agsamantha/node_modules/@langchain/openai/dist/azure/chat_models.js
2024-10-02 15:15:21 -05:00

550 lines
16 KiB
JavaScript

import { AzureOpenAI as AzureOpenAIClient } from "openai";
import { ChatOpenAI } from "../chat_models.js";
import { getEndpoint } from "../utils/azure.js";
/**
* Azure OpenAI chat model integration.
*
* Setup:
* Install `@langchain/openai` and set the following environment variables:
*
* ```bash
* npm install @langchain/openai
* export AZURE_OPENAI_API_KEY="your-api-key"
* export AZURE_OPENAI_API_DEPLOYMENT_NAME="your-deployment-name"
* export AZURE_OPENAI_API_VERSION="your-version"
* export AZURE_OPENAI_BASE_PATH="your-base-path"
* ```
*
* ## [Constructor args](https://api.js.langchain.com/classes/langchain_openai.AzureChatOpenAI.html#constructor)
*
* ## [Runtime args](https://api.js.langchain.com/interfaces/langchain_openai.ChatOpenAICallOptions.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(
* [...],
* {
* tool_choice: "auto",
* }
* );
* ```
*
* ## Examples
*
* <details open>
* <summary><strong>Instantiate</strong></summary>
*
* ```typescript
* import { AzureChatOpenAI } from '@langchain/openai';
*
* const llm = new AzureChatOpenAI({
* azureOpenAIApiKey: process.env.AZURE_OPENAI_API_KEY, // In Node.js defaults to process.env.AZURE_OPENAI_API_KEY
* azureOpenAIApiInstanceName: process.env.AZURE_OPENAI_API_INSTANCE_NAME, // In Node.js defaults to process.env.AZURE_OPENAI_API_INSTANCE_NAME
* azureOpenAIApiDeploymentName: process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME, // In Node.js defaults to process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME
* azureOpenAIApiVersion: process.env.AZURE_OPENAI_API_VERSION, // In Node.js defaults to process.env.AZURE_OPENAI_API_VERSION
* temperature: 0,
* maxTokens: undefined,
* timeout: undefined,
* maxRetries: 2,
* // apiKey: "...",
* // baseUrl: "...",
* // 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 {
* "id": "chatcmpl-9u4Mpu44CbPjwYFkTbeoZgvzB00Tz",
* "content": "J'adore la programmation.",
* "response_metadata": {
* "tokenUsage": {
* "completionTokens": 5,
* "promptTokens": 28,
* "totalTokens": 33
* },
* "finish_reason": "stop",
* "system_fingerprint": "fp_3aa7262c27"
* },
* "usage_metadata": {
* "input_tokens": 28,
* "output_tokens": 5,
* "total_tokens": 33
* }
* }
* ```
* </details>
*
* <br />
*
* <details>
* <summary><strong>Streaming Chunks</strong></summary>
*
* ```typescript
* for await (const chunk of await llm.stream(input)) {
* console.log(chunk);
* }
* ```
*
* ```txt
* AIMessageChunk {
* "id": "chatcmpl-9u4NWB7yUeHCKdLr6jP3HpaOYHTqs",
* "content": ""
* }
* AIMessageChunk {
* "content": "J"
* }
* AIMessageChunk {
* "content": "'adore"
* }
* AIMessageChunk {
* "content": " la"
* }
* AIMessageChunk {
* "content": " programmation",,
* }
* AIMessageChunk {
* "content": ".",,
* }
* AIMessageChunk {
* "content": "",
* "response_metadata": {
* "finish_reason": "stop",
* "system_fingerprint": "fp_c9aa9c0491"
* },
* }
* AIMessageChunk {
* "content": "",
* "usage_metadata": {
* "input_tokens": 28,
* "output_tokens": 5,
* "total_tokens": 33
* }
* }
* ```
* </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 {
* "id": "chatcmpl-9u4PnX6Fy7OmK46DASy0bH6cxn5Xu",
* "content": "J'adore la programmation.",
* "response_metadata": {
* "prompt": 0,
* "completion": 0,
* "finish_reason": "stop",
* },
* "usage_metadata": {
* "input_tokens": 28,
* "output_tokens": 5,
* "total_tokens": 33
* }
* }
* ```
