koishi-plugin-chatluna-google-gemini-adapter 1.3.0-alpha.13 → 1.3.0-alpha.15

package/lib/index.cjs

@@ -23,14 +23,14 @@ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: tru
 // src/locales/zh-CN.schema.yml
 var require_zh_CN_schema = __commonJS({
   "src/locales/zh-CN.schema.yml"(exports2, module2) {
-    module2.exports = { $inner: [{}, { $desc: "请求选项", platform: "适配器的平台名。（不懂请不要修改）", apiKeys: { $inner: ["Gemini 的 API Key", "Gemini API 的请求地址"], $desc: "Gemini 的 API Key 和请求地址列表。" } }, { $desc: "模型配置", maxContextRatio: "最大上下文使用比例（0~1），控制可用的模型上下文窗口大小的最大百分比。例如 0.35 表示最多使用模型上下文的 35%。", temperature: "回复的随机性程度，数值越高，回复越随机（范围：0~2）。", googleSearch: "为模型启用谷歌搜索。", thinkingBudget: "思考预算，范围：(-1~24576)，设置的数值越大，思考时花费的 Token 越多,-1 为动态思考。目前仅支持 gemini 2.5 系列模型。", groundingContentDisplay: "是否显示谷歌搜索结果。", imageGeneration: "为模型启用图像生成。目前仅支持 `gemini-2.0-flash-exp` 和 `gemini-2.5-flash-image-preview` 模型。", searchThreshold: "搜索的[置信度阈值](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval)，范围：0~1，设置的数值越低，则越倾向于使用谷歌搜索。（仅支持 `gemini-1.5` 系列模型。gemini 2.0 模型起使用动态的工具调用）", includeThoughts: "是否获取模型的思考内容。", codeExecution: "为模型启用代码执行工具。", urlContext: "为模型启用 URL 内容获取工具。", nonStreaming: "强制不启用流式返回。开启后，将总是以非流式发起请求，即便配置了 stream 参数。" }] };
+    module2.exports = { $inner: [{}, { $desc: "请求选项", platform: "适配器的平台名。（不懂请不要修改）", apiKeys: { $inner: ["Gemini API Key", "Gemini API 请求地址", "是否启用此配置"], $desc: "Gemini 的 API Key 和请求地址列表。" } }, { $desc: "模型配置", maxContextRatio: "最大上下文使用比例（0~1），控制可用的模型上下文窗口大小的最大百分比。例如 0.35 表示最多使用模型上下文的 35%。", temperature: "回复的随机性程度，数值越高，回复越随机（范围：0~2）。", googleSearch: "为模型启用谷歌搜索。", thinkingBudget: "思考预算，范围：(-1~24576)，设置的数值越大，思考时花费的 Token 越多,-1 为动态思考。目前仅支持 gemini 2.5 系列模型。", groundingContentDisplay: "是否显示谷歌搜索结果。", imageGeneration: "为模型启用图像生成。目前仅支持 `gemini-2.0-flash-exp` 和 `gemini-2.5-flash-image-preview` 模型。", searchThreshold: "搜索的[置信度阈值](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval)，范围：0~1，设置的数值越低，则越倾向于使用谷歌搜索。（仅支持 `gemini-1.5` 系列模型。gemini 2.0 模型起使用动态的工具调用）", includeThoughts: "是否获取模型的思考内容。", codeExecution: "为模型启用代码执行工具。", urlContext: "为模型启用 URL 内容获取工具。", nonStreaming: "强制不启用流式返回。开启后，将总是以非流式发起请求，即便配置了 stream 参数。" }] };
   }
 });
 
