@langchain/google-genai 0.0.25 → 0.1.0

package/README.md

@@ -5,7 +5,7 @@ This package contains the LangChain.js integrations for Gemini through their gen
 ## Installation
 
 ```bash npm2yarn
-npm install @langchain/google-genai
+npm install @langchain/google-genai @langchain/core
 ```
 
 This package, along with the main LangChain package, depends on [`@langchain/core`](https://npmjs.com/package/@langchain/core/).
@@ -17,18 +17,18 @@ You can do so by adding appropriate field to your project's `package.json` like
   "name": "your-project",
   "version": "0.0.0",
   "dependencies": {
-    "@langchain/google-genai": "^0.0.0",
-    "langchain": "0.0.207"
+    "@langchain/core": "^0.3.0",
+    "@langchain/google-genai": "^0.0.0"
   },
   "resolutions": {
-    "@langchain/core": "0.1.5"
+    "@langchain/core": "^0.3.0"
   },
   "overrides": {
-    "@langchain/core": "0.1.5"
+    "@langchain/core": "^0.3.0"
   },
   "pnpm": {
     "overrides": {
-      "@langchain/core": "0.1.5"
+      "@langchain/core": "^0.3.0"
     }
   }
 }

package/dist/chat_models.cjs

@@ -10,33 +10,339 @@ const zod_to_genai_parameters_js_1 = require("./utils/zod_to_genai_parameters.cj
 const common_js_1 = require("./utils/common.cjs");
 const output_parsers_js_1 = require("./output_parsers.cjs");
 /**
- * A class that wraps the Google Palm chat model.
- * @example
+ * Google Generative AI chat model integration.
+ *
+ * Setup:
+ * Install `@langchain/google-genai` and set an environment variable named `GOOGLE_API_KEY`.
+ *
+ * ```bash
+ * npm install @langchain/google-genai
+ * export GOOGLE_API_KEY="your-api-key"
+ * ```
+ *
+ * ## [Constructor args](https://api.js.langchain.com/classes/langchain_google_genai.ChatGoogleGenerativeAI.html#constructor)
+ *
+ * ## [Runtime args](https://api.js.langchain.com/interfaces/langchain_google_genai.GoogleGenerativeAIChatCallOptions.html)
+ *
+ * Runtime args can be passed as the second argument to any of the base runnable methods `.invoke`. `.stream`, `.batch`, etc.
+ * They can also be passed via `.bind`, or the second arg in `.bindTools`, like shown in the examples below:
+ *
+ * ```typescript
+ * // When calling `.bind`, call options should be passed via the first argument
+ * const llmWithArgsBound = llm.bind({
+ *   stop: ["\n"],
+ *   tools: [...],
+ * });
+ *
+ * // When calling `.bindTools`, call options should be passed via the second argument
+ * const llmWithTools = llm.bindTools(
+ *   [...],
+ *   {
+ *     stop: ["\n"],
+ *   }
+ * );
+ * ```
+ *
+ * ## Examples
+ *
+ * <details open>
+ * <summary><strong>Instantiate</strong></summary>
+ *
  * ```typescript
- * const model = new ChatGoogleGenerativeAI({
- *   apiKey: "<YOUR API KEY>",
- *   temperature: 0.7,
- *   modelName: "gemini-pro",
- *   topK: 40,
- *   topP: 1,
+ * import { ChatGoogleGenerativeAI } from '@langchain/google-genai';
+ *
+ * const llm = new ChatGoogleGenerativeAI({
+ *   model: "gemini-1.5-flash",
+ *   temperature: 0,
+ *   maxRetries: 2,
+ *   // apiKey: "...",
+ *   // other params...
