Skip to content

Commit

Permalink
azure-openai[minor]: update docs and fix openai api usage (#4898)
Browse files Browse the repository at this point in the history 
* feat: add Azure OpenAI examples

docs: add Azure OpenAI SDK documentation

docs: update docs and example for new package name

docs: update readme

docs: add Azure LLM integrations

docs: update doc

docs: add managed identity usage

fix: add missing model name in example

docs: add Azure OpenAI embeddings docs

* docs: fix embedding example

* chore: cleanup old changes

* docs: fix docs and add missing sections

* docs: fix build

* fix: openai key usage with LLM

* docs: add missing model name

* fix: openai key usage with embeddings

* feat: disable stripNewLines option by default

BREAKING CHANGE: stripNewLines option is now disabled by default

* chore: missing only in test

* fix: openai key usage with chat

* chore: formatting

* chore: formatting

* Update index.mdx

---------

Co-authored-by: Jacob Lee <[email protected]>
  • Loading branch information
@sinedied @jacoblee93
sinedied and jacoblee93 authored Mar 30, 2024
1 parent 14632b7 commit c694a5d
Show file tree
Hide file tree
Showing 10 changed files with 305 additions and 184 deletions.
122 changes: 71 additions & 51 deletions docs/core_docs/docs/integrations/llms/azure.mdx
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,56 +6,6 @@ LangChain.js supports integration with [Azure OpenAI](https://azure.microsoft.co

You can learn more about Azure OpenAI and its difference with the OpenAI API on [this page](https://learn.microsoft.com/azure/ai-services/openai/overview). If you don't have an Azure account, you can [create a free account](https://azure.microsoft.com/free/) to get started.

## Using the OpenAI SDK

You can also use the `OpenAI` class to call OpenAI models hosted on Azure.

For example, if your Azure instance is hosted under `https://{MY_INSTANCE_NAME}.openai.azure.com/openai/deployments/{DEPLOYMENT_NAME}`, you could initialize your instance like this:

import IntegrationInstallTooltip from "@mdx_components/integration_install_tooltip.mdx";

<IntegrationInstallTooltip></IntegrationInstallTooltip>

```bash npm2yarn
npm install @langchain/openai
```

```typescript
import { OpenAI } from "@langchain/openai";

const model = new OpenAI({
temperature: 0.9,
azureOpenAIApiKey: "YOUR-API-KEY",
azureOpenAIApiVersion: "YOUR-API-VERSION",
azureOpenAIApiInstanceName: "{MY_INSTANCE_NAME}",
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}",
});
const res = await model.invoke(
"What would be a good company name for a company that makes colorful socks?"
);
console.log({ res });
```

If your instance is hosted under a domain other than the default `openai.azure.com`, you'll need to use the alternate `AZURE_OPENAI_BASE_PATH` environment variable.
For example, here's how you would connect to the domain `https://westeurope.api.microsoft.com/openai/deployments/{DEPLOYMENT_NAME}`:

```typescript
import { OpenAI } from "@langchain/openai";

const model = new OpenAI({
temperature: 0.9,
azureOpenAIApiKey: "YOUR-API-KEY",
azureOpenAIApiVersion: "YOUR-API-VERSION",
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}",
azureOpenAIBasePath:
"https://westeurope.api.microsoft.com/openai/deployments", // In Node.js defaults to process.env.AZURE_OPENAI_BASE_PATH
});
const res = await model.invoke(
"What would be a good company name for a company that makes colorful socks?"
);
console.log({ res });
```

## Using the Azure OpenAI SDK

You'll first need to install the [`@langchain/azure-openai`](https://www.npmjs.com/package/@langchain/azure-openai) package:
Expand Down Expand Up @@ -87,7 +37,21 @@ const model = new AzureOpenAI({
azureOpenAIEndpoint: "<your_endpoint>",
azureOpenAIApiKey: "<your_key>",
azureOpenAIApiDeploymentName: "<your_deployment_name",
modelName: "<your_model>",
});
```

If you're using Azure Managed Identity, you can also pass the credentials directly to the constructor:

```typescript
import { DefaultAzureCredential } from "@azure/identity";
import { AzureOpenAI } from "@langchain/azure-openai";

const credentials = new DefaultAzureCredential();

const model = new AzureOpenAI({
credentials,
azureOpenAIEndpoint: "<your_endpoint>",
azureOpenAIApiDeploymentName: "<your_deployment_name",
});
```

Expand All @@ -113,3 +77,59 @@ import CodeBlock from "@theme/CodeBlock";
import LLMExample from "@examples/llms/azure_openai.ts";

<CodeBlock language="text">{LLMExample}</CodeBlock>

### Chat usage example

import ChatExample from "@examples/llms/azure_openai-chat.ts";

<CodeBlock language="text">{ChatExample}</CodeBlock>

## Using OpenAI SDK

You can also use the `OpenAI` class to call OpenAI models hosted on Azure.

For example, if your Azure instance is hosted under `https://{MY_INSTANCE_NAME}.openai.azure.com/openai/deployments/{DEPLOYMENT_NAME}`, you could initialize your instance like this:

import IntegrationInstallTooltip from "@mdx_components/integration_install_tooltip.mdx";

<IntegrationInstallTooltip></IntegrationInstallTooltip>