 // src/locales/en-US.schema.yml
 var require_en_US_schema = __commonJS({
   "src/locales/en-US.schema.yml"(exports2, module2) {
-    module2.exports = { $inner: [{}, { $desc: "API Configuration", platform: "Adapter platform name. (Do not modify if you do not understand)", apiKeys: { $inner: ["Gemini API Key", "Gemini API Endpoint (optional)"], $desc: "Gemini API access credentials" } }, { $desc: "Model Parameters", maxContextRatio: "Maximum context usage ratio (0-1). Controls the maximum percentage of model context window available for use. For example, 0.35 means at most 35% of the model context can be used.", temperature: "Sampling temperature (0-2). Higher: more random, Lower: more deterministic", googleSearch: "Enable Google search", thinkingBudget: "Thinking budget (-1-24576). (0: dynamic thinking) Higher: more tokens spent on thinking. Currently only supports `gemini-2.5` series models.", groundingContentDisplay: "Enable display of search results", imageGeneration: "Enable image generation (only for `gemini-2.0-flash-exp` and `gemini-2.5-flash-image-preview` model)", searchThreshold: "Search confidence [threshold](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval) (0-1). Lower: more likely to use Google search", includeThoughts: "Enable retrieval of model thoughts", codeExecution: "Enable code execution tool", urlContext: "Enable URL context retrieval tool", nonStreaming: "Force disable streaming response. When enabled, requests will always be made in non-streaming mode, even if the stream parameter is configured." }] };
+    module2.exports = { $inner: [{}, { $desc: "API Configuration", platform: "Adapter platform name. (Do not modify if you do not understand)", apiKeys: { $inner: ["Gemini API Key", "Gemini API Endpoint (optional)", "Enabled"], $desc: "Gemini API access credentials" } }, { $desc: "Model Parameters", maxContextRatio: "Maximum context usage ratio (0-1). Controls the maximum percentage of model context window available for use. For example, 0.35 means at most 35% of the model context can be used.", temperature: "Sampling temperature (0-2). Higher: more random, Lower: more deterministic", googleSearch: "Enable Google search", thinkingBudget: "Thinking budget (-1-24576). (0: dynamic thinking) Higher: more tokens spent on thinking. Currently only supports `gemini-2.5` series models.", groundingContentDisplay: "Enable display of search results", imageGeneration: "Enable image generation (only for `gemini-2.0-flash-exp` and `gemini-2.5-flash-image-preview` model)", searchThreshold: "Search confidence [threshold](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval) (0-1). Lower: more likely to use Google search", includeThoughts: "Enable retrieval of model thoughts", codeExecution: "Enable code execution tool", urlContext: "Enable URL context retrieval tool", nonStreaming: "Force disable streaming response. When enabled, requests will always be made in non-streaming mode, even if the stream parameter is configured." }] };
   }
 });
 
@@ -685,6 +685,18 @@ var GeminiRequester = class extends import_api.ModelRequester {
           return;
         }
         const transformValue = typeof chunk === "string" ? JSON.parse(chunk) : chunk;
+        if (transformValue.usageMetadata) {
+          const promptTokens = transformValue.usageMetadata.promptTokenCount;
+          const totalTokens = transformValue.usageMetadata.totalTokenCount;
+          const completionTokens = transformValue.usageMetadata.candidatesTokenCount ?? totalTokens - promptTokens;
+          controller.enqueue({
+            usage: {
+              promptTokens,
+              completionTokens,
+              totalTokens
+            }
+          });
+        }
         if (!transformValue?.candidates) {
           return;
         }
@@ -723,6 +735,20 @@ var GeminiRequester = class extends import_api.ModelRequester {
     let errorCount = 0;
     let functionIndex = 0;
     for await (const chunk of iterable) {
+      let parsedChunk;
+      if (parsedChunk = partAsTypeCheck(
+        chunk,
+        (chunk2) => chunk2["usage"] != null
+      )) {
+        const generationChunk = new import_outputs.ChatGenerationChunk({
+          message: new import_messages.AIMessageChunk(""),
+          text: "",
+          generationInfo: {
+            tokenUsage: parsedChunk.usage
+          }
+        });
+        yield { type: "generation", generation: generationChunk };
+      }
       try {
         const { updatedContent, updatedReasoning, updatedToolCalling } = await this._processChunk(
           chunk,
@@ -998,11 +1024,13 @@ var import_logger2 = require("koishi-plugin-chatluna/utils/logger");
 var logger;
 var reusable = true;
 function apply(ctx, config) {
-  const plugin = new import_chat.ChatLunaPlugin(ctx, config, config.platform);
   logger = (0, import_logger2.createLogger)(ctx, "chatluna-gemini-adapter");
   ctx.on("ready", async () => {
+    const plugin = new import_chat.ChatLunaPlugin(ctx, config, config.platform);
     plugin.parseConfig((config2) => {
-      return config2.apiKeys.map(([apiKey, apiEndpoint]) => {
+      return config2.apiKeys.filter(([apiKey, _, enabled]) => {
+        return apiKey.length > 0 && enabled;
+      }).map(([apiKey, apiEndpoint]) => {
         return {
           apiKey,
           apiEndpoint,
@@ -1025,12 +1053,13 @@ var Config4 = import_koishi.Schema.intersect([
     platform: import_koishi.Schema.string().default("gemini"),
     apiKeys: import_koishi.Schema.array(
       import_koishi.Schema.tuple([
-        import_koishi.Schema.string().role("secret"),
+        import_koishi.Schema.string().role("secret").required(true),
         import_koishi.Schema.string().default(
           "https://generativelanguage.googleapis.com/v1beta"
-        )
+        ),
+        import_koishi.Schema.boolean().default(true)
       ])
-    ).default([["", "https://generativelanguage.googleapis.com/v1beta"]])
+    ).default([[]]).role("table")
   }),
   import_koishi.Schema.object({
     maxContextRatio: import_koishi.Schema.number().min(0).max(1).step(1e-4).role("slider").default(0.35),

