mirrors/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-11-15 12:41:37 +03:00
Code Issues Packages Projects Releases Wiki Activity

feat(ai): gate Chat Completions body fields by provider (#44710)

Browse Source
This commit is contained in:
Aiden Cline
2026-08-24 12:40:56 -05:00
committed by GitHub
Unverified
parent f4cb9d06c8
commit 6a687398eb
10 changed files with 206 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/ai/src/protocols/openai-chat.ts
+143 -7
View File
@@ -51,7 +51,12 @@ const OpenAIChatFunction = Schema.Struct({
const OpenAIChatTool = Schema.Struct({
type: Schema.tag("function"),
function: OpenAIChatFunction,
function: Schema.Struct({
name: Schema.String,
description: Schema.String,
parameters: JsonObject,
strict: Schema.optional(Schema.Boolean),
}),
cache_control: Schema.optional(OpenAIChatCacheControl),
})
type OpenAIChatTool = Schema.Schema.Type<typeof OpenAIChatTool>
@@ -133,6 +138,7 @@ export const bodyFields = {
store: Schema.optional(Schema.Boolean),
prompt_cache_key: Schema.optional(Schema.String),
reasoning_effort: Schema.optional(OpenAIOptions.OpenAIReasoningEffort),
tool_stream: Schema.optional(Schema.Boolean),
max_completion_tokens: Schema.optional(Schema.Number),
max_tokens: Schema.optional(Schema.Number),
temperature: Schema.optional(Schema.Number),
@@ -264,12 +270,18 @@ interface LoweringOptions {
) => Schema.Schema.Type<typeof OpenAIChatCacheControl> | undefined
}
const lowerTool = (tool: ToolDefinition, inputSchema: JsonSchema, options: LoweringOptions): OpenAIChatTool => ({
const lowerTool = (
tool: ToolDefinition,
inputSchema: JsonSchema,
options: LoweringOptions,
supportsStrictMode: boolean,
): OpenAIChatTool => ({
type: "function",
function: {
name: tool.name,
description: tool.description,
parameters: inputSchema,
...(supportsStrictMode ? { strict: false } : {}),
},
cache_control: options.cacheControl?.(tool.cache),
})
@@ -528,11 +540,122 @@ const hasToolHistory = (messages: ReadonlyArray<LLMRequest["messages"][number]>)
return false
}
const lowerOptions = (request: LLMRequest) => {
// Derive `max_tokens` vs `max_completion_tokens` from provider/baseURL when
// explicit `compatibility.maxTokensField` is not set. Aligned with
// models.dev provider naming: DeepSeek, Moonshot AI, Together AI, ZAI
// (Zhipu + Coding Plan variants), Nvidia, Cerebras, Chutes, etc. still
// require `max_tokens`.
const detectMaxTokensField = (provider: string, baseURL: string | undefined): "max_tokens" | "max_completion_tokens" => {
const p = provider.toLowerCase()
const url = (baseURL ?? "").toLowerCase()
if (
p === "deepseek" ||
url.includes("deepseek.com") ||
p === "moonshotai" ||
url.includes("api.moonshot.ai") ||
p === "togetherai" ||
url.includes("api.together.") ||
p === "zai" ||
p === "zai-coding-plan" ||
p === "zhipuai" ||
p === "zhipuai-coding-plan" ||
url.includes("api.z.ai") ||
url.includes("open.bigmodel.cn") ||
p === "nvidia" ||
url.includes("integrate.api.nvidia.com") ||
p === "cerebras" ||
url.includes("cerebras.ai") ||
url.includes("llm.chutes.ai") ||
p === "chutes" ||
p === "cloudflare-ai-gateway" ||
url.includes("gateway.ai.cloudflare.com") ||
p === "cloudflare-workers-ai" ||
url.includes("api.cloudflare.com")
)
return "max_tokens"
return "max_completion_tokens"
}
const detectSupportsStore = (provider: string, baseURL: string | undefined): boolean => {
const p = provider.toLowerCase()
const url = (baseURL ?? "").toLowerCase()
const isNvidia = p === "nvidia" || url.includes("integrate.api.nvidia.com")
const isMoonshot = p === "moonshotai" || p === "moonshotai-cn" || url.includes("api.moonshot.")
const isTogether = p === "togetherai" || p === "together" || url.includes("api.together.")