  * });
- * const questions = [
- *   new HumanMessage({
- *     content: [
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Invoking</strong></summary>
+ *
+ * ```typescript
+ * const input = `Translate "I love programming" into French.`;
+ *
+ * // Models also accept a list of chat messages or a formatted prompt
+ * const result = await llm.invoke(input);
+ * console.log(result);
+ * ```
+ *
+ * ```txt
+ * AIMessage {
+ *   "content": "There are a few ways to translate \"I love programming\" into French, depending on the level of formality and nuance you want to convey:\n\n**Formal:**\n\n* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and your intended audience. \n",
+ *   "response_metadata": {
+ *     "finishReason": "STOP",
+ *     "index": 0,
+ *     "safetyRatings": [
+ *       {
+ *         "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+ *         "probability": "NEGLIGIBLE"
+ *       },
  *       {
- *         type: "text",
- *         text: "You are a funny assistant that answers in pirate language.",
+ *         "category": "HARM_CATEGORY_HATE_SPEECH",
+ *         "probability": "NEGLIGIBLE"
  *       },
  *       {
- *         type: "text",
- *         text: "What is your favorite food?",
+ *         "category": "HARM_CATEGORY_HARASSMENT",
+ *         "probability": "NEGLIGIBLE"
  *       },
+ *       {
+ *         "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+ *         "probability": "NEGLIGIBLE"
+ *       }
  *     ]
- *   })
- * ];
- * const res = await model.invoke(questions);
- * console.log({ res });
+ *   },
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 149,
+ *     "total_tokens": 159
+ *   }
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Streaming Chunks</strong></summary>
+ *
+ * ```typescript
+ * for await (const chunk of await llm.stream(input)) {
+ *   console.log(chunk);
+ * }
+ * ```
+ *
+ * ```txt
+ * AIMessageChunk {
+ *   "content": "There",
+ *   "response_metadata": {
+ *     "index": 0
+ *   }
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 1,
+ *     "total_tokens": 11
+ *   }
+ * }
+ * AIMessageChunk {
+ *   "content": " are a few ways to translate \"I love programming\" into French, depending on",
+ * }
+ * AIMessageChunk {
+ *   "content": " the level of formality and nuance you want to convey:\n\n**Formal:**\n\n",
+ * }
+ * AIMessageChunk {
+ *   "content": "* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This",
+ * }
+ * AIMessageChunk {
+ *   "content": " is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More",
+ * }
+ * AIMessageChunk {
+ *   "content": " specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and",
+ * }
+ * AIMessageChunk {
+ *   "content": " your intended audience. \n",
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Aggregate Streamed Chunks</strong></summary>
+ *
+ * ```typescript
+ * import { AIMessageChunk } from '@langchain/core/messages';
+ * import { concat } from '@langchain/core/utils/stream';
+ *
+ * const stream = await llm.stream(input);
+ * let full: AIMessageChunk | undefined;
+ * for await (const chunk of stream) {
+ *   full = !full ? chunk : concat(full, chunk);
+ * }
+ * console.log(full);
+ * ```
+ *
+ * ```txt
+ * AIMessageChunk {
+ *   "content": "There are a few ways to translate \"I love programming\" into French, depending on the level of formality and nuance you want to convey:\n\n**Formal:**\n\n* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and your intended audience. \n",
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 277,
+ *     "total_tokens": 287
+ *   }
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Bind tools</strong></summary>
+ *
+ * ```typescript
+ * import { z } from 'zod';
+ *
+ * const GetWeather = {
+ *   name: "GetWeather",
+ *   description: "Get the current weather in a given location",
+ *   schema: z.object({
+ *     location: z.string().describe("The city and state, e.g. San Francisco, CA")
+ *   }),
+ * }
+ *
+ * const GetPopulation = {
+ *   name: "GetPopulation",
+ *   description: "Get the current population in a given location",
+ *   schema: z.object({
+ *     location: z.string().describe("The city and state, e.g. San Francisco, CA")
+ *   }),
+ * }
+ *
+ * const llmWithTools = llm.bindTools([GetWeather, GetPopulation]);
+ * const aiMsg = await llmWithTools.invoke(
+ *   "Which city is hotter today and which is bigger: LA or NY?"
+ * );
+ * console.log(aiMsg.tool_calls);
+ * ```
+ *
+ * ```txt
+ * [
+ *   {
+ *     name: 'GetWeather',
+ *     args: { location: 'Los Angeles, CA' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetWeather',
+ *     args: { location: 'New York, NY' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetPopulation',
+ *     args: { location: 'Los Angeles, CA' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetPopulation',
+ *     args: { location: 'New York, NY' },
+ *     type: 'tool_call'
+ *   }
+ * ]
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Structured Output</strong></summary>
+ *
+ * ```typescript
+ * const Joke = z.object({
+ *   setup: z.string().describe("The setup of the joke"),
+ *   punchline: z.string().describe("The punchline to the joke"),
+ *   rating: z.number().optional().describe("How funny the joke is, from 1 to 10")
+ * }).describe('Joke to tell user.');
+ *
+ * const structuredLlm = llm.withStructuredOutput(Joke, { name: "Joke" });
+ * const jokeResult = await structuredLlm.invoke("Tell me a joke about cats");
+ * console.log(jokeResult);
+ * ```
+ *
+ * ```txt
+ * {
+ *   setup: "Why don\\'t cats play poker?",
+ *   punchline: "Why don\\'t cats play poker? Because they always have an ace up their sleeve!"
