeURIComponent(requireAccountId(input))}/ai/v1`
}
// Cloudflare's REST API rejects parts of the Anthropic Messages and OpenAI Responses request shapes, such as
// system prompt blocks and `minimal` reasoning effort, so these models use the provider-native gateway endpoints.
const passthroughURL = (input: GatewayURL & GatewayOptions, provider: "anthropic/v1" | "openai") =>
`https://gateway.ai.cloudflare.com/v1/${encodeURIComponent(requireAccountId(input))}/${encodeURIComponent(gatewayId(input))}/${provider}`
const requireAccountId = (input: GatewayURL) => {
if (input.accountId) return input.accountId
throw new ProviderConfigurationError({
provider: id,
message: "CloudflareAIGateway.configure requires accountId unless baseURL is supplied",
})
}
const gatewayId = (input: GatewayOptions) => input.gatewayId?.trim() || "default"
export const responsesRoute = OpenAIResponses.route.with({
id: "cloudflare-ai-gateway-responses",
provider: id,
@@ -70,16 +80,12 @@ export const route = OpenAIChat.route.with({
export const routes = [responsesRoute, messagesRoute, route]
const auth = (input: LanguageModelOptions) => {
if ("auth" in input && input.auth) return input.auth
return Auth.optional(input.gatewayApiKey ?? ("apiKey" in input ? input.apiKey : undefined), "apiKey")
const credential = (input: LanguageModelOptions) =>
Auth.optional(input.gatewayApiKey ?? ("apiKey" in input ? input.apiKey : undefined), "apiKey")
.orElse(Auth.config(authEnvVars[0]))
.orElse(Auth.config(authEnvVars[1]))
.bearer()
}
const headers = (input: LanguageModelOptions) => ({
...(input.gatewayId === undefined ? {} : { "cf-aig-gateway-id": input.gatewayId.trim() || "default" }),
...(input.metadata === undefined
? {}
: { "cf-aig-metadata": Schema.encodeSync(Schema.fromJsonString(Schema.Unknown))(input.metadata) }),
@@ -90,31 +96,42 @@ const headers = (input: LanguageModelOptions) => ({
...input.headers,
})
const modelID = (input: string | ModelID) => {
const value = String(input)
if (value.startsWith("workers-ai/")) return value.slice("workers-ai/".length)
if (value.startsWith("anthropic/")) return `anthropic/${value.slice("anthropic/".length).replaceAll(".", "-")}`
return value
}
export const configure = (input: LanguageModelOptions) => {
const defaults = {
const custom = "auth" in input && input.auth ? input.auth : undefined
const rest = {
endpoint: { baseURL: baseURL(input) },
auth: auth(input),
headers: headers(input),
auth: custom ?? credential(input).bearer(),
headers: { "cf-aig-gateway-id": gatewayId(input), ...headers(input) },
http: input.http,
providerOptions: input.providerOptions,
}
const responses = responsesRoute.with(defaults)
const messages = messagesRoute.with(defaults)
const chat = route.with(defaults)
// A custom baseURL stands in for the REST API, so every model keeps using it with gateway model IDs.
