@lobehub/lobehub 2.0.0-next.100 → 2.0.0-next.102

package/packages/model-runtime/src/index.ts

@@ -31,6 +31,7 @@ export { LobeQwenAI } from './providers/qwen';
 export { LobeStepfunAI } from './providers/stepfun';
 export { LobeTogetherAI } from './providers/togetherai';
 export { LobeVolcengineAI } from './providers/volcengine';
+export { LobeZenMuxAI } from './providers/zenmux';
 export { LobeZeroOneAI } from './providers/zeroone';
 export { LobeZhipuAI } from './providers/zhipu';
 export * from './types';
@@ -40,4 +41,3 @@ export { AgentRuntimeError } from './utils/createError';
 export { getModelPropertyWithFallback } from './utils/getFallbackModelProperty';
 export { getModelPricing } from './utils/getModelPricing';
 export { parseDataUri } from './utils/uriParser';
-

package/packages/model-runtime/src/providers/anthropic/index.ts

@@ -21,6 +21,7 @@ import { MODEL_LIST_CONFIGS, processModelList } from '../../utils/modelParse';
 import { StreamingResponse } from '../../utils/response';
 import { createAnthropicGenerateObject } from './generateObject';
 import { handleAnthropicError } from './handleAnthropicError';
+import { resolveCacheTTL } from './resolveCacheTTL';
 
 export interface AnthropicModelCard {
   created_at: string;
@@ -33,44 +34,6 @@ type anthropicTools = Anthropic.Tool | Anthropic.WebSearchTool20250305;
 const modelsWithSmallContextWindow = new Set(['claude-3-opus-20240229', 'claude-3-haiku-20240307']);
 
 const DEFAULT_BASE_URL = 'https://api.anthropic.com';
-const DEFAULT_CACHE_TTL = '5m' as const;
-
-type CacheTTL = Anthropic.Messages.CacheControlEphemeral['ttl'];
-
-/**
- * Resolves cache TTL from Anthropic payload or request settings
- * Returns the first valid TTL found in system messages or content blocks
- */
-const resolveCacheTTL = (
-  requestPayload: ChatStreamPayload,
-  anthropicPayload: Anthropic.MessageCreateParams,
-): CacheTTL | undefined => {
-  // Check system messages for cache TTL
-  if (Array.isArray(anthropicPayload.system)) {
-    for (const block of anthropicPayload.system) {
-      const ttl = block.cache_control?.ttl;
-      if (ttl) return ttl;
-    }
-  }
-
-  // Check message content blocks for cache TTL
-  for (const message of anthropicPayload.messages ?? []) {
-    if (!Array.isArray(message.content)) continue;
-
-    for (const block of message.content) {
-      // Message content blocks might have cache_control property
-      const ttl = ('cache_control' in block && block.cache_control?.ttl) as CacheTTL | undefined;
-      if (ttl) return ttl;
-    }
-  }
-
-  // Use default TTL if context caching is enabled
-  if (requestPayload.enabledContextCaching) {
-    return DEFAULT_CACHE_TTL;
-  }
-
-  return undefined;
-};
 
 interface AnthropicAIParams extends ClientOptions {
   id?: string;

package/packages/model-runtime/src/providers/anthropic/resolveCacheTTL.ts

@@ -0,0 +1,44 @@
+import Anthropic from '@anthropic-ai/sdk';
+
+import { ChatStreamPayload } from '../../types';
+
+type CacheTTL = Anthropic.Messages.CacheControlEphemeral['ttl'];
+
+const DEFAULT_CACHE_TTL = '5m' as const;
+
+/**
+ * Resolves cache TTL from Anthropic payload or request settings.
+ * Returns the first valid TTL found in system messages or content blocks.
+ */
+export const resolveCacheTTL = (
+  requestPayload: ChatStreamPayload,
+  anthropicPayload: {
+    messages: Anthropic.MessageCreateParams['messages'];
+    system: Anthropic.MessageCreateParams['system'];
+  },
+): CacheTTL | undefined => {
+  // Check system messages for cache TTL
+  if (Array.isArray(anthropicPayload.system)) {
+    for (const block of anthropicPayload.system) {
+      const ttl = block.cache_control?.ttl;
+      if (ttl) return ttl;
+    }
+  }
+
+  // Check message content blocks for cache TTL
+  for (const message of anthropicPayload.messages ?? []) {
+    if (!Array.isArray(message.content)) continue;
+
+    for (const block of message.content) {
+      const ttl = ('cache_control' in block && block.cache_control?.ttl) as CacheTTL | undefined;
+      if (ttl) return ttl;
+    }
+  }
+
+  // Use default TTL if context caching is enabled
+  if (requestPayload.enabledContextCaching) {
+    return DEFAULT_CACHE_TTL;
+  }
+
+  return undefined;
+};