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Multimodal</strong></summary>
+ *
+ * ```typescript
+ * import { HumanMessage } from '@langchain/core/messages';
+ *
+ * const imageUrl = "https://example.com/image.jpg";
+ * const imageData = await fetch(imageUrl).then(res => res.arrayBuffer());
+ * const base64Image = Buffer.from(imageData).toString('base64');
+ *
+ * const message = new HumanMessage({
+ *   content: [
+ *     { type: "text", text: "describe the weather in this image" },
+ *     {
+ *       type: "image_url",
+ *       image_url: { url: `data:image/jpeg;base64,${base64Image}` },
+ *     },
+ *   ]
+ * });
+ *
+ * const imageDescriptionAiMsg = await llm.invoke([message]);
+ * console.log(imageDescriptionAiMsg.content);
+ * ```
+ *
+ * ```txt
+ * The weather in the image appears to be clear and sunny. The sky is mostly blue with a few scattered white clouds, indicating fair weather. The bright sunlight is casting shadows on the green, grassy hill, suggesting it is a pleasant day with good visibility. There are no signs of rain or stormy conditions.
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Usage Metadata</strong></summary>
+ *
+ * ```typescript
+ * const aiMsgForMetadata = await llm.invoke(input);
+ * console.log(aiMsgForMetadata.usage_metadata);
+ * ```
+ *
+ * ```txt
+ * { input_tokens: 10, output_tokens: 149, total_tokens: 159 }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Response Metadata</strong></summary>
+ *
+ * ```typescript
+ * const aiMsgForResponseMetadata = await llm.invoke(input);
+ * console.log(aiMsgForResponseMetadata.response_metadata);
+ * ```
+ *
+ * ```txt
+ * {
+ *   finishReason: 'STOP',
+ *   index: 0,
+ *   safetyRatings: [
+ *     {
+ *       category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT',
+ *       probability: 'NEGLIGIBLE'
+ *     },
+ *     {
+ *       category: 'HARM_CATEGORY_HATE_SPEECH',
+ *       probability: 'NEGLIGIBLE'
+ *     },
+ *     { category: 'HARM_CATEGORY_HARASSMENT', probability: 'NEGLIGIBLE' },
+ *     {
+ *       category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
+ *       probability: 'NEGLIGIBLE'
+ *     }
+ *   ]
+ * }
  * ```
+ * </details>
+ *
+ * <br />
  */
 class ChatGoogleGenerativeAI extends chat_models_1.BaseChatModel {
     static lc_name() {
@@ -191,6 +497,7 @@ class ChatGoogleGenerativeAI extends chat_models_1.BaseChatModel {
                 temperature: this.temperature,
                 topP: this.topP,
                 topK: this.topK,
+                ...(fields?.json ? { responseMimeType: "application/json" } : {}),
             },
         }, {
             apiVersion: fields?.apiVersion,
@@ -328,7 +635,7 @@ class ChatGoogleGenerativeAI extends chat_models_1.BaseChatModel {
     async completionWithRetry(request, options) {
         return this.caller.callWithOptions({ signal: options?.signal }, async () => {
             try {
-                return this.client.generateContent(request);
+                return await this.client.generateContent(request);
                 // eslint-disable-next-line @typescript-eslint/no-explicit-any
             }
             catch (e) {

package/dist/chat_models.d.ts

@@ -25,6 +25,8 @@ export interface GoogleGenerativeAIChatCallOptions extends BaseChatModelCallOpti
  */
 export interface GoogleGenerativeAIChatInput extends BaseChatModelParams, Pick<GoogleGenerativeAIChatCallOptions, "streamUsage"> {
     /**
+     * @deprecated Use "model" instead.
+     *
      * Model Name to use
      *
      * Alias for `model`
@@ -107,35 +109,347 @@ export interface GoogleGenerativeAIChatInput extends BaseChatModelParams, Pick<G
     baseUrl?: string;
     /** Whether to stream the results or not */
     streaming?: boolean;
+    /**
+     * Whether or not to force the model to respond with JSON.
+     * Available for `gemini-1.5` models and later.
+     * @default false
+     */
+    json?: boolean;
 }
 /**
- * A class that wraps the Google Palm chat model.
- * @example
+ * Google Generative AI chat model integration.
+ *
+ * Setup:
+ * Install `@langchain/google-genai` and set an environment variable named `GOOGLE_API_KEY`.
+ *
+ * ```bash
+ * npm install @langchain/google-genai
+ * export GOOGLE_API_KEY="your-api-key"
+ * ```
+ *
+ * ## [Constructor args](https://api.js.langchain.com/classes/langchain_google_genai.ChatGoogleGenerativeAI.html#constructor)
+ *
+ * ## [Runtime args](https://api.js.langchain.com/interfaces/langchain_google_genai.GoogleGenerativeAIChatCallOptions.html)
+ *
+ * Runtime args can be passed as the second argument to any of the base runnable methods `.invoke`. `.stream`, `.batch`, etc.
