fix(core): expand reasoning option variants (#36894)

This commit is contained in:
Aiden Cline 2026-07-14 22:12:02 -05:00 committed by GitHub
commit 7eb7fe0763
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 988 additions and 53 deletions

View file

@ -13,7 +13,7 @@
"attachment": false,
"reasoning": true,
"reasoning_options": [
{ "type": "effort", "values": ["low", "high"] },
{ "type": "effort", "values": [null, "null", "low", "high"] },
{ "type": "budget_tokens", "min": 1024, "max": 64000 },
{ "type": "toggle" }
],
@ -27,6 +27,11 @@
"headers": { "x-mode": "high" },
"body": { "service_tier": "priority" }
}
},
"pro": {
"provider": {
"body": { "reasoning": { "mode": "pro" } }
}
}
}
}
@ -49,18 +54,296 @@
"reasoning_options": [{ "type": "budget_tokens", "min": 1024, "max": 64000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 8192 }
"limit": { "context": 128000, "output": 64000 }
},
"claude-effort": {
"id": "claude-effort",
"id": "claude-opus-4.7",
"name": "Claude Effort",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "effort", "values": ["low"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 8192 }
},
"claude-toggle": {
"id": "claude-toggle",
"name": "Claude Toggle",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 8192 }
},
"claude-opus-4-5": {
"id": "claude-opus-4-5",
"name": "Claude Opus 4.5",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [
{ "type": "effort", "values": ["low", "high"] },
{ "type": "budget_tokens", "min": 1024 }
],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 8192 }
}
}
},
"xai": {
"id": "xai",
"name": "xAI",
"env": ["XAI_API_KEY"],
"npm": "@ai-sdk/xai",
"models": {
"grok-4.5": {
"id": "grok-4.5",
"name": "Grok 4.5",
"release_date": "2026-07-08",
"attachment": true,
"reasoning": true,
"reasoning_options": [{ "type": "effort", "values": ["low", "medium", "high"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 500000, "output": 500000 }
}
}
},
"opencode-go": {
"id": "opencode-go",
"name": "OpenCode Go",
"env": ["OPENCODE_API_KEY"],
"npm": "@ai-sdk/openai-compatible",
"models": {
"minimax-m3": {
"id": "minimax-m3",
"name": "MiniMax-M3",
"release_date": "2026-05-31",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }],
"temperature": true,
"tool_call": true,
"limit": { "context": 1000000, "output": 131072 },
"provider": { "npm": "@ai-sdk/anthropic" }
}
}
},
"alibaba": {
"id": "alibaba",
"name": "Alibaba",
"env": ["ALIBABA_API_KEY"],
"npm": "@ai-sdk/alibaba",
"models": {
"toggle-only": {
"id": "toggle-only",
"name": "Toggle Only",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
},
"toggle-budget": {
"id": "toggle-budget",
"name": "Toggle Budget",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "budget_tokens", "max": 16000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
},
"vercel": {
"id": "vercel",
"name": "Vercel AI Gateway",
"env": ["AI_GATEWAY_API_KEY"],
"npm": "@ai-sdk/gateway",
"models": {
"alibaba/qwen-toggle": {
"id": "alibaba/qwen-toggle",
"name": "Gateway Alibaba Toggle",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "budget_tokens", "max": 16000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
},
"amazon/nova-2-lite": {
"id": "amazon/nova-2-lite",
"name": "Gateway Nova 2 Lite",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "effort", "values": ["low", "high"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
},
"deepseek/deepseek-toggle": {
"id": "deepseek/deepseek-toggle",
"name": "Gateway DeepSeek Toggle",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "effort", "values": ["low", "high"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
},
"openrouter": {
"id": "openrouter",
"name": "OpenRouter",
"env": ["OPENROUTER_API_KEY"],
"npm": "@openrouter/ai-sdk-provider",
"models": {
"openrouter-toggle": {
"id": "openrouter-toggle",