const native = input.baseURL === undefined
const passthrough = (provider: "anthropic/v1" | "openai") =>
native
? {
...rest,
endpoint: { baseURL: passthroughURL(input, provider) },
auth: custom ?? Auth.bearerHeader("cf-aig-authorization", credential(input)),
headers: headers(input),
}
: rest
const responses = responsesRoute.with(passthrough("openai"))
const messages = messagesRoute.with(passthrough("anthropic/v1"))
const chat = route.with(rest)
return {
id,
model: (input: string | ModelID) => {
const wire = modelID(input)
if (String(input).startsWith("openai/")) return responses.model<OpenAIProviderOptionsInput>({ id: wire })
if (String(input).startsWith("anthropic/")) return messages.model<OpenAIProviderOptionsInput>({ id: wire })
return chat.model<OpenAIProviderOptionsInput>({ id: wire })
const value = String(input)
if (value.startsWith("openai/"))
return responses.model<OpenAIProviderOptionsInput>({ id: native ? value.slice("openai/".length) : value })
if (value.startsWith("anthropic/"))
return messages.model<OpenAIProviderOptionsInput>({
id: native ? value.slice("anthropic/".length).replaceAll(".", "-") : value,
})
if (value.startsWith("workers-ai/"))
return chat.model<OpenAIProviderOptionsInput>({ id: value.slice("workers-ai/".length) })
return chat.model<OpenAIProviderOptionsInput>({ id: value })
},
configure,
}
@@ -0,0 +1,38 @@
{
"version": 1,
"metadata": {
"provider": "cloudflare-ai-gateway",
"route": "cloudflare-ai-gateway-messages",
"transport": "http",
"model": "claude-haiku-4-5",
"tags": [
"prefix:cloudflare-ai-gateway",
"provider:cloudflare-ai-gateway",
"text",
"golden"
],
"name": "cloudflare-ai-gateway/cloudflare-ai-gateway-claude-haiku-4-5-text",
"recordedAt": "2026-09-30T04:35:43.657Z"
},
"interactions": [
{
"transport": "http",
"request": {
"method": "POST",
"url": "https://gateway.ai.cloudflare.com/v1/{account}/{gateway}/anthropic/v1/messages",
"headers": {
"anthropic-version": "2023-06-01",
"content-type": "application/json"
},
"body": "{\"model\":\"claude-haiku-4-5\",\"system\":[{\"type\":\"text\",\"text\":\"You are concise.\"}],\"messages\":[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"Reply exactly with: Hello!\"}]}],\"stream\":true,\"max_tokens\":40,\"temperature\":0}"
},
"response": {
"status": 200,
"headers": {
"content-type": "text/event-stream; charset=utf-8"
},
"body": "event: message_start\ndata: {\"p\": \"6247a7b95b5e3f\", \"type\":\"message_start\",\"message\":{\"model\":\"claude-haiku-4-5-20251001\",\"id\":\"msg_011CfZ16hXSVRfaXBDtUBKis\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[],\"container\":null,\"stop_reason\":null,\"stop_sequence\":null,\"stop_details\":null,\"usage\":{\"input_tokens\":18,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"cache_creation\":{\"ephemeral_5m_input_tokens\":0,\"ephemeral_1h_input_tokens\":0},\"output_tokens\":2,\"service_tier\":\"standard\",\"inference_geo\":\"not_available\"},\"diagnostics\":null}}\n\nevent: content_block_start\ndata: {\"p\": \"f9b\", \"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"} }\n\nevent: ping\ndata: {\"p\": \"cc76aa8\", \"type\": \"ping\"}\n\nevent: content_block_delta\ndata: {\"p\": \"94f\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"Hello!\"} }\n\nevent: content_block_stop\ndata: {\"p\": \"1a9d8\", \"type\":\"content_block_stop\",\"index\":0 }\n\nevent: message_delta\ndata: {\"p\": \"0cb3c23\", \"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"stop_details\":null,\"container\":null},\"usage\":{\"input_tokens\":18,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"output_tokens\":5} }\n\nevent: message_stop\ndata: {\"p\": \"6ef9de29973ed\", \"type\":\"message_stop\" }\n\n"
}
}
]
}
@@ -0,0 +1,58 @@
{
"version": 1,
"metadata": {
"provider": "cloudflare-ai-gateway",
"route": "cloudflare-ai-gateway-messages",
"transport": "http",
"model": "claude-haiku-4-5",
"tags": [
"prefix:cloudflare-ai-gateway",
"provider:cloudflare-ai-gateway",
"tool",
