@aigne/gemini 0.14.16-beta.13 → 0.14.16-beta.16

package/CHANGELOG.md

@@ -1,5 +1,49 @@
 # Changelog
 
+## [0.14.16-beta.16](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.14.16-beta.15...gemini-v0.14.16-beta.16) (2026-01-10)
+
+
+### Bug Fixes
+
+* **core:** simplify token-estimator logic for remaining characters ([45d43cc](https://github.com/AIGNE-io/aigne-framework/commit/45d43ccd3afd636cfb459eea2e6551e8f9c53765))
+
+
+### Dependencies
+
+* The following workspace dependencies were updated
+  * dependencies
+    * @aigne/core bumped to 1.72.0-beta.15
+  * devDependencies
+    * @aigne/test-utils bumped to 0.5.69-beta.15
+
+## [0.14.16-beta.15](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.14.16-beta.14...gemini-v0.14.16-beta.15) (2026-01-09)
+
+
+### Dependencies
+
+* The following workspace dependencies were updated
+  * dependencies
+    * @aigne/core bumped to 1.72.0-beta.14
+  * devDependencies
+    * @aigne/test-utils bumped to 0.5.69-beta.14
+
+## [0.14.16-beta.14](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.14.16-beta.13...gemini-v0.14.16-beta.14) (2026-01-08)
+
+
+### Bug Fixes
+
+* bump version ([696560f](https://github.com/AIGNE-io/aigne-framework/commit/696560fa2673eddcb4d00ac0523fbbbde7273cb3))
+
+
+### Dependencies
+
+* The following workspace dependencies were updated
+  * dependencies
+    * @aigne/core bumped to 1.72.0-beta.13
+    * @aigne/platform-helpers bumped to 0.6.7-beta.1
+  * devDependencies
+    * @aigne/test-utils bumped to 0.5.69-beta.13
+
 ## [0.14.16-beta.13](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.14.16-beta.12...gemini-v0.14.16-beta.13) (2026-01-07)
 
 

package/lib/cjs/gemini-chat-model.d.ts

@@ -67,6 +67,8 @@ export declare class GeminiChatModel extends ChatModel {
             $get: string;
         } | undefined;
     }> | undefined;
+    countTokens(input: ChatModelInput): Promise<number>;
+    private contentUnionToContent;
     process(input: ChatModelInput, options: AgentInvokeOptions): PromiseOrValue<AgentProcessResult<ChatModelOutput>>;
     protected thinkingBudgetModelMap: ({
         pattern: RegExp;
@@ -104,6 +106,7 @@ export declare class GeminiChatModel extends ChatModel {
         budget?: number;
         level?: ThinkingLevel;
     };
+    private getParameters;
     private processInput;
     private buildConfig;
     private buildTools;