+ * They can also be passed via `.bind`, or the second arg in `.bindTools`, like shown in the examples below:
+ *
  * ```typescript
- * const model = new ChatGoogleGenerativeAI({
- *   apiKey: "<YOUR API KEY>",
- *   temperature: 0.7,
- *   modelName: "gemini-pro",
- *   topK: 40,
- *   topP: 1,
+ * // When calling `.bind`, call options should be passed via the first argument
+ * const llmWithArgsBound = llm.bind({
+ *   stop: ["\n"],
+ *   tools: [...],
  * });
- * const questions = [
- *   new HumanMessage({
- *     content: [
+ *
+ * // When calling `.bindTools`, call options should be passed via the second argument
+ * const llmWithTools = llm.bindTools(
+ *   [...],
+ *   {
+ *     stop: ["\n"],
+ *   }
+ * );
+ * ```
+ *
+ * ## Examples
+ *
+ * <details open>
+ * <summary><strong>Instantiate</strong></summary>
+ *
+ * ```typescript
+ * import { ChatGoogleGenerativeAI } from '@langchain/google-genai';
+ *
+ * const llm = new ChatGoogleGenerativeAI({
+ *   model: "gemini-1.5-flash",
+ *   temperature: 0,
+ *   maxRetries: 2,
+ *   // apiKey: "...",
+ *   // other params...
+ * });
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Invoking</strong></summary>
+ *
+ * ```typescript
+ * const input = `Translate "I love programming" into French.`;
+ *
+ * // Models also accept a list of chat messages or a formatted prompt
+ * const result = await llm.invoke(input);
+ * console.log(result);
+ * ```
+ *
+ * ```txt
+ * AIMessage {
+ *   "content": "There are a few ways to translate \"I love programming\" into French, depending on the level of formality and nuance you want to convey:\n\n**Formal:**\n\n* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and your intended audience. \n",
+ *   "response_metadata": {
+ *     "finishReason": "STOP",
+ *     "index": 0,
+ *     "safetyRatings": [
+ *       {
+ *         "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+ *         "probability": "NEGLIGIBLE"
+ *       },
  *       {
- *         type: "text",
- *         text: "You are a funny assistant that answers in pirate language.",
+ *         "category": "HARM_CATEGORY_HATE_SPEECH",
+ *         "probability": "NEGLIGIBLE"
  *       },
  *       {
- *         type: "text",
- *         text: "What is your favorite food?",
+ *         "category": "HARM_CATEGORY_HARASSMENT",
+ *         "probability": "NEGLIGIBLE"
  *       },
+ *       {
+ *         "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+ *         "probability": "NEGLIGIBLE"
+ *       }
  *     ]
- *   })
- * ];
- * const res = await model.invoke(questions);
- * console.log({ res });
+ *   },
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 149,
+ *     "total_tokens": 159
+ *   }
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Streaming Chunks</strong></summary>
+ *
+ * ```typescript
+ * for await (const chunk of await llm.stream(input)) {
+ *   console.log(chunk);
+ * }
+ * ```
+ *
+ * ```txt
+ * AIMessageChunk {
+ *   "content": "There",
+ *   "response_metadata": {
+ *     "index": 0
+ *   }
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 1,
+ *     "total_tokens": 11
+ *   }
+ * }
+ * AIMessageChunk {
+ *   "content": " are a few ways to translate \"I love programming\" into French, depending on",
+ * }
+ * AIMessageChunk {
+ *   "content": " the level of formality and nuance you want to convey:\n\n**Formal:**\n\n",
+ * }
+ * AIMessageChunk {
+ *   "content": "* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This",
+ * }
+ * AIMessageChunk {
+ *   "content": " is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More",
+ * }
+ * AIMessageChunk {
+ *   "content": " specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and",
+ * }
+ * AIMessageChunk {
+ *   "content": " your intended audience. \n",
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Aggregate Streamed Chunks</strong></summary>
+ *
+ * ```typescript
+ * import { AIMessageChunk } from '@langchain/core/messages';
+ * import { concat } from '@langchain/core/utils/stream';
+ *
+ * const stream = await llm.stream(input);
+ * let full: AIMessageChunk | undefined;
+ * for await (const chunk of stream) {
+ *   full = !full ? chunk : concat(full, chunk);
+ * }
+ * console.log(full);
+ * ```
+ *
+ * ```txt
+ * AIMessageChunk {
+ *   "content": "There are a few ways to translate \"I love programming\" into French, depending on the level of formality and nuance you want to convey:\n\n**Formal:**\n\n* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and your intended audience. \n",
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 277,
+ *     "total_tokens": 287
+ *   }
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Bind tools</strong></summary>
+ *
+ * ```typescript
+ * import { z } from 'zod';
+ *
+ * const GetWeather = {
+ *   name: "GetWeather",
+ *   description: "Get the current weather in a given location",
+ *   schema: z.object({
+ *     location: z.string().describe("The city and state, e.g. San Francisco, CA")
+ *   }),
+ * }
+ *
+ * const GetPopulation = {
+ *   name: "GetPopulation",
+ *   description: "Get the current population in a given location",
+ *   schema: z.object({
+ *     location: z.string().describe("The city and state, e.g. San Francisco, CA")
+ *   }),
+ * }
+ *
+ * const llmWithTools = llm.bindTools([GetWeather, GetPopulation]);
+ * const aiMsg = await llmWithTools.invoke(
+ *   "Which city is hotter today and which is bigger: LA or NY?"