package/packages/model-runtime/src/providers/bedrock/index.test.ts

@@ -173,7 +173,18 @@ describe('LobeBedrockAI', () => {
           body: JSON.stringify({
             anthropic_version: 'bedrock-2023-05-31',
             max_tokens: 4096,
-            messages: [{ content: 'Hello', role: 'user' }],
+            messages: [
+              {
+                content: [
+                  {
+                    cache_control: { type: 'ephemeral' },
+                    text: 'Hello',
+                    type: 'text',
+                  },
+                ],
+                role: 'user',
+              },
+            ],
             temperature: 0,
             top_p: 1,
           }),
@@ -211,8 +222,25 @@ describe('LobeBedrockAI', () => {
           body: JSON.stringify({
             anthropic_version: 'bedrock-2023-05-31',
             max_tokens: 4096,
-            messages: [{ content: 'Hello', role: 'user' }],
-            system: 'You are an awesome greeter',
+            messages: [
+              {
+                content: [
+                  {
+                    cache_control: { type: 'ephemeral' },
+                    text: 'Hello',
+                    type: 'text',
+                  },
+                ],
+                role: 'user',
+              },
+            ],
+            system: [
+              {
+                cache_control: { type: 'ephemeral' },
+                text: 'You are an awesome greeter',
+                type: 'text',
+              },
+            ],
             temperature: 0,
             top_p: 1,
           }),
@@ -248,7 +276,18 @@ describe('LobeBedrockAI', () => {
           body: JSON.stringify({
             anthropic_version: 'bedrock-2023-05-31',
             max_tokens: 2048,
-            messages: [{ content: 'Hello', role: 'user' }],
+            messages: [
+              {
+                content: [
+                  {
+                    cache_control: { type: 'ephemeral' },
+                    text: 'Hello',
+                    type: 'text',
+                  },
+                ],
+                role: 'user',
+              },
+            ],
             temperature: 0.25,
             top_p: 1,
           }),
@@ -327,7 +366,18 @@ describe('LobeBedrockAI', () => {
           body: JSON.stringify({
             anthropic_version: 'bedrock-2023-05-31',
             max_tokens: 4096,
-            messages: [{ content: 'Hello', role: 'user' }],
+            messages: [
+              {
+                content: [
+                  {
+                    cache_control: { type: 'ephemeral' },
+                    text: 'Hello',
+                    type: 'text',
+                  },
+                ],
+                role: 'user',
+              },
+            ],
             temperature: 0,
           }),
           contentType: 'application/json',
@@ -363,7 +413,18 @@ describe('LobeBedrockAI', () => {
           body: JSON.stringify({
             anthropic_version: 'bedrock-2023-05-31',
             max_tokens: 2048,
-            messages: [{ content: 'Hello', role: 'user' }],
+            messages: [
+              {
+                content: [
+                  {
+                    cache_control: { type: 'ephemeral' },
+                    text: 'Hello',
+                    type: 'text',
+                  },
+                ],
+                role: 'user',
+              },
+            ],
             temperature: 0.25,
             top_p: 1,
           }),
@@ -418,7 +479,18 @@ describe('LobeBedrockAI', () => {
             body: JSON.stringify({
               anthropic_version: 'bedrock-2023-05-31',
               max_tokens: 4096,
-              messages: [{ content: 'Hello', role: 'user' }],
+              messages: [
+                {
+                  content: [
+                    {
+                      cache_control: { type: 'ephemeral' },
+                      text: 'Hello',
+                      type: 'text',
+                    },
+                  ],
+                  role: 'user',
+                },
+              ],
               temperature: 0.4, // temperature / 2, top_p omitted due to conflict
             }),
             contentType: 'application/json',
@@ -450,7 +522,18 @@ describe('LobeBedrockAI', () => {
             body: JSON.stringify({
               anthropic_version: 'bedrock-2023-05-31',
               max_tokens: 4096,
-              messages: [{ content: 'Hello', role: 'user' }],
+              messages: [
+                {
+                  content: [
+                    {
+                      cache_control: { type: 'ephemeral' },
+                      text: 'Hello',
+                      type: 'text',
+                    },
+                  ],
+                  role: 'user',
+                },
+              ],
               top_p: 0.9, // temperature omitted since not provided
             }),
             contentType: 'application/json',
@@ -483,7 +566,18 @@ describe('LobeBedrockAI', () => {
             body: JSON.stringify({
               anthropic_version: 'bedrock-2023-05-31',
               max_tokens: 4096,
-              messages: [{ content: 'Hello', role: 'user' }],
+              messages: [
+                {
+                  content: [
+                    {
+                      cache_control: { type: 'ephemeral' },
+                      text: 'Hello',
+                      type: 'text',
+                    },
+                  ],
+                  role: 'user',
+                },
+              ],
               temperature: 0.4, // temperature / 2
               top_p: 0.9, // both parameters allowed for older models
             }),
@@ -517,7 +611,18 @@ describe('LobeBedrockAI', () => {
             body: JSON.stringify({
               anthropic_version: 'bedrock-2023-05-31',
               max_tokens: 4096,
-              messages: [{ content: 'Hello', role: 'user' }],
+              messages: [
+                {
+                  content: [
+                    {
+                      cache_control: { type: 'ephemeral' },
+                      text: 'Hello',
+                      type: 'text',
+                    },
+                  ],
+                  role: 'user',
+                },
+              ],
               temperature: 0.3, // temperature / 2, top_p omitted due to conflict
             }),
             contentType: 'application/json',
@@ -550,7 +655,18 @@ describe('LobeBedrockAI', () => {
             body: JSON.stringify({
               anthropic_version: 'bedrock-2023-05-31',
               max_tokens: 4096,
-              messages: [{ content: 'Hello', role: 'user' }],
+              messages: [
+                {
+                  content: [
+                    {
+                      cache_control: { type: 'ephemeral' },
+                      text: 'Hello',
+                      type: 'text',
+                    },
+                  ],
+                  role: 'user',
+                },
+              ],
               temperature: 0.35, // temperature / 2, top_p omitted due to conflict
             }),
             contentType: 'application/json',