package/lib/index.d.ts

@@ -4,7 +4,7 @@ export declare let logger: Logger;
 export declare const reusable = true;
 export declare function apply(ctx: Context, config: Config): void;
 export interface Config extends ChatLunaPlugin.Config {
-    apiKeys: [string, string][];
+    apiKeys: [string, string, boolean][];
     maxContextRatio: number;
     platform: string;
     temperature: number;

package/lib/index.mjs

@@ -8,14 +8,14 @@ var __commonJS = (cb, mod) => function __require() {
 // src/locales/zh-CN.schema.yml
 var require_zh_CN_schema = __commonJS({
   "src/locales/zh-CN.schema.yml"(exports, module) {
-    module.exports = { $inner: [{}, { $desc: "请求选项", platform: "适配器的平台名。（不懂请不要修改）", apiKeys: { $inner: ["Gemini 的 API Key", "Gemini API 的请求地址"], $desc: "Gemini 的 API Key 和请求地址列表。" } }, { $desc: "模型配置", maxContextRatio: "最大上下文使用比例（0~1），控制可用的模型上下文窗口大小的最大百分比。例如 0.35 表示最多使用模型上下文的 35%。", temperature: "回复的随机性程度，数值越高，回复越随机（范围：0~2）。", googleSearch: "为模型启用谷歌搜索。", thinkingBudget: "思考预算，范围：(-1~24576)，设置的数值越大，思考时花费的 Token 越多,-1 为动态思考。目前仅支持 gemini 2.5 系列模型。", groundingContentDisplay: "是否显示谷歌搜索结果。", imageGeneration: "为模型启用图像生成。目前仅支持 `gemini-2.0-flash-exp` 和 `gemini-2.5-flash-image-preview` 模型。", searchThreshold: "搜索的[置信度阈值](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval)，范围：0~1，设置的数值越低，则越倾向于使用谷歌搜索。（仅支持 `gemini-1.5` 系列模型。gemini 2.0 模型起使用动态的工具调用）", includeThoughts: "是否获取模型的思考内容。", codeExecution: "为模型启用代码执行工具。", urlContext: "为模型启用 URL 内容获取工具。", nonStreaming: "强制不启用流式返回。开启后，将总是以非流式发起请求，即便配置了 stream 参数。" }] };
+    module.exports = { $inner: [{}, { $desc: "请求选项", platform: "适配器的平台名。（不懂请不要修改）", apiKeys: { $inner: ["Gemini API Key", "Gemini API 请求地址", "是否启用此配置"], $desc: "Gemini 的 API Key 和请求地址列表。" } }, { $desc: "模型配置", maxContextRatio: "最大上下文使用比例（0~1），控制可用的模型上下文窗口大小的最大百分比。例如 0.35 表示最多使用模型上下文的 35%。", temperature: "回复的随机性程度，数值越高，回复越随机（范围：0~2）。", googleSearch: "为模型启用谷歌搜索。", thinkingBudget: "思考预算，范围：(-1~24576)，设置的数值越大，思考时花费的 Token 越多,-1 为动态思考。目前仅支持 gemini 2.5 系列模型。", groundingContentDisplay: "是否显示谷歌搜索结果。", imageGeneration: "为模型启用图像生成。目前仅支持 `gemini-2.0-flash-exp` 和 `gemini-2.5-flash-image-preview` 模型。", searchThreshold: "搜索的[置信度阈值](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval)，范围：0~1，设置的数值越低，则越倾向于使用谷歌搜索。（仅支持 `gemini-1.5` 系列模型。gemini 2.0 模型起使用动态的工具调用）", includeThoughts: "是否获取模型的思考内容。", codeExecution: "为模型启用代码执行工具。", urlContext: "为模型启用 URL 内容获取工具。", nonStreaming: "强制不启用流式返回。开启后，将总是以非流式发起请求，即便配置了 stream 参数。" }] };
   }
 });
 