+ * );
+ * console.log(aiMsg.tool_calls);
+ * ```
+ *
+ * ```txt
+ * [
+ *   {
+ *     name: 'GetWeather',
+ *     args: { location: 'Los Angeles, CA' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetWeather',
+ *     args: { location: 'New York, NY' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetPopulation',
+ *     args: { location: 'Los Angeles, CA' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetPopulation',
+ *     args: { location: 'New York, NY' },
+ *     type: 'tool_call'
+ *   }
+ * ]
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Structured Output</strong></summary>
+ *
+ * ```typescript
+ * const Joke = z.object({
+ *   setup: z.string().describe("The setup of the joke"),
+ *   punchline: z.string().describe("The punchline to the joke"),
+ *   rating: z.number().optional().describe("How funny the joke is, from 1 to 10")
+ * }).describe('Joke to tell user.');
+ *
+ * const structuredLlm = llm.withStructuredOutput(Joke, { name: "Joke" });
+ * const jokeResult = await structuredLlm.invoke("Tell me a joke about cats");
+ * console.log(jokeResult);
+ * ```
+ *
+ * ```txt
+ * {
+ *   setup: "Why don\\'t cats play poker?",
+ *   punchline: "Why don\\'t cats play poker? Because they always have an ace up their sleeve!"
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Multimodal</strong></summary>
+ *
+ * ```typescript
+ * import { HumanMessage } from '@langchain/core/messages';
+ *
+ * const imageUrl = "https://example.com/image.jpg";
+ * const imageData = await fetch(imageUrl).then(res => res.arrayBuffer());
+ * const base64Image = Buffer.from(imageData).toString('base64');
+ *
+ * const message = new HumanMessage({
+ *   content: [
+ *     { type: "text", text: "describe the weather in this image" },
+ *     {
+ *       type: "image_url",
+ *       image_url: { url: `data:image/jpeg;base64,${base64Image}` },
+ *     },
+ *   ]
+ * });
+ *
+ * const imageDescriptionAiMsg = await llm.invoke([message]);
+ * console.log(imageDescriptionAiMsg.content);
+ * ```
+ *
+ * ```txt
+ * The weather in the image appears to be clear and sunny. The sky is mostly blue with a few scattered white clouds, indicating fair weather. The bright sunlight is casting shadows on the green, grassy hill, suggesting it is a pleasant day with good visibility. There are no signs of rain or stormy conditions.
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Usage Metadata</strong></summary>
+ *
+ * ```typescript
+ * const aiMsgForMetadata = await llm.invoke(input);
+ * console.log(aiMsgForMetadata.usage_metadata);
+ * ```
+ *
+ * ```txt
+ * { input_tokens: 10, output_tokens: 149, total_tokens: 159 }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Response Metadata</strong></summary>
+ *
+ * ```typescript
+ * const aiMsgForResponseMetadata = await llm.invoke(input);
+ * console.log(aiMsgForResponseMetadata.response_metadata);
+ * ```
+ *
+ * ```txt
+ * {
+ *   finishReason: 'STOP',
+ *   index: 0,
+ *   safetyRatings: [
+ *     {
+ *       category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT',
+ *       probability: 'NEGLIGIBLE'
+ *     },
+ *     {
+ *       category: 'HARM_CATEGORY_HATE_SPEECH',
+ *       probability: 'NEGLIGIBLE'
+ *     },
+ *     { category: 'HARM_CATEGORY_HARASSMENT', probability: 'NEGLIGIBLE' },
+ *     {
+ *       category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
+ *       probability: 'NEGLIGIBLE'
+ *     }
+ *   ]
+ * }
  * ```
+ * </details>
+ *
+ * <br />
  */
 export declare class ChatGoogleGenerativeAI extends BaseChatModel<GoogleGenerativeAIChatCallOptions, AIMessageChunk> implements GoogleGenerativeAIChatInput {
     static lc_name(): string;

package/dist/chat_models.js

@@ -7,33 +7,339 @@ import { zodToGenerativeAIParameters } from "./utils/zod_to_genai_parameters.js"
 import { convertBaseMessagesToContent, convertResponseContentToChatGenerationChunk, convertToGenerativeAITools, mapGenerateContentResultToChatResult, } from "./utils/common.js";
 import { GoogleGenerativeAIToolsOutputParser } from "./output_parsers.js";
 /**
- * A class that wraps the Google Palm chat model.