"name": "OpenRouter Toggle",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
},
"google": {
"id": "google",
"name": "Google",
"env": ["GOOGLE_GENERATIVE_AI_API_KEY"],
"npm": "@ai-sdk/google",
"models": {
"gemini-2.5-flash": {
"id": "gemini-2.5-flash",
"name": "Gemini 2.5 Flash",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "budget_tokens", "min": 0, "max": 16000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
},
"google-vertex": {
"id": "google-vertex",
"name": "Google Vertex",
"env": ["GOOGLE_VERTEX_PROJECT"],
"npm": "@ai-sdk/google-vertex",
"models": {
"gemini-2.5-flash-lite": {
"id": "gemini-2.5-flash-lite",
"name": "Gemini 2.5 Flash Lite",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "budget_tokens", "min": 512, "max": 16000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
},
"amazon-bedrock": {
"id": "amazon-bedrock",
"name": "Amazon Bedrock",
"env": ["AWS_ACCESS_KEY_ID"],
"npm": "@ai-sdk/amazon-bedrock",
"models": {
"amazon.nova-2-lite-v1:0": {
"id": "amazon.nova-2-lite-v1:0",
"name": "Nova 2 Lite",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "effort", "values": ["low", "high"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
},
"sap-ai-core": {
"id": "sap-ai-core",
"name": "SAP AI Core",
"env": ["AICORE_SERVICE_KEY"],
"npm": "@jerome-benoit/sap-ai-provider-v2",
"models": {
"gemini-2.5-flash": {
"id": "gemini-2.5-flash",
"name": "Gemini 2.5 Flash",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "budget_tokens", "min": 0, "max": 16000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
},
"amazon--nova-lite": {
"id": "amazon--nova-lite",
"name": "Nova Lite",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "toggle" }, { "type": "effort", "values": ["low", "high"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
},
"cohere--command-a-reasoning": {
"id": "cohere--command-a-reasoning",
"name": "Command A Reasoning",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [
{ "type": "toggle" },
{ "type": "effort", "values": ["low", "high"] },
{ "type": "budget_tokens", "min": 1 }
],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
},
"anthropic--claude-4.7-opus": {
"id": "anthropic--claude-4.7-opus",
"name": "Claude 4.7 Opus",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "effort", "values": ["low"] }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 8192 }
"limit": { "context": 128000, "output": 20000 }
},
"anthropic--claude-4-sonnet": {
"id": "anthropic--claude-4-sonnet",
"name": "Claude 4 Sonnet",
"release_date": "2026-01-01",
"attachment": false,
"reasoning": true,
"reasoning_options": [{ "type": "budget_tokens", "min": 1024, "max": 16000 }],
"temperature": true,
"tool_call": true,
"limit": { "context": 128000, "output": 20000 }
}
}
}

View file

@ -292,6 +292,12 @@ describe("ModelsDevPlugin", () => {
ModelV2.VariantID.make("high"),
])
const pro = yield* catalog.model.get(ProviderV2.ID.openai, ModelV2.ID.make("gpt-reasoning-pro"))
expect(pro).toMatchObject({
id: "gpt-reasoning-pro",
body: { reasoning: { mode: "pro" } },
})
const budgetModel = yield* catalog.model.get(ProviderV2.ID.anthropic, ModelV2.ID.make("claude-budget"))
expect(budgetModel?.variants).toContainEqual({
id: ModelV2.VariantID.make("high"),
@ -299,17 +305,292 @@ describe("ModelsDevPlugin", () => {
})
expect(budgetModel?.variants).toContainEqual({
id: ModelV2.VariantID.make("max"),
settings: { thinking: { type: "enabled", budgetTokens: 64000 } },
settings: { thinking: { type: "enabled", budgetTokens: 31999 } },
})
const anthropicEffortModel = yield* catalog.model.get(
ProviderV2.ID.anthropic,
ModelV2.ID.make("claude-effort"),
ModelV2.ID.make("claude-opus-4.7"),
)
expect(anthropicEffortModel?.variants).toContainEqual({
id: ModelV2.VariantID.make("low"),
settings: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
})