"tool-loop",
"golden"
],
"name": "cloudflare-ai-gateway/cloudflare-ai-gateway-claude-haiku-4-5-tool-loop",
"recordedAt": "2026-09-30T04:35:45.222Z"
},
"interactions": [
{
"transport": "http",
"request": {
"method": "POST",
"url": "https://gateway.ai.cloudflare.com/v1/{account}/{gateway}/anthropic/v1/messages",
"headers": {
"anthropic-version": "2023-06-01",
"content-type": "application/json"
},
"body": "{\"model\":\"claude-haiku-4-5\",\"system\":[{\"type\":\"text\",\"text\":\"Use the get_weather tool exactly once. After the tool result, reply exactly: Paris is sunny.\"}],\"messages\":[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"What is the weather in Paris?\"}]}],\"tools\":[{\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"input_schema\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false}}],\"stream\":true,\"max_tokens\":80,\"temperature\":0}"
},
"response": {
"status": 200,
"headers": {
"content-type": "text/event-stream; charset=utf-8"
},
"body": "event: message_start\ndata: {\"p\": \"f2f6a212\", \"type\":\"message_start\",\"message\":{\"model\":\"claude-haiku-4-5-20251001\",\"id\":\"msg_011CfZ16jzmxrCrHLxeCxrL8\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[],\"container\":null,\"stop_reason\":null,\"stop_sequence\":null,\"stop_details\":null,\"usage\":{\"input_tokens\":593,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"cache_creation\":{\"ephemeral_5m_input_tokens\":0,\"ephemeral_1h_input_tokens\":0},\"output_tokens\":26,\"service_tier\":\"standard\",\"inference_geo\":\"not_available\"},\"diagnostics\":null} }\n\nevent: content_block_start\ndata: {\"p\": \"7\", \"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"tool_use\",\"id\":\"toolu_019hb1DDwKBNMMeaKjZf1CpS\",\"name\":\"get_weather\",\"input\":{},\"caller\":{\"type\":\"direct\"}} }\n\nevent: ping\ndata: {\"p\": \"e112b39d9a4869\", \"type\": \"ping\"}\n\nevent: content_block_delta\ndata: {\"p\": \"12d\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"input_json_delta\",\"partial_json\":\"\"} }\n\nevent: content_block_delta\ndata: {\"p\": \"2\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"input_json_delta\",\"partial_json\":\"{\\\"city\\\"\"} }\n\nevent: content_block_delta\ndata: {\"p\": \"f6bb01e5ae5c1c7\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"input_json_delta\",\"partial_json\":\": \\\"Pari\"} }\n\nevent: content_block_delta\ndata: {\"p\": \"c26aa17\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"input_json_delta\",\"partial_json\":\"s\\\"}\"} }\n\nevent: content_block_stop\ndata: {\"p\": \"16f665\", \"type\":\"content_block_stop\",\"index\":0 }\n\nevent: message_delta\ndata: {\"p\": \"e00cc845875\", \"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"tool_use\",\"stop_sequence\":null,\"stop_details\":null,\"container\":null},\"usage\":{\"input_tokens\":593,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"output_tokens\":54} }\n\nevent: message_stop\ndata: {\"p\": \"5082e73ce8055a\", \"type\":\"message_stop\" }\n\n"
}
},
{
"transport": "http",
"request": {
"method": "POST",
"url": "https://gateway.ai.cloudflare.com/v1/{account}/{gateway}/anthropic/v1/messages",
"headers": {
"anthropic-version": "2023-06-01",
"content-type": "application/json"
},
"body": "{\"model\":\"claude-haiku-4-5\",\"system\":[{\"type\":\"text\",\"text\":\"Use the get_weather tool exactly once. After the tool result, reply exactly: Paris is sunny.\"}],\"messages\":[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"What is the weather in Paris?\"}]},{\"role\":\"assistant\",\"content\":[{\"type\":\"tool_use\",\"id\":\"toolu_019hb1DDwKBNMMeaKjZf1CpS\",\"name\":\"get_weather\",\"input\":{\"city\":\"Paris\"}}]},{\"role\":\"user\",\"content\":[{\"type\":\"tool_result\",\"tool_use_id\":\"toolu_019hb1DDwKBNMMeaKjZf1CpS\",\"content\":\"{\\\"temperature\\\":22,\\\"condition\\\":\\\"sunny\\\"}\"}]}],\"tools\":[{\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"input_schema\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false}}],\"stream\":true,\"max_tokens\":80,\"temperature\":0}"
