koishi-plugin-chatluna-google-gemini-adapter 1.0.0-beta.16 → 1.0.0-beta.19

package/lib/index.cjs

@@ -404,24 +404,33 @@ var GeminiRequester = class extends import_api.ModelRequester {
   }
   async embeddings(params) {
     let data;
+    if (typeof params.input === "string") {
+      params.input = [params.input];
+    }
     try {
       const response = await this._post(
-        `models/${params.model}:embedContent`,
+        `models/${params.model}:batchEmbedContents`,
         {
-          model: `models/${params.model}`,
-          content: {
-            parts: [
-              {
-                text: params.input
+          requests: params.input.map((input) => {
+            return {
+              model: `models/${params.model}`,
+              content: {
+                parts: [
+                  {
+                    text: input
+                  }
+                ]
               }
-            ]
-          }
+            };
+          })
         }
       );
       data = await response.text();
       data = JSON.parse(data);
-      if (data.embedding && data.embedding.values?.length > 0) {
-        return data.embedding.values;
+      if (data.embeddings && data.embeddings.length > 0) {
+        return data.embeddings.map((embedding) => {
+          return embedding.values;
+        });
       }
       throw new Error(
         "error when calling gemini embeddings, Result: " + JSON.stringify(data)
@@ -529,7 +538,18 @@ var GeminiClient = class extends import_client.PlatformModelAndEmbeddingsClient
       return rawModels.map((model) => model.replace("models/", "")).map((model) => {
         return {
           name: model,
-          maxTokens: model.includes("vision") ? 12288 : 30720,
+          maxTokens: ((model2) => {
+            if (model2.includes("gemini-1.5-pro")) {
+              return 1048576;
+            }
+            if (model2.includes("gemini-1.5-flash")) {
+              return 2097152;
+            }
+            if (model2.includes("gemini-1.0-pro")) {
+              return 30720;
+            }
+            return 30720;
+          })(model),
           type: model.includes("embedding") ? import_types.ModelType.embeddings : import_types.ModelType.llm,
           functionCall: !model.includes("vision"),
           supportMode: ["all"]
@@ -559,6 +579,7 @@ var GeminiClient = class extends import_client.PlatformModelAndEmbeddingsClient
         modelInfo: info,
         requester: this._requester,
         model,
+        modelMaxContextSize: info.maxTokens,
         maxTokens: this._config.maxTokens,
         timeout: this._config.timeout,
         temperature: this._config.temperature,
@@ -614,8 +635,8 @@ var Config = import_koishi.Schema.intersect([
   }).description("请求设置"),
   import_koishi.Schema.object({
     maxTokens: import_koishi.Schema.number().description(
-      "回复的最大 Token 数（16~32800，必须是16的倍数）（注意如果你目前使用的模型的最大 Token 为 8000 及以上的话才建议设置超过 512 token）"
-    ).min(16).max(128e3).step(16).default(1024),
+      "回复的最大 Token 数（16~2097000，必须是16的倍数）（注意如果你目前使用的模型的最大 Token 为 8000 及以上的话才建议设置超过 512 token）"
+    ).min(16).max(2097e3).step(16).default(8064),
     temperature: import_koishi.Schema.percent().description("回复温度，越高越随机").min(0).max(1).step(0.1).default(0.8)
   }).description("模型设置")
   // eslint-disable-next-line @typescript-eslint/no-explicit-any

package/lib/index.d.ts

@@ -73,9 +73,9 @@ export interface ChatCompletionMessageFunctionCall {
     args?: any;
 }
 export interface CreateEmbeddingResponse {
-    embedding: {
+    embeddings: {
         values: number[];
-    };
+    }[];
 }
 export type ChatCompletionResponseMessageRoleEnum = 'system' | 'model' | 'user' | 'function';
 export function langchainMessageToGeminiMessage(messages: BaseMessage[], model?: string): Promise<ChatCompletionResponseMessage[]>;