expect(anthropicEffortModel?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { thinking: { type: "disabled" } } },
{
id: ModelV2.VariantID.make("low"),
settings: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
},
])
const anthropicToggleModel = yield* catalog.model.get(
ProviderV2.ID.anthropic,
ModelV2.ID.make("claude-toggle"),
)
expect(anthropicToggleModel?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { thinking: { type: "disabled" } } },
{
id: ModelV2.VariantID.make("thinking"),
settings: { thinking: { type: "adaptive", display: "summarized" } },
},
])
const opus45 = yield* catalog.model.get(ProviderV2.ID.anthropic, ModelV2.ID.make("claude-opus-4-5"))
expect(opus45?.variants).toEqual([
{ id: ModelV2.VariantID.make("low"), settings: { effort: "low" } },
{ id: ModelV2.VariantID.make("high"), settings: { effort: "high" } },
])
const grok = yield* catalog.model.get(ProviderV2.ID.make("xai"), ModelV2.ID.make("grok-4.5"))
expect(grok?.variants).toEqual(
["low", "medium", "high"].map((id) => ({
id: ModelV2.VariantID.make(id),
settings: { reasoningEffort: id },
})),
)
const minimax = yield* catalog.model.get(ProviderV2.ID.make("opencode-go"), ModelV2.ID.make("minimax-m3"))
expect(minimax?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { thinking: { type: "disabled" } } },
{
id: ModelV2.VariantID.make("thinking"),
settings: { thinking: { type: "adaptive", display: "summarized" } },
},
])
const toggle = yield* catalog.model.get(ProviderV2.ID.make("alibaba"), ModelV2.ID.make("toggle-only"))
expect(toggle?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { enableThinking: false } },
{ id: ModelV2.VariantID.make("thinking"), settings: { enableThinking: true } },
])
const combined = yield* catalog.model.get(ProviderV2.ID.make("alibaba"), ModelV2.ID.make("toggle-budget"))
expect(combined?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { enableThinking: false } },
{
id: ModelV2.VariantID.make("high"),
settings: { enableThinking: true, thinkingBudget: 8000 },
},
{
id: ModelV2.VariantID.make("max"),
settings: { enableThinking: true, thinkingBudget: 16000 },
},
])
const gateway = yield* catalog.model.get(ProviderV2.ID.make("vercel"), ModelV2.ID.make("alibaba/qwen-toggle"))
expect(gateway?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { enableThinking: false } },
{
id: ModelV2.VariantID.make("high"),
settings: { enableThinking: true, thinkingBudget: 8000 },
},
{
id: ModelV2.VariantID.make("max"),
settings: { enableThinking: true, thinkingBudget: 16000 },
},
])
const gatewayNova = yield* catalog.model.get(
ProviderV2.ID.make("vercel"),
ModelV2.ID.make("amazon/nova-2-lite"),
)
expect(gatewayNova?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { additionalModelRequestFields: { reasoningConfig: { type: "disabled" } } },
},
{
id: ModelV2.VariantID.make("low"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "low" } },
},
{
id: ModelV2.VariantID.make("high"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "high" } },
},
])
const gatewayFallback = yield* catalog.model.get(
ProviderV2.ID.make("vercel"),
ModelV2.ID.make("deepseek/deepseek-toggle"),
)
expect(gatewayFallback?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { reasoning: { enabled: false } },
},
{
id: ModelV2.VariantID.make("low"),
settings: { reasoningEffort: "low" },
},
{
id: ModelV2.VariantID.make("high"),
settings: { reasoningEffort: "high" },
},
])
const openrouter = yield* catalog.model.get(
ProviderV2.ID.make("openrouter"),
ModelV2.ID.make("openrouter-toggle"),
)
expect(openrouter?.variants).toEqual([
{ id: ModelV2.VariantID.make("none"), settings: { reasoning: { enabled: false } } },
{ id: ModelV2.VariantID.make("thinking"), settings: { reasoning: { enabled: true } } },
])
const google = yield* catalog.model.get(ProviderV2.ID.make("google"), ModelV2.ID.make("gemini-2.5-flash"))