 // src/locales/en-US.schema.yml
 var require_en_US_schema = __commonJS({
   "src/locales/en-US.schema.yml"(exports, module) {
-    module.exports = { $inner: [{}, { $desc: "API Configuration", platform: "Adapter platform name. (Do not modify if you do not understand)", apiKeys: { $inner: ["Gemini API Key", "Gemini API Endpoint (optional)"], $desc: "Gemini API access credentials" } }, { $desc: "Model Parameters", maxContextRatio: "Maximum context usage ratio (0-1). Controls the maximum percentage of model context window available for use. For example, 0.35 means at most 35% of the model context can be used.", temperature: "Sampling temperature (0-2). Higher: more random, Lower: more deterministic", googleSearch: "Enable Google search", thinkingBudget: "Thinking budget (-1-24576). (0: dynamic thinking) Higher: more tokens spent on thinking. Currently only supports `gemini-2.5` series models.", groundingContentDisplay: "Enable display of search results", imageGeneration: "Enable image generation (only for `gemini-2.0-flash-exp` and `gemini-2.5-flash-image-preview` model)", searchThreshold: "Search confidence [threshold](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval) (0-1). Lower: more likely to use Google search", includeThoughts: "Enable retrieval of model thoughts", codeExecution: "Enable code execution tool", urlContext: "Enable URL context retrieval tool", nonStreaming: "Force disable streaming response. When enabled, requests will always be made in non-streaming mode, even if the stream parameter is configured." }] };
+    module.exports = { $inner: [{}, { $desc: "API Configuration", platform: "Adapter platform name. (Do not modify if you do not understand)", apiKeys: { $inner: ["Gemini API Key", "Gemini API Endpoint (optional)", "Enabled"], $desc: "Gemini API access credentials" } }, { $desc: "Model Parameters", maxContextRatio: "Maximum context usage ratio (0-1). Controls the maximum percentage of model context window available for use. For example, 0.35 means at most 35% of the model context can be used.", temperature: "Sampling temperature (0-2). Higher: more random, Lower: more deterministic", googleSearch: "Enable Google search", thinkingBudget: "Thinking budget (-1-24576). (0: dynamic thinking) Higher: more tokens spent on thinking. Currently only supports `gemini-2.5` series models.", groundingContentDisplay: "Enable display of search results", imageGeneration: "Enable image generation (only for `gemini-2.0-flash-exp` and `gemini-2.5-flash-image-preview` model)", searchThreshold: "Search confidence [threshold](https://ai.google.dev/gemini-api/docs/grounding?lang=rest#dynamic-retrieval) (0-1). Lower: more likely to use Google search", includeThoughts: "Enable retrieval of model thoughts", codeExecution: "Enable code execution tool", urlContext: "Enable URL context retrieval tool", nonStreaming: "Force disable streaming response. When enabled, requests will always be made in non-streaming mode, even if the stream parameter is configured." }] };
   }
 });
 