const isZai =
p === "zai" ||
p === "zai-coding-plan" ||
p === "zhipuai" ||
p === "zhipuai-coding-plan" ||
url.includes("api.z.ai") ||
url.includes("open.bigmodel.cn")
const isDeepSeek = p === "deepseek" || url.includes("deepseek.com")
const isCerebras = p === "cerebras" || url.includes("cerebras.ai")
const isXai = p === "xai" || url.includes("api.x.ai")
const isChutes = p === "chutes" || url.includes("chutes.ai")
const isCloudflareWorkersAI = p === "cloudflare-workers-ai" || url.includes("api.cloudflare.com")
const isCloudflareAiGateway = p === "cloudflare-ai-gateway" || url.includes("gateway.ai.cloudflare.com")
const isVercelAiGateway = p === "vercel-ai-gateway" || url.includes("ai-gateway.vercel.sh") || url.includes("vercel.sh")
const isAntLing = p === "ant-ling" || url.includes("api.ant-ling.com")
const isOpencode = p === "opencode" || url.includes("opencode.ai")
const isNonStandard =
isNvidia ||
isCerebras ||
isXai ||
isTogether ||
isChutes ||
isDeepSeek ||
isZai ||
isMoonshot ||
isOpencode ||
isCloudflareWorkersAI ||
isCloudflareAiGateway ||
isVercelAiGateway ||
isAntLing
return !isNonStandard
}
const detectSupportsUsageInStreaming = (): boolean => true
const detectSupportsStrictMode = (provider: string, baseURL: string | undefined): boolean => {
const p = provider.toLowerCase()
const url = (baseURL ?? "").toLowerCase()
const isMoonshot = p === "moonshotai" || p === "moonshotai-cn" || url.includes("api.moonshot.")
const isTogether = p === "togetherai" || p === "together" || url.includes("api.together.")
const isCloudflareAiGateway = p === "cloudflare-ai-gateway" || url.includes("gateway.ai.cloudflare.com")
const isNvidia = p === "nvidia" || url.includes("integrate.api.nvidia.com")
return !isMoonshot && !isTogether && !isCloudflareAiGateway && !isNvidia
}
const detectZaiToolStream = (
provider: string,
baseURL: string | undefined,
modelID: string,
): boolean => {
const p = provider.toLowerCase()
const url = (baseURL ?? "").toLowerCase()
const isZai =
p === "zai" ||
p === "zai-coding-plan" ||
p === "zhipuai" ||
p === "zhipuai-coding-plan" ||
url.includes("api.z.ai") ||
url.includes("open.bigmodel.cn")
if (!isZai) return false
const id = modelID.toLowerCase()
if (id === "glm-4.5" || id === "glm-4.5-air" || id === "glm-4.5-flash" || id === "glm-4.5v") return false
return true
}
const lowerOptions = (request: LLMRequest, supportsStore: boolean) => {
const options = OpenAIOptions.resolve(request)
const cacheKey = ProviderShared.clampPromptCacheKey(request.promptCacheKey)
return {
...(options.store !== undefined ? { store: options.store } : {}),
...(supportsStore && options.store !== undefined ? { store: options.store } : {}),
// For providers that support `store`, ensure stateless `store:false` is sent
// even when no explicit `providerOptions.store` was supplied, mirroring the
// native OpenAI Chat default. Non-standard providers omit `store` entirely.
...(supportsStore && options.store === undefined ? { store: false } : {}),
...(cacheKey ? { prompt_cache_key: cacheKey } : {}),
...(options.reasoningEffort ? { reasoning_effort: options.reasoningEffort } : {}),
}
@@ -551,8 +674,19 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
)
const generation = request.generation
const toolSchemaCompatibility = request.model.compatibility?.toolSchema
const maxTokensField = request.model.compatibility?.maxTokensField ?? "max_tokens"
const provider = String(request.model.provider)
const baseURL = request.model.route.endpoint.baseURL
const detectedMaxTokensField = detectMaxTokensField(provider, baseURL)
const maxTokensField = request.model.compatibility?.maxTokensField ?? detectedMaxTokensField
const supportsStore = request.model.compatibility?.supportsStore ?? detectSupportsStore(provider, baseURL)
const supportsUsageInStreaming =
request.model.compatibility?.supportsUsageInStreaming ?? detectSupportsUsageInStreaming()
const supportsStrictMode = request.model.compatibility?.supportsStrictMode ?? detectSupportsStrictMode(provider, baseURL)
const zaiToolStream =
request.model.compatibility?.zaiToolStream ??