- * @example
+ * Google Generative AI chat model integration.
+ *
+ * Setup:
+ * Install `@langchain/google-genai` and set an environment variable named `GOOGLE_API_KEY`.
+ *
+ * ```bash
+ * npm install @langchain/google-genai
+ * export GOOGLE_API_KEY="your-api-key"
+ * ```
+ *
+ * ## [Constructor args](https://api.js.langchain.com/classes/langchain_google_genai.ChatGoogleGenerativeAI.html#constructor)
+ *
+ * ## [Runtime args](https://api.js.langchain.com/interfaces/langchain_google_genai.GoogleGenerativeAIChatCallOptions.html)
+ *
+ * Runtime args can be passed as the second argument to any of the base runnable methods `.invoke`. `.stream`, `.batch`, etc.
+ * They can also be passed via `.bind`, or the second arg in `.bindTools`, like shown in the examples below:
+ *
+ * ```typescript
+ * // When calling `.bind`, call options should be passed via the first argument
+ * const llmWithArgsBound = llm.bind({
+ *   stop: ["\n"],
+ *   tools: [...],
+ * });
+ *
+ * // When calling `.bindTools`, call options should be passed via the second argument
+ * const llmWithTools = llm.bindTools(
+ *   [...],
+ *   {
+ *     stop: ["\n"],
+ *   }
+ * );
+ * ```
+ *
+ * ## Examples
+ *
+ * <details open>
+ * <summary><strong>Instantiate</strong></summary>
+ *
  * ```typescript
- * const model = new ChatGoogleGenerativeAI({
- *   apiKey: "<YOUR API KEY>",
- *   temperature: 0.7,
- *   modelName: "gemini-pro",
- *   topK: 40,
- *   topP: 1,
+ * import { ChatGoogleGenerativeAI } from '@langchain/google-genai';
+ *
+ * const llm = new ChatGoogleGenerativeAI({
+ *   model: "gemini-1.5-flash",
+ *   temperature: 0,
+ *   maxRetries: 2,
+ *   // apiKey: "...",
+ *   // other params...
  * });
- * const questions = [
- *   new HumanMessage({
- *     content: [
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Invoking</strong></summary>
+ *
+ * ```typescript
+ * const input = `Translate "I love programming" into French.`;
+ *
+ * // Models also accept a list of chat messages or a formatted prompt
+ * const result = await llm.invoke(input);
+ * console.log(result);
+ * ```
+ *
+ * ```txt
+ * AIMessage {
+ *   "content": "There are a few ways to translate \"I love programming\" into French, depending on the level of formality and nuance you want to convey:\n\n**Formal:**\n\n* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and your intended audience. \n",
+ *   "response_metadata": {
+ *     "finishReason": "STOP",
+ *     "index": 0,
+ *     "safetyRatings": [
+ *       {
+ *         "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+ *         "probability": "NEGLIGIBLE"
+ *       },
  *       {
- *         type: "text",
- *         text: "You are a funny assistant that answers in pirate language.",
+ *         "category": "HARM_CATEGORY_HATE_SPEECH",
+ *         "probability": "NEGLIGIBLE"
  *       },
  *       {
- *         type: "text",
- *         text: "What is your favorite food?",
+ *         "category": "HARM_CATEGORY_HARASSMENT",
+ *         "probability": "NEGLIGIBLE"
  *       },
+ *       {
+ *         "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+ *         "probability": "NEGLIGIBLE"
+ *       }
  *     ]
- *   })
- * ];
- * const res = await model.invoke(questions);
- * console.log({ res });
+ *   },
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 149,
+ *     "total_tokens": 159
+ *   }
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Streaming Chunks</strong></summary>
+ *
+ * ```typescript
+ * for await (const chunk of await llm.stream(input)) {
+ *   console.log(chunk);
+ * }
+ * ```
+ *
+ * ```txt
+ * AIMessageChunk {
+ *   "content": "There",
+ *   "response_metadata": {
+ *     "index": 0
+ *   }
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 1,
+ *     "total_tokens": 11
+ *   }
+ * }
+ * AIMessageChunk {
+ *   "content": " are a few ways to translate \"I love programming\" into French, depending on",
+ * }
+ * AIMessageChunk {