@@ -682,6 +682,18 @@ var GeminiRequester = class extends ModelRequester {
           return;
         }
         const transformValue = typeof chunk === "string" ? JSON.parse(chunk) : chunk;
+        if (transformValue.usageMetadata) {
+          const promptTokens = transformValue.usageMetadata.promptTokenCount;
+          const totalTokens = transformValue.usageMetadata.totalTokenCount;
+          const completionTokens = transformValue.usageMetadata.candidatesTokenCount ?? totalTokens - promptTokens;
+          controller.enqueue({
+            usage: {
+              promptTokens,
+              completionTokens,
+              totalTokens
+            }
+          });
+        }
         if (!transformValue?.candidates) {
           return;
         }
@@ -720,6 +732,20 @@ var GeminiRequester = class extends ModelRequester {
     let errorCount = 0;
     let functionIndex = 0;
     for await (const chunk of iterable) {
+      let parsedChunk;
+      if (parsedChunk = partAsTypeCheck(
+        chunk,
+        (chunk2) => chunk2["usage"] != null
+      )) {
+        const generationChunk = new ChatGenerationChunk({
+          message: new AIMessageChunk(""),
+          text: "",
+          generationInfo: {
+            tokenUsage: parsedChunk.usage
+          }
+        });
+        yield { type: "generation", generation: generationChunk };
+      }
       try {
         const { updatedContent, updatedReasoning, updatedToolCalling } = await this._processChunk(
           chunk,
@@ -995,11 +1021,13 @@ import { createLogger } from "koishi-plugin-chatluna/utils/logger";
 var logger;
 var reusable = true;
 function apply(ctx, config) {
-  const plugin = new ChatLunaPlugin(ctx, config, config.platform);
   logger = createLogger(ctx, "chatluna-gemini-adapter");
   ctx.on("ready", async () => {
+    const plugin = new ChatLunaPlugin(ctx, config, config.platform);
     plugin.parseConfig((config2) => {
-      return config2.apiKeys.map(([apiKey, apiEndpoint]) => {
+      return config2.apiKeys.filter(([apiKey, _, enabled]) => {
+        return apiKey.length > 0 && enabled;
+      }).map(([apiKey, apiEndpoint]) => {
         return {
           apiKey,
           apiEndpoint,
@@ -1022,12 +1050,13 @@ var Config4 = Schema.intersect([
     platform: Schema.string().default("gemini"),
     apiKeys: Schema.array(
       Schema.tuple([
-        Schema.string().role("secret"),
+        Schema.string().role("secret").required(true),
         Schema.string().default(
           "https://generativelanguage.googleapis.com/v1beta"
-        )
+        ),
+        Schema.boolean().default(true)
       ])
-    ).default([["", "https://generativelanguage.googleapis.com/v1beta"]])
+    ).default([[]]).role("table")
   }),
   Schema.object({
     maxContextRatio: Schema.number().min(0).max(1).step(1e-4).role("slider").default(0.35),

package/lib/types.d.ts

@@ -2,11 +2,18 @@ export interface ChatCompletionResponseMessage {
     role: string;
     parts?: ChatPart[];
 }
-export type ChatPart = ChatMessagePart | ChatInlineDataPart | ChatFunctionCallingPart | ChatFunctionResponsePart | ChatUploadDataPart;
+export type ChatPart = ChatMessagePart | ChatInlineDataPart | ChatFunctionCallingPart | ChatFunctionResponsePart | ChatUploadDataPart | ChatUsageMetadataPart;
 export type ChatMessagePart = {
     text: string;
     thought?: boolean;
 };
+export type ChatUsageMetadataPart = {
+    usage: {
+        promptTokens: number;
+        completionTokens: number;
+        totalTokens: number;
+    };
+};
 export type ChatInlineDataPart = {
     inlineData: {
         mimeType: string;

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "koishi-plugin-chatluna-google-gemini-adapter",
   "description": "google-gemini adapter for chatluna",
-  "version": "1.3.0-alpha.13",
+  "version": "1.3.0-alpha.15",
   "main": "lib/index.cjs",
   "module": "lib/index.mjs",
   "typings": "lib/index.d.ts",
@@ -62,7 +62,7 @@
     "adapter"
   ],
   "dependencies": {
-    "@chatluna/v1-shared-adapter": "^1.0.11",
+    "@chatluna/v1-shared-adapter": "^1.0.13",
     "@langchain/core": "0.3.62",
     "zod": "3.25.76",
     "zod-to-json-schema": "^3.24.6"
@@ -73,7 +73,7 @@
   },
   "peerDependencies": {
     "koishi": "^4.18.9",
-    "koishi-plugin-chatluna": "^1.3.0-alpha.49",
+    "koishi-plugin-chatluna": "^1.3.0-alpha.56",
     "koishi-plugin-chatluna-storage-service": "^0.0.9"
   },
   "peerDependenciesMeta": {