* </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',
* id: 'call_uPU4FiFzoKAtMxfmPnfQL6UK'
* },
* {
* name: 'GetWeather',
* args: { location: 'New York, NY' },
* type: 'tool_call',
* id: 'call_UNkEwuQsHrGYqgDQuH9nPAtX'
* },
* {
* name: 'GetPopulation',
* args: { location: 'Los Angeles, CA' },
* type: 'tool_call',
* id: 'call_kL3OXxaq9OjIKqRTpvjaCH14'
* },
* {
* name: 'GetPopulation',
* args: { location: 'New York, NY' },
* type: 'tool_call',
* id: 'call_s9KQB1UWj45LLGaEnjz0179q'
* }
* ]
* ```
* </details>
*
* <br />
*
* <details>
* <summary><strong>Structured Output</strong></summary>
*
* ```typescript
* import { z } from 'zod';
*
* 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 was the cat sitting on the computer?',
* punchline: 'Because it wanted to keep an eye on the mouse!',
* rating: 7
* }
* ```
* </details>
*
* <br />
*
* <details>
* <summary><strong>JSON Object Response Format</strong></summary>
*
* ```typescript
* const jsonLlm = llm.bind({ response_format: { type: "json_object" } });
* const jsonLlmAiMsg = await jsonLlm.invoke(
* "Return a JSON object with key 'randomInts' and a value of 10 random ints in [0-99]"
* );
* console.log(jsonLlmAiMsg.content);
* ```
*
* ```txt
* {
* "randomInts": [23, 87, 45, 12, 78, 34, 56, 90, 11, 67]
* }
* ```
* </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: 28, output_tokens: 5, total_tokens: 33 }
* ```
* </details>
*
* <br />
*
* <details>
* <summary><strong>Logprobs</strong></summary>
*
* ```typescript
* const logprobsLlm = new ChatOpenAI({ logprobs: true });
* const aiMsgForLogprobs = await logprobsLlm.invoke(input);
* console.log(aiMsgForLogprobs.response_metadata.logprobs);
* ```
*
* ```txt
* {
* content: [
* {
* token: 'J',
* logprob: -0.000050616763,
* bytes: [Array],
* top_logprobs: []
* },
* {
* token: "'",
* logprob: -0.01868736,
* bytes: [Array],
* top_logprobs: []
* },
* {
* token: 'ad',
* logprob: -0.0000030545007,
* bytes: [Array],
* top_logprobs: []
* },
* { token: 'ore', logprob: 0, bytes: [Array], top_logprobs: [] },
* {
* token: ' la',
* logprob: -0.515404,
* bytes: [Array],
* top_logprobs: []
* },
* {
* token: ' programm',
* logprob: -0.0000118755715,
* bytes: [Array],
* top_logprobs: []
* },
* { token: 'ation', logprob: 0, bytes: [Array], top_logprobs: [] },
* {
* token: '.',
* logprob: -0.0000037697225,
* bytes: [Array],
* top_logprobs: []
* }
* ],
* refusal: null
* }
* ```
* </details>
*
* <br />
*
* <details>
* <summary><strong>Response Metadata</strong></summary>
*
* ```typescript
* const aiMsgForResponseMetadata = await llm.invoke(input);
* console.log(aiMsgForResponseMetadata.response_metadata);
* ```
*
* ```txt
* {
* tokenUsage: { completionTokens: 5, promptTokens: 28, totalTokens: 33 },
* finish_reason: 'stop',
* system_fingerprint: 'fp_3aa7262c27'
* }
* ```
* </details>
*/
export class AzureChatOpenAI extends ChatOpenAI {
_llmType() {
return "azure_openai";
}
get lc_aliases() {
return {
openAIApiKey: "openai_api_key",
openAIApiVersion: "openai_api_version",
openAIBasePath: "openai_api_base",
deploymentName: "deployment_name",
azureOpenAIEndpoint: "azure_endpoint",
azureOpenAIApiVersion: "openai_api_version",
azureOpenAIBasePath: "openai_api_base",
azureOpenAIApiDeploymentName: "deployment_name",
};
}
constructor(fields) {
const newFields = fields ? { ...fields } : fields;
if (newFields) {
// don't rewrite the fields if they are already set
newFields.azureOpenAIApiDeploymentName =