package/lib/cjs/gemini-chat-model.js

@@ -61,6 +61,40 @@ class GeminiChatModel extends core_1.ChatModel {
     get modelOptions() {
         return this.options?.modelOptions;
     }
+    async countTokens(input) {
+        const { model, ...request } = await this.getParameters(input);
+        const contents = [];
+        const { systemInstruction, tools } = request.config ?? {};
+        if (systemInstruction)
+            contents.push(this.contentUnionToContent(systemInstruction));
+        if (tools?.length)
+            contents.push({ role: "system", parts: [{ text: JSON.stringify(tools) }] });
+        contents.push(...[request.contents].flat().map(this.contentUnionToContent));
+        const tokens = (await this.googleClient.models.countTokens({
+            model,
+            contents,
+        })).totalTokens;
+        if (!(0, type_utils_js_1.isNil)(tokens))
+            return tokens;
+        return super.countTokens(input);
+    }
+    contentUnionToContent(content) {
+        if (typeof content === "object" && "parts" in content) {
+            return { role: "system", parts: content.parts };
+        }
+        else if (typeof content === "string") {
+            return { role: "system", parts: [{ text: content }] };
+        }
+        else if (Array.isArray(content)) {
+            return {
+                role: "system",
+                parts: content.map((i) => (typeof i === "string" ? { text: i } : i)),
+            };
+        }
+        else {
+            return { role: "system", parts: [content] };
+        }
+    }
     process(input, options) {
         return this.processInput(input, options);
     }
@@ -135,10 +169,10 @@ class GeminiChatModel extends core_1.ChatModel {
             budget = Math.min(m.max, budget);
         return { support: true, budget };
     }
-    async *processInput(input, options) {
+    async getParameters(input) {
         const { modelOptions = {} } = input;
         const model = modelOptions.model || this.credential.model;
-        const { contents, config } = await this.buildContents(input, options);
+        const { contents, config } = await this.buildContents(input);
         const thinkingBudget = this.getThinkingBudget(model, modelOptions.reasoningEffort);
         const parameters = {
             model,
@@ -160,6 +194,10 @@ class GeminiChatModel extends core_1.ChatModel {
                 ...(await this.buildConfig(input)),
             },
         };
+        return parameters;
+    }
+    async *processInput(input, options) {
+        const parameters = await this.getParameters(input);
         const response = await this.googleClient.models.generateContentStream(parameters);
         let usage = {
             inputTokens: 0,
@@ -211,7 +249,7 @@ class GeminiChatModel extends core_1.ChatModel {
                                     },
                                 };
                                 // Preserve thought_signature for 3.x models
-                                if (part.thoughtSignature && model.includes("gemini-3")) {
+                                if (part.thoughtSignature && parameters.model.includes("gemini-3")) {
                                     toolCall.metadata = {
                                         thoughtSignature: part.thoughtSignature,
                                     };
@@ -362,8 +400,8 @@ class GeminiChatModel extends core_1.ChatModel {
                         };
         return { tools, toolConfig: { functionCallingConfig } };
     }
-    async buildVideoContentParts(media, options) {
-        const { path: filePath, mimeType: fileMimeType } = await this.transformFileType("local", media, options);
+    async buildVideoContentParts(media) {
+        const { path: filePath, mimeType: fileMimeType } = await this.transformFileType("local", media);
         if (filePath) {
             const stats = await index_js_1.nodejs.fs.stat(filePath);
             const fileSizeInBytes = stats.size;
@@ -394,7 +432,7 @@ class GeminiChatModel extends core_1.ChatModel {
             }
         }
     }
-    async buildContents(input, options) {
+    async buildContents(input) {
         const result = {
             contents: [],
         };
@@ -472,12 +510,12 @@ class GeminiChatModel extends core_1.ChatModel {
                     functionResponse.response = toolResult;
                 }
                 else {
-                    functionResponse.parts = await this.contentToParts(msg.content, options);
+                    functionResponse.parts = await this.contentToParts(msg.content);
                 }
                 content.parts = [{ functionResponse }];
             }
             else if (msg.content) {
-                content.parts = await this.contentToParts(msg.content, options);
+                content.parts = await this.contentToParts(msg.content);
             }
             return content;
         }))).filter(type_utils_js_1.isNonNullable);
@@ -488,7 +526,7 @@ class GeminiChatModel extends core_1.ChatModel {
         }
         return result;
     }
-    async contentToParts(content, options) {
+    async contentToParts(content) {
         if (typeof content === "string")
             return [{ text: content }];
         return Promise.all(content.map(async (item) => {
@@ -498,7 +536,7 @@ class GeminiChatModel extends core_1.ChatModel {
                 case "url":
                     return { fileData: { fileUri: item.url, mimeType: item.mimeType } };
                 case "file": {
-                    const part = await this.buildVideoContentParts(item, options);
+                    const part = await this.buildVideoContentParts(item);
                     if (part)
                         return part;
                     return { inlineData: { data: item.data, mimeType: item.mimeType } };