},
"response": {
"status": 200,
"headers": {
"content-type": "text/event-stream; charset=utf-8"
},
"body": "event: message_start\ndata: {\"p\": \"ff048fa113d17ffc\", \"type\":\"message_start\",\"message\":{\"model\":\"claude-haiku-4-5-20251001\",\"id\":\"msg_011CfZ16ooU6seiMzxt69doW\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[],\"container\":null,\"stop_reason\":null,\"stop_sequence\":null,\"stop_details\":null,\"usage\":{\"input_tokens\":668,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"cache_creation\":{\"ephemeral_5m_input_tokens\":0,\"ephemeral_1h_input_tokens\":0},\"output_tokens\":1,\"service_tier\":\"standard\",\"inference_geo\":\"not_available\"},\"diagnostics\":null} }\n\nevent: content_block_start\ndata: {\"p\": \"5172b1\", \"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"} }\n\nevent: ping\ndata: {\"p\": \"d\", \"type\": \"ping\"}\n\nevent: content_block_delta\ndata: {\"p\": \"7f847273c\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"Paris\"} }\n\nevent: content_block_delta\ndata: {\"p\": \"1de4\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\" is sunny\"}}\n\nevent: content_block_delta\ndata: {\"p\": \"d61\", \"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\".\"}}\n\nevent: content_block_stop\ndata: {\"p\": \"29638a9b46e21ecd\", \"type\":\"content_block_stop\",\"index\":0 }\n\nevent: message_delta\ndata: {\"p\": \"8e5f9442b03f76\", \"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"stop_details\":null,\"container\":null},\"usage\":{\"input_tokens\":668,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"output_tokens\":7} }\n\nevent: message_stop\ndata: {\"p\": \"97d9afcead79c7a\", \"type\":\"message_stop\" }\n\n"
}
}
]
}
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
+1 -1
View File
@@ -214,7 +214,7 @@ describe("provider package entrypoints", () => {
})
expect(selected.provider).toBe(provider.id)
expect(selected.route.endpoint.baseURL).toBe(provider.baseURL({ accountId: "account" }))
expect(selected.route.defaults.headers).toEqual({ "x-application": "opencode" })
expect(selected.route.defaults.headers).toMatchObject({ "x-application": "opencode" })
expect(selected.route.defaults.http?.body).toEqual({ custom: true })
expect(selected.route.defaults.providerOptions).toEqual({ reasoningEffort: "high" })
}
+59 -3
View File
@@ -39,9 +39,16 @@ describe("Cloudflare", () => {
)
expect(responses.route).toBe("cloudflare-ai-gateway-responses")
expect(responses.body).toMatchObject({ model: "openai/gpt-5.4", stream: true })
expect(responses.model.route.endpoint.baseURL).toBe(
"https://gateway.ai.cloudflare.com/v1/test-account/test-gateway/openai",
)
expect(responses.body).toMatchObject({ model: "gpt-5.4", stream: true })
expect(messages.route).toBe("cloudflare-ai-gateway-messages")
expect(messages.body).toMatchObject({ model: "anthropic/claude-haiku-4-5" })
expect(messages.model.route.endpoint.baseURL).toBe(
"https://gateway.ai.cloudflare.com/v1/test-account/test-gateway/anthropic/v1",
)
expect(messages.body).toMatchObject({ model: "claude-haiku-4-5" })
expect(chat.model.route.endpoint.baseURL).toBe("https://api.cloudflare.com/client/v4/accounts/test-account/ai/v1")
expect(chat.route).toBe("cloudflare-ai-gateway-chat")
expect(chat.body).toMatchObject({ model: "xai/grok-4.6", stream: true })
expect(workers.route).toBe("cloudflare-ai-gateway-chat")
@@ -100,6 +107,54 @@ describe("Cloudflare", () => {
}),
)
it.effect("posts Claude and OpenAI models to the provider-native gateway endpoints", () =>
Effect.gen(function* () {
const gateway = CloudflareAIGateway.configure({ accountId: "test-account", apiKey: "test-token", cacheTtl: 300 })
const requests: Array<Request> = []
const capture = dynamicResponse((input) =>
Effect.gen(function* () {
requests.push(yield* HttpClientRequest.toWeb(input.request).pipe(Effect.orDie))