package/lib/index.mjs

@@ -397,24 +397,33 @@ var GeminiRequester = class extends ModelRequester {
   }
   async embeddings(params) {
     let data;
+    if (typeof params.input === "string") {
+      params.input = [params.input];
+    }
     try {
       const response = await this._post(
-        `models/${params.model}:embedContent`,
+        `models/${params.model}:batchEmbedContents`,
         {
-          model: `models/${params.model}`,
-          content: {
-            parts: [
-              {
-                text: params.input
+          requests: params.input.map((input) => {
+            return {
+              model: `models/${params.model}`,
+              content: {
+                parts: [
+                  {
+                    text: input
+                  }
+                ]
               }
-            ]
-          }
+            };
+          })
         }
       );
       data = await response.text();
       data = JSON.parse(data);
-      if (data.embedding && data.embedding.values?.length > 0) {
-        return data.embedding.values;
+      if (data.embeddings && data.embeddings.length > 0) {
+        return data.embeddings.map((embedding) => {
+          return embedding.values;
+        });
       }
       throw new Error(
         "error when calling gemini embeddings, Result: " + JSON.stringify(data)
@@ -522,7 +531,18 @@ var GeminiClient = class extends PlatformModelAndEmbeddingsClient {
       return rawModels.map((model) => model.replace("models/", "")).map((model) => {
         return {
           name: model,
-          maxTokens: model.includes("vision") ? 12288 : 30720,
+          maxTokens: ((model2) => {
+            if (model2.includes("gemini-1.5-pro")) {
+              return 1048576;
+            }
+            if (model2.includes("gemini-1.5-flash")) {
+              return 2097152;
+            }
+            if (model2.includes("gemini-1.0-pro")) {
+              return 30720;
+            }
+            return 30720;
+          })(model),
           type: model.includes("embedding") ? ModelType.embeddings : ModelType.llm,
           functionCall: !model.includes("vision"),
           supportMode: ["all"]
@@ -552,6 +572,7 @@ var GeminiClient = class extends PlatformModelAndEmbeddingsClient {
         modelInfo: info,
         requester: this._requester,
         model,
+        modelMaxContextSize: info.maxTokens,
         maxTokens: this._config.maxTokens,
         timeout: this._config.timeout,
         temperature: this._config.temperature,
@@ -607,8 +628,8 @@ var Config = Schema.intersect([
   }).description("请求设置"),
   Schema.object({
     maxTokens: Schema.number().description(
-      "回复的最大 Token 数（16~32800，必须是16的倍数）（注意如果你目前使用的模型的最大 Token 为 8000 及以上的话才建议设置超过 512 token）"
-    ).min(16).max(128e3).step(16).default(1024),
+      "回复的最大 Token 数（16~2097000，必须是16的倍数）（注意如果你目前使用的模型的最大 Token 为 8000 及以上的话才建议设置超过 512 token）"
+    ).min(16).max(2097e3).step(16).default(8064),
     temperature: Schema.percent().description("回复温度，越高越随机").min(0).max(1).step(0.1).default(0.8)
   }).description("模型设置")
   // eslint-disable-next-line @typescript-eslint/no-explicit-any

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "koishi-plugin-chatluna-google-gemini-adapter",
   "description": "google-gemini adapter for chatluna",
-  "version": "1.0.0-beta.16",
+  "version": "1.0.0-beta.19",
   "main": "lib/index.cjs",
   "module": "lib/index.mjs",
   "typings": "lib/index.d.ts",
@@ -48,18 +48,18 @@
     "adapter"
   ],
   "dependencies": {
-    "@langchain/core": "^0.2.12",
+    "@langchain/core": "^0.2.17",
     "@streamparser/json": "^0.0.21",
     "zod": "^3.24.0-canary.20240701T200529",
     "zod-to-json-schema": "^3.23.1"
   },
   "devDependencies": {
     "atsc": "^2.1.0",
-    "koishi": "^4.17.9"
+    "koishi": "^4.17.10"
   },
   "peerDependencies": {
     "koishi": "^4.17.9",
-    "koishi-plugin-chatluna": "^1.0.0-beta.56"
+    "koishi-plugin-chatluna": "^1.0.0-beta.62"
   },
   "koishi": {
     "description": {