newFields.azureOpenAIApiDeploymentName ?? newFields.deploymentName;
newFields.azureOpenAIApiKey =
newFields.azureOpenAIApiKey ?? newFields.openAIApiKey;
newFields.azureOpenAIApiVersion =
newFields.azureOpenAIApiVersion ?? newFields.openAIApiVersion;
}
super(newFields);
}
getLsParams(options) {
const params = super.getLsParams(options);
params.ls_provider = "azure";
return params;
}
_getClientOptions(options) {
if (!this.client) {
const openAIEndpointConfig = {
azureOpenAIApiDeploymentName: this.azureOpenAIApiDeploymentName,
azureOpenAIApiInstanceName: this.azureOpenAIApiInstanceName,
azureOpenAIApiKey: this.azureOpenAIApiKey,
azureOpenAIBasePath: this.azureOpenAIBasePath,
azureADTokenProvider: this.azureADTokenProvider,
baseURL: this.clientConfig.baseURL,
azureOpenAIEndpoint: this.azureOpenAIEndpoint,
};
const endpoint = getEndpoint(openAIEndpointConfig);
const params = {
...this.clientConfig,
baseURL: endpoint,
timeout: this.timeout,
maxRetries: 0,
};
if (!this.azureADTokenProvider) {
params.apiKey = openAIEndpointConfig.azureOpenAIApiKey;
}
if (!params.baseURL) {
delete params.baseURL;
}
params.defaultHeaders = {
...params.defaultHeaders,
"User-Agent": params.defaultHeaders?.["User-Agent"]
? `${params.defaultHeaders["User-Agent"]}: langchainjs-azure-openai-v2`
: `langchainjs-azure-openai-v2`,
};
this.client = new AzureOpenAIClient({
apiVersion: this.azureOpenAIApiVersion,
azureADTokenProvider: this.azureADTokenProvider,
deployment: this.azureOpenAIApiDeploymentName,
...params,
});
}
const requestOptions = {
...this.clientConfig,
...options,
};
if (this.azureOpenAIApiKey) {
requestOptions.headers = {
"api-key": this.azureOpenAIApiKey,
...requestOptions.headers,
};
requestOptions.query = {
"api-version": this.azureOpenAIApiVersion,
...requestOptions.query,
};
}
return requestOptions;
}
// eslint-disable-next-line @typescript-eslint/no-explicit-any
toJSON() {
const json = super.toJSON();
function isRecord(obj) {
return typeof obj === "object" && obj != null;
}
if (isRecord(json) && isRecord(json.kwargs)) {
delete json.kwargs.azure_openai_base_path;
delete json.kwargs.azure_openai_api_deployment_name;
delete json.kwargs.azure_openai_api_key;
delete json.kwargs.azure_openai_api_version;
delete json.kwargs.azure_open_ai_base_path;
if (!json.kwargs.azure_endpoint && this.azureOpenAIEndpoint) {
json.kwargs.azure_endpoint = this.azureOpenAIEndpoint;
}
if (!json.kwargs.azure_endpoint && this.azureOpenAIBasePath) {
const parts = this.azureOpenAIBasePath.split("/openai/deployments/");
if (parts.length === 2 && parts[0].startsWith("http")) {
const [endpoint] = parts;
json.kwargs.azure_endpoint = endpoint;
}
}
if (!json.kwargs.azure_endpoint && this.azureOpenAIApiInstanceName) {
json.kwargs.azure_endpoint = `https://${this.azureOpenAIApiInstanceName}.openai.azure.com/`;
}
if (!json.kwargs.deployment_name && this.azureOpenAIApiDeploymentName) {
json.kwargs.deployment_name = this.azureOpenAIApiDeploymentName;
}
if (!json.kwargs.deployment_name && this.azureOpenAIBasePath) {
const parts = this.azureOpenAIBasePath.split("/openai/deployments/");
if (parts.length === 2) {
const [, deployment] = parts;
json.kwargs.deployment_name = deployment;
}
}
if (json.kwargs.azure_endpoint &&
json.kwargs.deployment_name &&
json.kwargs.openai_api_base) {
delete json.kwargs.openai_api_base;
}
if (json.kwargs.azure_openai_api_instance_name &&
json.kwargs.azure_endpoint) {
delete json.kwargs.azure_openai_api_instance_name;
}
}
return json;
}
}