package/lib/cjs/gemini-image-model.d.ts

@@ -28,7 +28,7 @@ export declare class GeminiImageModel extends ImageModel<GeminiImageModelInput,
      * @param input The input to process
      * @returns The generated response
      */
-    process(input: GeminiImageModelInput, options: AgentInvokeOptions): Promise<ImageModelOutput>;
+    process(input: GeminiImageModelInput, _options: AgentInvokeOptions): Promise<ImageModelOutput>;
     private generateImageByImagenModel;
     private generateImageByGeminiModel;
 }

package/lib/cjs/gemini-image-model.js

@@ -52,7 +52,7 @@ class GeminiImageModel extends core_1.ImageModel {
      * @param input The input to process
      * @returns The generated response
      */
-    async process(input, options) {
+    async process(input, _options) {
         const model = input.modelOptions?.model || this.credential.model;
         const responseFormat = input.responseFormat || "base64";
         if (responseFormat === "url") {
@@ -61,7 +61,7 @@ class GeminiImageModel extends core_1.ImageModel {
         if (model.includes("imagen")) {
             return this.generateImageByImagenModel(input);
         }
-        return this.generateImageByGeminiModel(input, options);
+        return this.generateImageByGeminiModel(input);
     }
     async generateImageByImagenModel(input) {
         const model = input.modelOptions?.model || this.credential.model;
@@ -100,7 +100,7 @@ class GeminiImageModel extends core_1.ImageModel {
             model,
         };
     }
-    async generateImageByGeminiModel(input, options) {
+    async generateImageByGeminiModel(input) {
         const model = input.modelOptions?.model || this.credential.model;
         const mergedInput = { ...this.modelOptions, ...input.modelOptions, ...input };
         const inputKeys = [
@@ -135,7 +135,7 @@ class GeminiImageModel extends core_1.ImageModel {
             "imageConfig",
         ];
         const images = await Promise.all((0, type_utils_js_1.flat)(input.image).map(async (image) => {
-            const { data, mimeType } = await this.transformFileType("file", image, options);
+            const { data, mimeType } = await this.transformFileType("file", image);
             return { inlineData: { data, mimeType } };
         }));
         const response = await this.client.models.generateContent({

package/lib/cjs/gemini-video-model.js

@@ -88,7 +88,7 @@ class GeminiVideoModel extends core_1.VideoModel {
         if (mergedInput.personGeneration)
             config.personGeneration = mergedInput.personGeneration;
         if (mergedInput.lastFrame) {
-            config.lastFrame = await this.transformFileType("file", mergedInput.lastFrame, options).then((file) => {
+            config.lastFrame = await this.transformFileType("file", mergedInput.lastFrame).then((file) => {
                 return {
                     imageBytes: file.data,
                     mimeType: file.mimeType,
@@ -97,7 +97,7 @@ class GeminiVideoModel extends core_1.VideoModel {
         }
         if (mergedInput.referenceImages) {
             config.referenceImages = await Promise.all(mergedInput.referenceImages.map(async (image) => {
-                return await this.transformFileType("file", image, options).then((file) => {
+                return await this.transformFileType("file", image).then((file) => {
                     return {
                         image: {
                             imageBytes: file.data,
@@ -113,7 +113,7 @@ class GeminiVideoModel extends core_1.VideoModel {
             config,
         };
         if (mergedInput.image) {
-            params.image = await this.transformFileType("file", mergedInput.image, options).then((file) => {
+            params.image = await this.transformFileType("file", mergedInput.image).then((file) => {
                 return {
                     imageBytes: file.data,
                     mimeType: file.mimeType,

package/lib/dts/gemini-chat-model.d.ts

@@ -67,6 +67,8 @@ export declare class GeminiChatModel extends ChatModel {
             $get: string;
         } | undefined;
     }> | undefined;
+    countTokens(input: ChatModelInput): Promise<number>;
+    private contentUnionToContent;
     process(input: ChatModelInput, options: AgentInvokeOptions): PromiseOrValue<AgentProcessResult<ChatModelOutput>>;
     protected thinkingBudgetModelMap: ({
         pattern: RegExp;
@@ -104,6 +106,7 @@ export declare class GeminiChatModel extends ChatModel {
         budget?: number;
         level?: ThinkingLevel;
     };
+    private getParameters;
     private processInput;
     private buildConfig;
     private buildTools;