package/packages/model-runtime/src/providers/bedrock/index.ts

@@ -23,7 +23,9 @@ import {
 import { AgentRuntimeErrorType } from '../../types/error';
 import { AgentRuntimeError } from '../../utils/createError';
 import { debugStream } from '../../utils/debugStream';
+import { getModelPricing } from '../../utils/getModelPricing';
 import { StreamingResponse } from '../../utils/response';
+import { resolveCacheTTL } from '../anthropic/resolveCacheTTL';
 
 /**
  * A prompt constructor for HuggingFace LLama 2 chat models.
@@ -148,7 +150,16 @@ export class LobeBedrockAI implements LobeRuntimeAI {
     payload: ChatStreamPayload,
     options?: ChatMethodOptions,
   ): Promise<Response> => {
-    const { max_tokens, messages, model, temperature, top_p, tools } = payload;
+    const {
+      enabledContextCaching = true,
+      max_tokens,
+      messages,
+      model,
+      temperature,
+      top_p,
+      tools,
+    } = payload;
+    const inputStartAt = Date.now();
     const system_message = messages.find((m) => m.role === 'system');
     const user_messages = messages.filter((m) => m.role !== 'system');
 
@@ -159,17 +170,29 @@ export class LobeBedrockAI implements LobeRuntimeAI {
       { hasConflict, normalizeTemperature: true, preferTemperature: true },
     );
 