+ *   "content": " the level of formality and nuance you want to convey:\n\n**Formal:**\n\n",
+ * }
+ * AIMessageChunk {
+ *   "content": "* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This",
+ * }
+ * AIMessageChunk {
+ *   "content": " is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More",
+ * }
+ * AIMessageChunk {
+ *   "content": " specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and",
+ * }
+ * AIMessageChunk {
+ *   "content": " your intended audience. \n",
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Aggregate Streamed Chunks</strong></summary>
+ *
+ * ```typescript
+ * import { AIMessageChunk } from '@langchain/core/messages';
+ * import { concat } from '@langchain/core/utils/stream';
+ *
+ * const stream = await llm.stream(input);
+ * let full: AIMessageChunk | undefined;
+ * for await (const chunk of stream) {
+ *   full = !full ? chunk : concat(full, chunk);
+ * }
+ * console.log(full);
+ * ```
+ *
+ * ```txt
+ * AIMessageChunk {
+ *   "content": "There are a few ways to translate \"I love programming\" into French, depending on the level of formality and nuance you want to convey:\n\n**Formal:**\n\n* **J'aime la programmation.** (This is the most literal and formal translation.)\n\n**Informal:**\n\n* **J'adore programmer.** (This is a more enthusiastic and informal translation.)\n* **J'aime beaucoup programmer.** (This is a slightly less enthusiastic but still informal translation.)\n\n**More specific:**\n\n* **J'aime beaucoup coder.** (This specifically refers to writing code.)\n* **J'aime beaucoup développer des logiciels.** (This specifically refers to developing software.)\n\nThe best translation will depend on the context and your intended audience. \n",
+ *   "usage_metadata": {
+ *     "input_tokens": 10,
+ *     "output_tokens": 277,
+ *     "total_tokens": 287
+ *   }
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Bind tools</strong></summary>
+ *
+ * ```typescript
+ * import { z } from 'zod';
+ *
+ * const GetWeather = {
+ *   name: "GetWeather",
+ *   description: "Get the current weather in a given location",
+ *   schema: z.object({
+ *     location: z.string().describe("The city and state, e.g. San Francisco, CA")
+ *   }),
+ * }
+ *
+ * const GetPopulation = {
+ *   name: "GetPopulation",
+ *   description: "Get the current population in a given location",
+ *   schema: z.object({
+ *     location: z.string().describe("The city and state, e.g. San Francisco, CA")
+ *   }),
+ * }
+ *
+ * const llmWithTools = llm.bindTools([GetWeather, GetPopulation]);
+ * const aiMsg = await llmWithTools.invoke(
+ *   "Which city is hotter today and which is bigger: LA or NY?"
+ * );
+ * console.log(aiMsg.tool_calls);
+ * ```
+ *
+ * ```txt
+ * [
+ *   {
+ *     name: 'GetWeather',
+ *     args: { location: 'Los Angeles, CA' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetWeather',
+ *     args: { location: 'New York, NY' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetPopulation',
+ *     args: { location: 'Los Angeles, CA' },
+ *     type: 'tool_call'
+ *   },
+ *   {
+ *     name: 'GetPopulation',
+ *     args: { location: 'New York, NY' },
+ *     type: 'tool_call'
+ *   }
+ * ]
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Structured Output</strong></summary>
+ *
+ * ```typescript
+ * const Joke = z.object({
+ *   setup: z.string().describe("The setup of the joke"),
+ *   punchline: z.string().describe("The punchline to the joke"),
+ *   rating: z.number().optional().describe("How funny the joke is, from 1 to 10")
+ * }).describe('Joke to tell user.');
+ *
+ * const structuredLlm = llm.withStructuredOutput(Joke, { name: "Joke" });
+ * const jokeResult = await structuredLlm.invoke("Tell me a joke about cats");
+ * console.log(jokeResult);
+ * ```
+ *
+ * ```txt
+ * {
+ *   setup: "Why don\\'t cats play poker?",
+ *   punchline: "Why don\\'t cats play poker? Because they always have an ace up their sleeve!"