package/lib/dts/gemini-image-model.d.ts

@@ -28,7 +28,7 @@ export declare class GeminiImageModel extends ImageModel<GeminiImageModelInput,
      * @param input The input to process
      * @returns The generated response
      */
-    process(input: GeminiImageModelInput, options: AgentInvokeOptions): Promise<ImageModelOutput>;
+    process(input: GeminiImageModelInput, _options: AgentInvokeOptions): Promise<ImageModelOutput>;
     private generateImageByImagenModel;
     private generateImageByGeminiModel;
 }

package/lib/esm/gemini-chat-model.d.ts

@@ -67,6 +67,8 @@ export declare class GeminiChatModel extends ChatModel {
             $get: string;
         } | undefined;
     }> | undefined;
+    countTokens(input: ChatModelInput): Promise<number>;
+    private contentUnionToContent;
     process(input: ChatModelInput, options: AgentInvokeOptions): PromiseOrValue<AgentProcessResult<ChatModelOutput>>;
     protected thinkingBudgetModelMap: ({
         pattern: RegExp;
@@ -104,6 +106,7 @@ export declare class GeminiChatModel extends ChatModel {
         budget?: number;
         level?: ThinkingLevel;
     };
+    private getParameters;
     private processInput;
     private buildConfig;
     private buildTools;