+    const systemPrompts = !!system_message?.content
+      ? ([
+          {
+            cache_control: enabledContextCaching ? { type: 'ephemeral' } : undefined,
+            text: system_message.content as string,
+            type: 'text',
+          },
+        ] as any)
+      : undefined;
+
+    const anthropicPayload = {
+      anthropic_version: 'bedrock-2023-05-31',
+      max_tokens: max_tokens || 4096,
+      messages: await buildAnthropicMessages(user_messages, { enabledContextCaching }),
+      system: systemPrompts,
+      temperature: resolvedParams.temperature,
+      tools: buildAnthropicTools(tools, { enabledContextCaching }),
+      top_p: resolvedParams.top_p,
+    };
+
     const command = new InvokeModelWithResponseStreamCommand({
       accept: 'application/json',
-      body: JSON.stringify({
-        anthropic_version: 'bedrock-2023-05-31',
-        max_tokens: max_tokens || 4096,
-        messages: await buildAnthropicMessages(user_messages),
-        system: system_message?.content as string,
-        temperature: resolvedParams.temperature,
-        tools: buildAnthropicTools(tools),
-        top_p: resolvedParams.top_p,
-      }),
+      body: JSON.stringify(anthropicPayload),
       contentType: 'application/json',
       modelId: model,
     });
@@ -186,10 +209,21 @@ export class LobeBedrockAI implements LobeRuntimeAI {
         debugStream(debug).catch(console.error);
       }
 
+      const pricing = await getModelPricing(payload.model, ModelProvider.Bedrock);
+      const cacheTTL = resolveCacheTTL({ ...payload, enabledContextCaching }, anthropicPayload);
+      const pricingOptions = cacheTTL ? { lookupParams: { ttl: cacheTTL } } : undefined;
+
       // Respond with the stream
-      return StreamingResponse(AWSBedrockClaudeStream(prod, options?.callback), {
-        headers: options?.headers,
-      });
+      return StreamingResponse(
+        AWSBedrockClaudeStream(prod, {
+          callbacks: options?.callback,
+          inputStartAt,
+          payload: { model, pricing, pricingOptions, provider: ModelProvider.Bedrock },
+        }),
+        {
+          headers: options?.headers,
+        },
+      );
     } catch (e) {
       const err = e as Error & { $metadata: any };
 

package/packages/model-runtime/src/providers/google/createImage.ts

@@ -147,6 +147,7 @@ async function generateImageByChatModel(
       ? {
           imageConfig: {
             aspectRatio: params.aspectRatio,
+            imageSize: params.resolution,
           },
         }
       : {}),

package/packages/model-runtime/src/providers/google/index.ts

@@ -38,6 +38,8 @@ const modelsWithModalities = new Set([
   'gemini-2.0-flash-preview-image-generation',
   'gemini-2.5-flash-image-preview',
   'gemini-2.5-flash-image',
+  'gemini-3-pro-image-preview',
+  'nano-banana-pro-preview',
 ]);
 
 const modelsDisableInstuction = new Set([
@@ -51,6 +53,11 @@ const modelsDisableInstuction = new Set([
   'gemma-3-12b-it',
   'gemma-3-27b-it',
   'gemma-3n-e4b-it',
+  // ZenMux
+  'google/gemini-2.5-flash-image-free',
+  'google/gemini-2.5-flash-image',
+  'google/gemini-3-pro-image-preview-free',
+  'google/gemini-3-pro-image-preview',
 ]);
 
 const PRO_THINKING_MIN = 128;
@@ -203,7 +210,10 @@ export class LobeGoogleAI implements LobeRuntimeAI {
         includeThoughts:
           (!!thinkingBudget ||
             !!thinkingLevel ||
-            (model && (model.includes('-3-pro-image') || model.includes('thinking')))) &&
+            (model &&
+              (model.includes('-3-pro-image') ||
+                model.includes('nano-banana-pro') ||
+                model.includes('thinking')))) &&
           resolvedThinkingBudget !== 0
             ? true
             : undefined,