return input.respond("", { status: 500 })
}),
)
yield* LLM.generate(LLM.request({ model: gateway.model("anthropic/claude-sonnet-4.6"), prompt: "Hi" })).pipe(
Effect.provide(capture),
Effect.ignore,
)
yield* LLM.generate(LLM.request({ model: gateway.model("openai/gpt-5-nano"), prompt: "Hi" })).pipe(
Effect.provide(capture),
Effect.ignore,
)
expect(requests.map((request) => request.url)).toEqual([
"https://gateway.ai.cloudflare.com/v1/test-account/default/anthropic/v1/messages",
"https://gateway.ai.cloudflare.com/v1/test-account/default/openai/responses",
])
requests.forEach((request) => {
expect(request.headers.get("cf-aig-authorization")).toBe("Bearer test-token")
expect(request.headers.get("authorization")).toBeNull()
expect(request.headers.get("x-api-key")).toBeNull()
expect(request.headers.get("cf-aig-gateway-id")).toBeNull()
expect(request.headers.get("cf-aig-cache-ttl")).toBe("300")
})
}),
)
it.effect("sends the default gateway ID to the REST API", () =>
Effect.gen(function* () {
const prepared = yield* compileRequest(
LLM.request({
model: CloudflareAIGateway.configure({ accountId: "test-account", apiKey: "test-token" }).model(
"workers-ai/@cf/meta/llama-3.3-70b-instruct",
),
prompt: "Say hello.",
}),
)
expect(prepared.model.route.defaults.headers).toMatchObject({ "cf-aig-gateway-id": "default" })
}),
)
it.effect("preserves reasoning details for AI Gateway continuation", () =>
Effect.gen(function* () {
const model = CloudflareAIGateway.configure({
@@ -160,12 +215,13 @@ describe("Cloudflare", () => {
model: CloudflareAIGateway.configure({
baseURL: "https://gateway.proxy.test/v1",
apiKey: "test-token",
}).model("xai/grok-4.6"),
}).model("anthropic/claude-sonnet-4.6"),
prompt: "Say hello.",
}),
)
expect(prepared.model.route.endpoint.baseURL).toBe("https://gateway.proxy.test/v1")
expect(prepared.body).toMatchObject({ model: "anthropic/claude-sonnet-4.6" })
}),
)
@@ -49,6 +49,8 @@ const cloudflareWorkers = CloudflareWorkersAI.configure({
})
const cloudflareAIGatewayWorkers = cloudflareAIGateway.model("workers-ai/@cf/meta/llama-3.1-8b-instruct")
const cloudflareAIGatewayWorkersTools = cloudflareAIGateway.model("workers-ai/@cf/openai/gpt-oss-20b")
const cloudflareAIGatewayClaude = cloudflareAIGateway.model("anthropic/claude-haiku-4.5")
const cloudflareAIGatewayOpenAI = cloudflareAIGateway.model("openai/gpt-5-nano")
const cloudflareWorkersAI = cloudflareWorkers.model("@cf/meta/llama-3.1-8b-instruct")
const cloudflareWorkersAITools = cloudflareWorkers.model("@cf/openai/gpt-oss-20b")
const deepseek = DeepSeek.configure({ apiKey: process.env.DEEPSEEK_API_KEY ?? "fixture" }).model("deepseek-chat")
@@ -76,7 +78,7 @@ const openrouterOpus = OpenRouter.configure({
const redactCloudflareURL = (url: string) =>
url
.replace(/\/client\/v4\/accounts\/[^/]+\/ai\/v1\//, "/client/v4/accounts/{account}/ai/v1/")
.replace(/\/v1\/[^/]+\/[^/]+\/compat\//, "/v1/{account}/{gateway}/compat/")
.replace(/\/v1\/[^/]+\/[^/]+\/(anthropic|openai)\//, "/v1/{account}/{gateway}/$1/")
const cloudflareOptions = {
redact: { url: redactCloudflareURL },
@@ -177,6 +179,25 @@ describeRecordedGoldenScenarios([
options: cloudflareOptions,
scenarios: [{ id: "tool-call", maxTokens: 120 }],
},
{
name: "Cloudflare AI Gateway Claude Haiku 4.5",
prefix: "cloudflare-ai-gateway",
model: cloudflareAIGatewayClaude,
requires: ["CLOUDFLARE_ACCOUNT_ID", "CLOUDFLARE_API_TOKEN"],
options: { redact: { url: redactCloudflareURL, allowRequestHeaders: ["anthropic-version"] } },
scenarios: ["text", "tool-loop"],
},
{
name: "Cloudflare AI Gateway OpenAI GPT-5 Nano",
prefix: "cloudflare-ai-gateway",
model: cloudflareAIGatewayOpenAI,
requires: ["CLOUDFLARE_ACCOUNT_ID", "CLOUDFLARE_API_TOKEN"],
options: cloudflareOptions,
scenarios: [
{ id: "text", temperature: false, maxTokens: 2000 },
{ id: "tool-loop", temperature: false, maxTokens: 2000, timeout: 30_000 },
],
},
{
name: "Cloudflare Workers AI Llama 3.1 8B",
prefix: "cloudflare-workers-ai",