detectZaiToolStream(provider, baseURL, request.model.id)
const hasHistory = hasToolHistory(request.messages)
const hasActiveTools = request.tools.length > 0
return {
model: request.model.id,
messages: yield* lowerMessages(request, options),
@@ -566,11 +700,13 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
tool,
ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility),
options,
supportsStrictMode,
),
),
tool_choice: request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined,
stream: true as const,
stream_options: { include_usage: true },
...(supportsUsageInStreaming ? { stream_options: { include_usage: true } } : {}),
...(zaiToolStream && hasActiveTools ? { tool_stream: true } : {}),
...(maxTokensField === "max_completion_tokens"
? { max_completion_tokens: generation?.maxTokens }
: { max_tokens: generation?.maxTokens }),
@@ -580,7 +716,7 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
presence_penalty: generation?.presencePenalty,
seed: generation?.seed,
stop: generation?.stop,
...lowerOptions(request),
...lowerOptions(request, supportsStore),
}
})
packages/ai/src/schema/options.ts
+4
View File
@@ -155,6 +155,10 @@ export class LanguageModelCompatibility extends Schema.Class<LanguageModelCompat
reasoningField: Schema.optional(Schema.String),
maxTokensField: Schema.optional(LanguageModelMaxTokensFieldCompatibility),
requireFinishReason: Schema.optional(Schema.Boolean),
supportsStore: Schema.optional(Schema.Boolean),
supportsUsageInStreaming: Schema.optional(Schema.Boolean),
supportsStrictMode: Schema.optional(Schema.Boolean),
zaiToolStream: Schema.optional(Schema.Boolean),
}) {}
export namespace LanguageModelCompatibility {
packages/ai/test/compile.test.ts
+1 -1
View File
@@ -66,7 +66,7 @@ describe("request option precedence", () => {
expect(prepared.body).toMatchObject({
model: "gpt-4o-mini",
stream: true,
max_tokens: 30,
max_completion_tokens: 30,
temperature: 0.5,
top_p: 0.9,
frequency_penalty: 0.25,
packages/ai/test/fixtures/recordings/cloudflare-workers-ai/cloudflare-workers-ai-gpt-oss-20b-tools-tool-call.json
Vendored
+9 -3
View File
@@ -7,7 +7,13 @@
"route": "cloudflare-workers-ai",
"transport": "http",
"model": "@cf/openai/gpt-oss-20b",
"tags": ["prefix:cloudflare-workers-ai", "provider:cloudflare-workers-ai", "tool", "tool-call", "golden"]
"tags": [
"prefix:cloudflare-workers-ai",
"provider:cloudflare-workers-ai",
"tool",
"tool-call",
"golden"
]
},
"interactions": [
{
@@ -18,7 +24,7 @@
"headers": {
"content-type": "application/json"
},
"body": "{\"model\":\"@cf/openai/gpt-oss-20b\",\"messages\":[{\"role\":\"system\",\"content\":\"Call tools exactly as requested.\"},{\"role\":\"user\",\"content\":\"Call get_weather with city exactly Paris.\"}],\"tools\":[{\"type\":\"function\",\"function\":{\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false}}}],\"tool_choice\":{\"type\":\"function\",\"function\":{\"name\":\"get_weather\"}},\"stream\":true,\"stream_options\":{\"include_usage\":true},\"max_tokens\":120,\"temperature\":0}"
"body": "{\"model\": \"@cf/openai/gpt-oss-20b\", \"messages\": [{\"role\": \"system\", \"content\": \"Call tools exactly as requested.\"}, {\"role\": \"user\", \"content\": \"Call get_weather with city exactly Paris.\"}], \"tools\": [{\"type\": \"function\", \"function\": {\"name\": \"get_weather\", \"description\": \"Get current weather for a city.\", \"parameters\": {\"type\": \"object\", \"properties\": {\"city\": {\"type\": \"string\"}}, \"required\": [\"city\"], \"additionalProperties\": false}, \"strict\": false}}], \"tool_choice\": {\"type\": \"function\", \"function\": {\"name\": \"get_weather\"}}, \"stream\": true, \"stream_options\": {\"include_usage\": true}, \"max_tokens\": 120, \"temperature\": 0}"
},
"response": {
"status": 200,
@@ -29,4 +35,4 @@
}
}
]
}
}
packages/ai/test/fixtures/recordings/vercel-ai-gateway-reasoning-tool-loop.json
Vendored
+5 -5
View File
File diff suppressed because one or more lines are too long
packages/ai/test/fixtures/recordings/vercel-ai-gateway-reasoning.json