+ * }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Multimodal</strong></summary>
+ *
+ * ```typescript
+ * import { HumanMessage } from '@langchain/core/messages';
+ *
+ * const imageUrl = "https://example.com/image.jpg";
+ * const imageData = await fetch(imageUrl).then(res => res.arrayBuffer());
+ * const base64Image = Buffer.from(imageData).toString('base64');
+ *
+ * const message = new HumanMessage({
+ *   content: [
+ *     { type: "text", text: "describe the weather in this image" },
+ *     {
+ *       type: "image_url",
+ *       image_url: { url: `data:image/jpeg;base64,${base64Image}` },
+ *     },
+ *   ]
+ * });
+ *
+ * const imageDescriptionAiMsg = await llm.invoke([message]);
+ * console.log(imageDescriptionAiMsg.content);
+ * ```
+ *
+ * ```txt
+ * The weather in the image appears to be clear and sunny. The sky is mostly blue with a few scattered white clouds, indicating fair weather. The bright sunlight is casting shadows on the green, grassy hill, suggesting it is a pleasant day with good visibility. There are no signs of rain or stormy conditions.
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Usage Metadata</strong></summary>
+ *
+ * ```typescript
+ * const aiMsgForMetadata = await llm.invoke(input);
+ * console.log(aiMsgForMetadata.usage_metadata);
+ * ```
+ *
+ * ```txt
+ * { input_tokens: 10, output_tokens: 149, total_tokens: 159 }
+ * ```
+ * </details>
+ *
+ * <br />
+ *
+ * <details>
+ * <summary><strong>Response Metadata</strong></summary>
+ *
+ * ```typescript
+ * const aiMsgForResponseMetadata = await llm.invoke(input);
+ * console.log(aiMsgForResponseMetadata.response_metadata);
+ * ```
+ *
+ * ```txt
+ * {
+ *   finishReason: 'STOP',
+ *   index: 0,
+ *   safetyRatings: [
+ *     {
+ *       category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT',
+ *       probability: 'NEGLIGIBLE'
+ *     },
+ *     {
+ *       category: 'HARM_CATEGORY_HATE_SPEECH',
+ *       probability: 'NEGLIGIBLE'
+ *     },
+ *     { category: 'HARM_CATEGORY_HARASSMENT', probability: 'NEGLIGIBLE' },
+ *     {
+ *       category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
+ *       probability: 'NEGLIGIBLE'
+ *     }
+ *   ]
+ * }
  * ```
+ * </details>
+ *
+ * <br />
  */
 export class ChatGoogleGenerativeAI extends BaseChatModel {
     static lc_name() {
@@ -188,6 +494,7 @@ export class ChatGoogleGenerativeAI extends BaseChatModel {
                 temperature: this.temperature,
                 topP: this.topP,
                 topK: this.topK,
+                ...(fields?.json ? { responseMimeType: "application/json" } : {}),
             },
         }, {
             apiVersion: fields?.apiVersion,
@@ -325,7 +632,7 @@ export class ChatGoogleGenerativeAI extends BaseChatModel {
     async completionWithRetry(request, options) {
         return this.caller.callWithOptions({ signal: options?.signal }, async () => {
             try {
-                return this.client.generateContent(request);
+                return await this.client.generateContent(request);
                 // eslint-disable-next-line @typescript-eslint/no-explicit-any
             }
             catch (e) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@langchain/google-genai",
-  "version": "0.0.25",
+  "version": "0.1.0",
   "description": "Google Generative AI integration for LangChain.js",
   "type": "module",
   "engines": {
@@ -15,7 +15,7 @@
   "homepage": "https://github.com/langchain-ai/langchainjs/tree/main/libs/langchain-google-genai/",
   "scripts": {
     "build": "yarn turbo:command build:internal --filter=@langchain/google-genai",
-    "build:internal": "yarn lc_build_v2 --create-entrypoints --pre --tree-shaking",
+    "build:internal": "yarn lc_build --create-entrypoints --pre --tree-shaking",
     "lint:eslint": "NODE_OPTIONS=--max-old-space-size=4096 eslint --cache --ext .ts,.js src/",
     "lint:dpdm": "dpdm --exit-code circular:1 --no-warning --no-tree src/*.ts src/**/*.ts",
     "lint": "yarn lint:eslint && yarn lint:dpdm",
@@ -36,12 +36,15 @@
   "license": "MIT",
   "dependencies": {
     "@google/generative-ai": "^0.7.0",
-    "@langchain/core": ">=0.2.21 <0.3.0",
     "zod-to-json-schema": "^3.22.4"
   },
+  "peerDependencies": {
+    "@langchain/core": ">=0.2.21 <0.4.0"
+  },
   "devDependencies": {
     "@jest/globals": "^29.5.0",
-    "@langchain/scripts": "~0.0.20",
+    "@langchain/core": "workspace:*",
+    "@langchain/scripts": ">=0.1.0 <0.2.0",
     "@langchain/standard-tests": "0.0.0",
     "@swc/core": "^1.3.90",
     "@swc/jest": "^0.2.29",