expect(google?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { thinkingConfig: { includeThoughts: false, thinkingBudget: 0 } },
},
{
id: ModelV2.VariantID.make("high"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 8000 } },
},
{
id: ModelV2.VariantID.make("max"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 16000 } },
},
])
const vertex = yield* catalog.model.get(
ProviderV2.ID.make("google-vertex"),
ModelV2.ID.make("gemini-2.5-flash-lite"),
)
expect(vertex?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { thinkingConfig: { includeThoughts: false, thinkingBudget: 0 } },
},
{
id: ModelV2.VariantID.make("high"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 8000 } },
},
{
id: ModelV2.VariantID.make("max"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 16000 } },
},
])
const bedrock = yield* catalog.model.get(
ProviderV2.ID.make("amazon-bedrock"),
ModelV2.ID.make("amazon.nova-2-lite-v1:0"),
)
expect(bedrock?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { additionalModelRequestFields: { reasoningConfig: { type: "disabled" } } },
},
{
id: ModelV2.VariantID.make("low"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "low" } },
},
{
id: ModelV2.VariantID.make("high"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "high" } },
},
])
const sapGemini = yield* catalog.model.get(
ProviderV2.ID.make("sap-ai-core"),
ModelV2.ID.make("gemini-2.5-flash"),
)
expect(sapGemini?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { modelParams: { thinkingConfig: { includeThoughts: false, thinkingBudget: 0 } } },
},
{
id: ModelV2.VariantID.make("high"),
settings: { modelParams: { thinkingConfig: { includeThoughts: true, thinkingBudget: 8000 } } },
},
{
id: ModelV2.VariantID.make("max"),
settings: { modelParams: { thinkingConfig: { includeThoughts: true, thinkingBudget: 16000 } } },
},
])
const sapNova = yield* catalog.model.get(
ProviderV2.ID.make("sap-ai-core"),
ModelV2.ID.make("amazon--nova-lite"),
)
expect(sapNova?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: {
modelParams: { additionalModelRequestFields: { thinking: { type: "disabled" } } },
},
},
{
id: ModelV2.VariantID.make("low"),
settings: {
modelParams: { additionalModelRequestFields: { output_config: { effort: "low" } } },
},
},
{
id: ModelV2.VariantID.make("high"),
settings: {
modelParams: { additionalModelRequestFields: { output_config: { effort: "high" } } },
},
},
])
const sapCohere = yield* catalog.model.get(
ProviderV2.ID.make("sap-ai-core"),
ModelV2.ID.make("cohere--command-a-reasoning"),
)
expect(sapCohere?.variants).toEqual([
{
id: ModelV2.VariantID.make("none"),
settings: { modelParams: { thinking: { type: "disabled" } } },
},
{
id: ModelV2.VariantID.make("low"),
settings: { modelParams: { reasoning_effort: "low" } },
},
{
id: ModelV2.VariantID.make("high"),
settings: { modelParams: { reasoning_effort: "high" } },
},
])
const sapAnthropicEffort = yield* catalog.model.get(
ProviderV2.ID.make("sap-ai-core"),
ModelV2.ID.make("anthropic--claude-4.7-opus"),
)
expect(sapAnthropicEffort?.variants).toEqual([
{
id: ModelV2.VariantID.make("low"),
settings: {
modelParams: {
additionalModelRequestFields: {
thinking: { type: "adaptive", display: "summarized" },
output_config: { effort: "low" },
},
},
},
},
])
const sapAnthropicBudget = yield* catalog.model.get(
ProviderV2.ID.make("sap-ai-core"),
ModelV2.ID.make("anthropic--claude-4-sonnet"),
)
expect(sapAnthropicBudget?.variants).toEqual([
{
id: ModelV2.VariantID.make("high"),
settings: {
modelParams: {
additionalModelRequestFields: { thinking: { type: "enabled", budget_tokens: 8000 } },
},
},
},
{
id: ModelV2.VariantID.make("max"),
settings: {
modelParams: {
additionalModelRequestFields: { thinking: { type: "enabled", budget_tokens: 16000 } },
},
},
},
])
}).pipe(Effect.provide(AppNodeBuilder.build(ModelsDev.node))),
(previous) =>
Effect.sync(() => {