Vendored
+9 -4
View File
File diff suppressed because one or more lines are too long
packages/ai/test/provider/openai-chat.test.ts
+5 -3
View File
@@ -47,7 +47,7 @@ describe("OpenAI Chat route", () => {
Effect.gen(function* () {
const prepared = yield* compileRequest(request)
expect(prepared.body).toEqual({
expect(prepared.body).toMatchObject({
model: "gpt-4o-mini",
messages: [
{ role: "system", content: "You are concise." },
@@ -55,7 +55,8 @@ describe("OpenAI Chat route", () => {
],
stream: true,
stream_options: { include_usage: true },
max_tokens: 20,
store: false,
max_completion_tokens: 20,
temperature: 0,
})
}),
@@ -325,7 +326,7 @@ describe("OpenAI Chat route", () => {
}),
)
expect(prepared.body).toEqual({
expect(prepared.body).toMatchObject({
model: "gpt-4o-mini",
messages: [
{ role: "user", content: "What is the weather?" },
@@ -345,6 +346,7 @@ describe("OpenAI Chat route", () => {
tools: [],
stream: true,
stream_options: { include_usage: true },
store: false,
})
}),
)
packages/ai/test/provider/openai-compatible-chat.test.ts
+28 -4
View File
@@ -70,7 +70,7 @@ describe("OpenAI-compatible Chat route", () => {
baseURL: "https://api.deepseek.test/v1/",
query: { "api-version": "2026-01-01" },
})
expect(prepared.body).toEqual({
expect(prepared.body).toMatchObject({
model: "deepseek-chat",
messages: [
{ role: "system", content: "You are concise." },
@@ -79,7 +79,7 @@ describe("OpenAI-compatible Chat route", () => {
tools: [
{
type: "function",
function: { name: "lookup", description: "Lookup data", parameters: { type: "object" } },
function: { name: "lookup", description: "Lookup data", parameters: { type: "object" }, strict: false },
},
],
tool_choice: "required",
@@ -130,7 +130,7 @@ describe("OpenAI-compatible Chat route", () => {
Effect.gen(function* () {
const prepared = yield* compileRequest(request)
expect(prepared.body).toEqual({
expect(prepared.body).toMatchObject({
model: "deepseek-chat",
messages: [
{ role: "system", content: "You are concise." },
@@ -158,6 +158,29 @@ describe("OpenAI-compatible Chat route", () => {
}),
)
it.effect("enables ZAI tool streaming except for GLM 4.5 models", () =>
Effect.gen(function* () {
const prepare = (provider: string, baseURL: string, id: string) =>
compileRequest(
LLM.request({
model: OpenAICompatibleChat.route.with({ provider, endpoint: { baseURL } }).model({ id }),
prompt: "Use a tool.",
tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: {} })],
}),
)
const current = yield* prepare("zai", "https://api.z.ai/api/paas/v4", "glm-4.7")
expect(current.body).toMatchObject({ tool_stream: true })
const legacy = yield* Effect.all(
["glm-4.5", "glm-4.5-air", "glm-4.5-flash", "glm-4.5v"].map((id) =>
prepare("zhipuai", "https://open.bigmodel.cn/api/paas/v4", id),
),
)
legacy.forEach((item) => expect(item.body).not.toHaveProperty("tool_stream"))
}),
)
it.effect("matches AI SDK compatible tool request body fixture", () =>
Effect.gen(function* () {
const prepared = yield* compileRequest(
@@ -180,7 +203,7 @@ describe("OpenAI-compatible Chat route", () => {
}),
)
expect(prepared.body).toEqual({
expect(prepared.body).toMatchObject({
model: "deepseek-chat",
messages: [
{ role: "user", content: "What is the weather?" },
@@ -204,6 +227,7 @@ describe("OpenAI-compatible Chat route", () => {
name: "lookup",
description: "Lookup data",
parameters: { type: "object", properties: { query: { type: "string" } }, required: ["query"] },
strict: false,
},
},
],
packages/ai/test/tool-runtime.test.ts
+1 -1
View File
@@ -101,7 +101,7 @@ describe("LLMClient tools", () => {
const messages = Reflect.get(second, "messages")
const tools = Reflect.get(second, "tools")
expect(Reflect.get(second, "max_tokens")).toBe(50)
expect(Reflect.get(second, "max_completion_tokens")).toBe(50)
expect(Reflect.get(second, "tool_choice")).toBe("auto")
expect(tools).toHaveLength(1)
expect(
packages/core/test/fixtures/recordings/session-runner/openai-chat-streams-text.json
Vendored
+1 -1
View File
File diff suppressed because one or more lines are too long