diff --git a/packages/types/src/providers/bedrock.ts b/packages/types/src/providers/bedrock.ts index 9ea52bced89..aa665ef1c48 100644 --- a/packages/types/src/providers/bedrock.ts +++ b/packages/types/src/providers/bedrock.ts @@ -167,6 +167,31 @@ export const bedrockModels = { }, ], }, + "anthropic.claude-opus-4-7": { + maxTokens: 8192, + contextWindow: 200_000, // Default 200K, extendable to 1M with beta flag 'context-1m-2025-08-07' + supportsImages: true, + supportsPromptCache: true, + supportsReasoningBudget: true, + supportsTemperature: false, // Temperature is deprecated for this model + inputPrice: 5.0, // $5 per million input tokens (≤200K context) + outputPrice: 25.0, // $25 per million output tokens (≤200K context) + cacheWritesPrice: 6.25, // $6.25 per million tokens + cacheReadsPrice: 0.5, // $0.50 per million tokens + minTokensPerCachePoint: 1024, + maxCachePoints: 4, + cachableFields: ["system", "messages", "tools"], + // Tiered pricing for extended context (requires beta flag 'context-1m-2025-08-07') + tiers: [ + { + contextWindow: 1_000_000, // 1M tokens with beta flag + inputPrice: 10.0, // $10 per million input tokens (>200K context) + outputPrice: 37.5, // $37.50 per million output tokens (>200K context) + cacheWritesPrice: 12.5, // $12.50 per million tokens (>200K context) + cacheReadsPrice: 1.0, // $1.00 per million tokens (>200K context) + }, + ], + }, "anthropic.claude-opus-4-5-20251101-v1:0": { maxTokens: 8192, contextWindow: 200_000, @@ -525,6 +550,7 @@ export const BEDROCK_1M_CONTEXT_MODEL_IDS = [ "anthropic.claude-sonnet-4-5-20250929-v1:0", "anthropic.claude-sonnet-4-6", "anthropic.claude-opus-4-6-v1", + "anthropic.claude-opus-4-7", ] as const // Amazon Bedrock models that support Global Inference profiles @@ -535,6 +561,7 @@ export const BEDROCK_1M_CONTEXT_MODEL_IDS = [ // - Claude Haiku 4.5 // - Claude Opus 4.5 // - Claude Opus 4.6 +// - Claude Opus 4.7 export const BEDROCK_GLOBAL_INFERENCE_MODEL_IDS = [ "anthropic.claude-sonnet-4-20250514-v1:0", "anthropic.claude-sonnet-4-5-20250929-v1:0", @@ -542,6 +569,7 @@ export const BEDROCK_GLOBAL_INFERENCE_MODEL_IDS = [ "anthropic.claude-haiku-4-5-20251001-v1:0", "anthropic.claude-opus-4-5-20251101-v1:0", "anthropic.claude-opus-4-6-v1", + "anthropic.claude-opus-4-7", ] as const // Amazon Bedrock Service Tier types diff --git a/src/api/providers/__tests__/bedrock-reasoning.spec.ts b/src/api/providers/__tests__/bedrock-reasoning.spec.ts index 9dd271744c2..ed86dcb9e61 100644 --- a/src/api/providers/__tests__/bedrock-reasoning.spec.ts +++ b/src/api/providers/__tests__/bedrock-reasoning.spec.ts @@ -282,6 +282,110 @@ describe("AwsBedrockHandler - Extended Thinking", () => { expect(reasoningChunks[1].text).toBe(" about this problem.") }) + it("should use adaptive thinking for Opus 4.7 instead of enabled with budget_tokens", async () => { + handler = new AwsBedrockHandler({ + apiProvider: "bedrock", + apiModelId: "anthropic.claude-opus-4-7", + awsRegion: "us-east-1", + enableReasoningEffort: true, + modelMaxTokens: 8192, + modelMaxThinkingTokens: 4096, + }) + + mockSend.mockResolvedValue({ + stream: (async function* () { + yield { messageStart: { role: "assistant" } } + yield { + contentBlockStart: { + content_block: { type: "thinking", thinking: "Thinking adaptively..." }, + contentBlockIndex: 0, + }, + } + yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } } + })(), + }) + + const messages = [{ role: "user" as const, content: "Test message" }] + const stream = handler.createMessage("System prompt", messages) + + const chunks = [] + for await (const chunk of stream) { + chunks.push(chunk) + } + + // Opus 4.7 must use thinking.type: "adaptive" with output_config.effort + expect(mockSend).toHaveBeenCalledTimes(1) + expect(capturedPayload).toBeDefined() + expect(capturedPayload.additionalModelRequestFields).toBeDefined() + expect(capturedPayload.additionalModelRequestFields.thinking).toEqual({ + type: "adaptive", + }) + expect(capturedPayload.additionalModelRequestFields.output_config).toEqual({ + effort: "high", + }) + + // Must NOT have budget_tokens (causes 400 error on Opus 4.7) + expect(capturedPayload.additionalModelRequestFields.thinking).not.toHaveProperty("budget_tokens") + }) + + it("should exclude temperature from inferenceConfig for Opus 4.7 (supportsTemperature: false)", async () => { + handler = new AwsBedrockHandler({ + apiProvider: "bedrock", + apiModelId: "anthropic.claude-opus-4-7", + awsRegion: "us-east-1", + modelTemperature: 0.7, + }) + + mockSend.mockResolvedValue({ + stream: (async function* () { + yield { messageStart: { role: "assistant" } } + yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } } + })(), + }) + + const messages = [{ role: "user" as const, content: "Test message" }] + const stream = handler.createMessage("System prompt", messages) + + for await (const chunk of stream) { + // consume stream + } + + expect(mockSend).toHaveBeenCalledTimes(1) + expect(capturedPayload).toBeDefined() + // Temperature must NOT be present for Opus 4.7 + expect(capturedPayload.inferenceConfig).not.toHaveProperty("temperature") + // maxTokens should still be present + expect(capturedPayload.inferenceConfig).toHaveProperty("maxTokens") + }) + + it("should include temperature in inferenceConfig for models that support it", async () => { + handler = new AwsBedrockHandler({ + apiProvider: "bedrock", + apiModelId: "anthropic.claude-sonnet-4-20250514-v1:0", + awsRegion: "us-east-1", + modelTemperature: 0.5, + }) + + mockSend.mockResolvedValue({ + stream: (async function* () { + yield { messageStart: { role: "assistant" } } + yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } } + })(), + }) + + const messages = [{ role: "user" as const, content: "Test message" }] + const stream = handler.createMessage("System prompt", messages) + + for await (const chunk of stream) { + // consume stream + } + + expect(mockSend).toHaveBeenCalledTimes(1) + expect(capturedPayload).toBeDefined() + // Temperature should be present for Sonnet 4 + expect(capturedPayload.inferenceConfig).toHaveProperty("temperature", 0.5) + }) + it("should support API key authentication", async () => { handler = new AwsBedrockHandler({ apiProvider: "bedrock", diff --git a/src/api/providers/bedrock.ts b/src/api/providers/bedrock.ts index 2116dff08a7..cd3d8192de1 100644 --- a/src/api/providers/bedrock.ts +++ b/src/api/providers/bedrock.ts @@ -59,14 +59,26 @@ interface BedrockInferenceConfig { // Define interface for Bedrock additional model request fields // This includes thinking configuration, 1M context beta, and other model-specific parameters interface BedrockAdditionalModelFields { - thinking?: { - type: "enabled" - budget_tokens: number + thinking?: + | { + type: "enabled" + budget_tokens: number + } + | { + type: "adaptive" + } + output_config?: { + effort?: "low" | "medium" | "high" } anthropic_beta?: string[] [key: string]: any // Add index signature to be compatible with DocumentType } +// Models that only support thinking.type: "adaptive" (not "enabled" with budget_tokens) +const BEDROCK_ADAPTIVE_THINKING_ONLY_MODEL_IDS = [ + "anthropic.claude-opus-4-7", +] as const + // Define interface for Bedrock payload interface BedrockPayload { modelId: BedrockModelId | string @@ -390,12 +402,28 @@ export class AwsBedrockHandler extends BaseProvider implements SingleCompletionH if ((isThinkingExplicitlyEnabled || isThinkingEnabledBySettings) && modelConfig.info.supportsReasoningBudget) { thinkingEnabled = true - additionalModelRequestFields = { - thinking: { - type: "enabled", - budget_tokens: metadata?.thinking?.maxThinkingTokens || modelConfig.reasoningBudget || 4096, - }, + + // Opus 4.7 only supports thinking.type: "adaptive" with output_config.effort + // (NOT "enabled" with budget_tokens which returns a 400 error) + const baseId = this.parseBaseModelId(modelConfig.id) + if (BEDROCK_ADAPTIVE_THINKING_ONLY_MODEL_IDS.includes(baseId as any)) { + additionalModelRequestFields = { + thinking: { + type: "adaptive", + }, + output_config: { + effort: "high", + }, + } + } else { + additionalModelRequestFields = { + thinking: { + type: "enabled", + budget_tokens: metadata?.thinking?.maxThinkingTokens || modelConfig.reasoningBudget || 4096, + }, + } } + logger.info("Extended thinking enabled for Bedrock request", { ctx: "bedrock", modelId: modelConfig.id, @@ -405,7 +433,10 @@ export class AwsBedrockHandler extends BaseProvider implements SingleCompletionH const inferenceConfig: BedrockInferenceConfig = { maxTokens: modelConfig.maxTokens || (modelConfig.info.maxTokens as number), - temperature: modelConfig.temperature ?? (this.options.modelTemperature as number), + // Only include temperature if the model supports it (Opus 4.7 deprecated temperature) + ...(modelConfig.info.supportsTemperature !== false && { + temperature: modelConfig.temperature ?? (this.options.modelTemperature as number), + }), } // Check if 1M context is enabled for supported Claude 4 models @@ -743,7 +774,10 @@ export class AwsBedrockHandler extends BaseProvider implements SingleCompletionH const inferenceConfig: BedrockInferenceConfig = { maxTokens: modelConfig.maxTokens || (modelConfig.info.maxTokens as number), - temperature: modelConfig.temperature ?? (this.options.modelTemperature as number), + // Only include temperature if the model supports it (Opus 4.7 deprecated temperature) + ...(modelConfig.info.supportsTemperature !== false && { + temperature: modelConfig.temperature ?? (this.options.modelTemperature as number), + }), } // For completePrompt, use a unique conversation ID based on the prompt