package/lib/esm/gemini-chat-model.js

@@ -1,7 +1,7 @@
 import { agentProcessResultToObject, ChatModel, StructuredOutputError, safeParseJSON, } from "@aigne/core";
 import { logger } from "@aigne/core/utils/logger.js";
 import { mergeUsage } from "@aigne/core/utils/model-utils.js";
-import { isNonNullable, isRecord } from "@aigne/core/utils/type-utils.js";
+import { isNil, isNonNullable, isRecord, } from "@aigne/core/utils/type-utils.js";
 import { nodejs } from "@aigne/platform-helpers/nodejs/index.js";
 import { v7 } from "@aigne/uuid";
 import { createPartFromUri, createUserContent, FunctionCallingConfigMode, GoogleGenAI, ThinkingLevel, } from "@google/genai";
@@ -58,6 +58,40 @@ export class GeminiChatModel extends ChatModel {
     get modelOptions() {
         return this.options?.modelOptions;
     }
+    async countTokens(input) {
+        const { model, ...request } = await this.getParameters(input);
+        const contents = [];
+        const { systemInstruction, tools } = request.config ?? {};
+        if (systemInstruction)
+            contents.push(this.contentUnionToContent(systemInstruction));
+        if (tools?.length)
+            contents.push({ role: "system", parts: [{ text: JSON.stringify(tools) }] });
+        contents.push(...[request.contents].flat().map(this.contentUnionToContent));
+        const tokens = (await this.googleClient.models.countTokens({
+            model,
+            contents,
+        })).totalTokens;
+        if (!isNil(tokens))
+            return tokens;
+        return super.countTokens(input);
+    }
+    contentUnionToContent(content) {
+        if (typeof content === "object" && "parts" in content) {
+            return { role: "system", parts: content.parts };
+        }
+        else if (typeof content === "string") {
+            return { role: "system", parts: [{ text: content }] };
+        }
+        else if (Array.isArray(content)) {
+            return {
+                role: "system",
+                parts: content.map((i) => (typeof i === "string" ? { text: i } : i)),
+            };
+        }
+        else {
+            return { role: "system", parts: [content] };
+        }
+    }
     process(input, options) {
         return this.processInput(input, options);
     }
@@ -132,10 +166,10 @@ export class GeminiChatModel extends ChatModel {
             budget = Math.min(m.max, budget);
         return { support: true, budget };
     }
-    async *processInput(input, options) {
+    async getParameters(input) {
         const { modelOptions = {} } = input;
         const model = modelOptions.model || this.credential.model;
-        const { contents, config } = await this.buildContents(input, options);
+        const { contents, config } = await this.buildContents(input);
         const thinkingBudget = this.getThinkingBudget(model, modelOptions.reasoningEffort);
         const parameters = {
             model,
@@ -157,6 +191,10 @@ export class GeminiChatModel extends ChatModel {
                 ...(await this.buildConfig(input)),
             },
         };
+        return parameters;
+    }
+    async *processInput(input, options) {
+        const parameters = await this.getParameters(input);
         const response = await this.googleClient.models.generateContentStream(parameters);
         let usage = {
             inputTokens: 0,
@@ -208,7 +246,7 @@ export class GeminiChatModel extends ChatModel {
                                     },
                                 };
                                 // Preserve thought_signature for 3.x models
-                                if (part.thoughtSignature && model.includes("gemini-3")) {
+                                if (part.thoughtSignature && parameters.model.includes("gemini-3")) {
                                     toolCall.metadata = {
                                         thoughtSignature: part.thoughtSignature,
                                     };
@@ -359,8 +397,8 @@ export class GeminiChatModel extends ChatModel {
                         };
         return { tools, toolConfig: { functionCallingConfig } };
     }
-    async buildVideoContentParts(media, options) {
-        const { path: filePath, mimeType: fileMimeType } = await this.transformFileType("local", media, options);
+    async buildVideoContentParts(media) {
+        const { path: filePath, mimeType: fileMimeType } = await this.transformFileType("local", media);
         if (filePath) {
             const stats = await nodejs.fs.stat(filePath);
             const fileSizeInBytes = stats.size;
@@ -391,7 +429,7 @@ export class GeminiChatModel extends ChatModel {
             }
         }
     }
-    async buildContents(input, options) {
+    async buildContents(input) {
         const result = {
             contents: [],
         };
@@ -469,12 +507,12 @@ export class GeminiChatModel extends ChatModel {
                     functionResponse.response = toolResult;
                 }
                 else {
-                    functionResponse.parts = await this.contentToParts(msg.content, options);
+                    functionResponse.parts = await this.contentToParts(msg.content);
                 }
                 content.parts = [{ functionResponse }];
             }
             else if (msg.content) {
-                content.parts = await this.contentToParts(msg.content, options);
+                content.parts = await this.contentToParts(msg.content);
             }
             return content;
         }))).filter(isNonNullable);
@@ -485,7 +523,7 @@ export class GeminiChatModel extends ChatModel {
         }
         return result;
     }
-    async contentToParts(content, options) {
+    async contentToParts(content) {
         if (typeof content === "string")
             return [{ text: content }];
         return Promise.all(content.map(async (item) => {
@@ -495,7 +533,7 @@ export class GeminiChatModel extends ChatModel {
                 case "url":
                     return { fileData: { fileUri: item.url, mimeType: item.mimeType } };
                 case "file": {
-                    const part = await this.buildVideoContentParts(item, options);
+                    const part = await this.buildVideoContentParts(item);
                     if (part)
                         return part;
                     return { inlineData: { data: item.data, mimeType: item.mimeType } };

package/lib/esm/gemini-image-model.d.ts

@@ -28,7 +28,7 @@ export declare class GeminiImageModel extends ImageModel<GeminiImageModelInput,
      * @param input The input to process
      * @returns The generated response
      */
-    process(input: GeminiImageModelInput, options: AgentInvokeOptions): Promise<ImageModelOutput>;
+    process(input: GeminiImageModelInput, _options: AgentInvokeOptions): Promise<ImageModelOutput>;
     private generateImageByImagenModel;
     private generateImageByGeminiModel;
 }

package/lib/esm/gemini-image-model.js

@@ -49,7 +49,7 @@ export class GeminiImageModel extends ImageModel {
      * @param input The input to process
      * @returns The generated response
      */
-    async process(input, options) {
+    async process(input, _options) {
         const model = input.modelOptions?.model || this.credential.model;
         const responseFormat = input.responseFormat || "base64";
         if (responseFormat === "url") {
@@ -58,7 +58,7 @@ export class GeminiImageModel extends ImageModel {
         if (model.includes("imagen")) {
             return this.generateImageByImagenModel(input);
         }
-        return this.generateImageByGeminiModel(input, options);
+        return this.generateImageByGeminiModel(input);
     }
     async generateImageByImagenModel(input) {
         const model = input.modelOptions?.model || this.credential.model;
@@ -97,7 +97,7 @@ export class GeminiImageModel extends ImageModel {
             model,
         };
     }
-    async generateImageByGeminiModel(input, options) {
+    async generateImageByGeminiModel(input) {
         const model = input.modelOptions?.model || this.credential.model;
         const mergedInput = { ...this.modelOptions, ...input.modelOptions, ...input };
         const inputKeys = [
@@ -132,7 +132,7 @@ export class GeminiImageModel extends ImageModel {
             "imageConfig",
         ];
         const images = await Promise.all(flat(input.image).map(async (image) => {
-            const { data, mimeType } = await this.transformFileType("file", image, options);
+            const { data, mimeType } = await this.transformFileType("file", image);
             return { inlineData: { data, mimeType } };
         }));
         const response = await this.client.models.generateContent({

package/lib/esm/gemini-video-model.js

@@ -85,7 +85,7 @@ export class GeminiVideoModel extends VideoModel {
         if (mergedInput.personGeneration)
             config.personGeneration = mergedInput.personGeneration;
         if (mergedInput.lastFrame) {
-            config.lastFrame = await this.transformFileType("file", mergedInput.lastFrame, options).then((file) => {
+            config.lastFrame = await this.transformFileType("file", mergedInput.lastFrame).then((file) => {
                 return {
                     imageBytes: file.data,
                     mimeType: file.mimeType,
@@ -94,7 +94,7 @@ export class GeminiVideoModel extends VideoModel {
         }
         if (mergedInput.referenceImages) {
             config.referenceImages = await Promise.all(mergedInput.referenceImages.map(async (image) => {
-                return await this.transformFileType("file", image, options).then((file) => {
+                return await this.transformFileType("file", image).then((file) => {
                     return {
                         image: {
                             imageBytes: file.data,
@@ -110,7 +110,7 @@ export class GeminiVideoModel extends VideoModel {
             config,
         };
         if (mergedInput.image) {
-            params.image = await this.transformFileType("file", mergedInput.image, options).then((file) => {
+            params.image = await this.transformFileType("file", mergedInput.image).then((file) => {
                 return {
                     imageBytes: file.data,
                     mimeType: file.mimeType,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aigne/gemini",
-  "version": "0.14.16-beta.13",
+  "version": "0.14.16-beta.16",
   "description": "AIGNE Gemini SDK for integrating with Google's Gemini AI models",
   "publishConfig": {
     "access": "public"
@@ -40,8 +40,8 @@
     "yaml": "^2.8.1",
     "zod": "^3.25.67",
     "zod-to-json-schema": "^3.24.6",
-    "@aigne/core": "^1.72.0-beta.12",
-    "@aigne/platform-helpers": "^0.6.7-beta"
+    "@aigne/core": "^1.72.0-beta.15",
+    "@aigne/platform-helpers": "^0.6.7-beta.1"
   },
   "devDependencies": {
     "@types/bun": "^1.2.22",
@@ -49,7 +49,7 @@
     "npm-run-all": "^4.1.5",
     "rimraf": "^6.0.1",
     "typescript": "^5.9.2",
-    "@aigne/test-utils": "^0.5.69-beta.12"
+    "@aigne/test-utils": "^0.5.69-beta.15"
   },
   "scripts": {
     "lint": "tsc --noEmit",