```bash npm2yarn
npm install @langchain/openai
```

```typescript
import { OpenAI } from "@langchain/openai";

const model = new OpenAI({
temperature: 0.9,
azureOpenAIApiKey: "YOUR-API-KEY",
azureOpenAIApiVersion: "YOUR-API-VERSION",
azureOpenAIApiInstanceName: "{MY_INSTANCE_NAME}",
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}",
});
const res = await model.invoke(
"What would be a good company name for a company that makes colorful socks?"
);
console.log({ res });
```

If your instance is hosted under a domain other than the default `openai.azure.com`, you'll need to use the alternate `AZURE_OPENAI_BASE_PATH` environment variable.
For example, here's how you would connect to the domain `https://westeurope.api.microsoft.com/openai/deployments/{DEPLOYMENT_NAME}`:

```typescript
import { OpenAI } from "@langchain/openai";

const model = new OpenAI({
temperature: 0.9,
azureOpenAIApiKey: "YOUR-API-KEY",
azureOpenAIApiVersion: "YOUR-API-VERSION",
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}",
azureOpenAIBasePath:
"https://westeurope.api.microsoft.com/openai/deployments", // In Node.js defaults to process.env.AZURE_OPENAI_BASE_PATH
});
const res = await model.invoke(
"What would be a good company name for a company that makes colorful socks?"
);
console.log({ res });
```
94 changes: 46 additions & 48 deletions docs/core_docs/docs/integrations/text_embedding/azure_openai.mdx
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,57 +10,14 @@ LangChain.js supports integration with [Azure OpenAI](https://azure.microsoft.co

You can learn more about Azure OpenAI and its difference with the OpenAI API on [this page](https://learn.microsoft.com/azure/ai-services/openai/overview). If you don't have an Azure account, you can [create a free account](https://azure.microsoft.com/free/) to get started.

## Using the OpenAI SDK
## Using Azure OpenAI SDK

The `OpenAIEmbeddings` class can also use the OpenAI API on Azure to generate embeddings for a given text. By default it strips new line characters from the text, as recommended by OpenAI, but you can disable this by passing `stripNewLines: false` to the constructor.

For example, if your Azure instance is hosted under `https://{MY_INSTANCE_NAME}.openai.azure.com/openai/deployments/{DEPLOYMENT_NAME}`, you
could initialize your instance like this:
You'll first need to install the [`@langchain/azure-openai`](https://www.npmjs.com/package/@langchain/azure-openai) package:

import IntegrationInstallTooltip from "@mdx_components/integration_install_tooltip.mdx";

<IntegrationInstallTooltip></IntegrationInstallTooltip>

```bash npm2yarn
npm install @langchain/openai
```

```typescript
import { OpenAIEmbeddings } from "@langchain/openai";

const embeddings = new OpenAIEmbeddings({
azureOpenAIApiKey: "YOUR-API-KEY", // In Node.js defaults to process.env.AZURE_OPENAI_API_KEY
azureOpenAIApiVersion: "YOUR-API-VERSION", // In Node.js defaults to process.env.AZURE_OPENAI_API_VERSION
azureOpenAIApiInstanceName: "{MY_INSTANCE_NAME}", // In Node.js defaults to process.env.AZURE_OPENAI_API_INSTANCE_NAME
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}", // In Node.js defaults to process.env.AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME
});
```

If you'd like to initialize using environment variable defaults, the `process.env.AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME`
will be used first, then `process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME`. This can be useful if you're using these embeddings
with another Azure OpenAI model.

If your instance is hosted under a domain other than the default `openai.azure.com`, you'll need to use the alternate `AZURE_OPENAI_BASE_PATH` environment variable.
For example, here's how you would connect to the domain `https://westeurope.api.microsoft.com/openai/deployments/{DEPLOYMENT_NAME}`:

```typescript
import { OpenAIEmbeddings } from "@langchain/openai";

const embeddings = new OpenAIEmbeddings({
azureOpenAIApiKey: "YOUR-API-KEY",
azureOpenAIApiVersion: "YOUR-API-VERSION",
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}",
azureOpenAIBasePath:
"https://westeurope.api.microsoft.com/openai/deployments", // In Node.js defaults to process.env.AZURE_OPENAI_BASE_PATH
});
```

## Using the Azure OpenAI SDK

You'll first need to install the [`@langchain/azure-openai`](https://www.npmjs.com/package/@langchain/azure-openai) package:

<IntegrationInstallTooltip></IntegrationInstallTooltip>

```bash npm2yarn
npm install -S @langchain/azure-openai
```
Expand All @@ -86,7 +43,6 @@ const model = new AzureOpenAI({
azureOpenAIEndpoint: "<your_endpoint>",
azureOpenAIApiKey: "<your_key>",
azureOpenAIApiDeploymentName: "<your_embedding_deployment_name",
modelName: "<your_model>",
});
```

Expand All @@ -102,13 +58,55 @@ const model = new AzureOpenAI({
credentials,
azureOpenAIEndpoint: "<your_endpoint>",
azureOpenAIApiDeploymentName: "<your_embedding_deployment_name",
modelName: "<your_model>",
});
```

### Usage example

import CodeBlock from "@theme/CodeBlock";
import Example from "@examples/llms/azure_openai.ts";
import Example from "@examples/embeddings/azure_openai.ts";

<CodeBlock language="text">{Example}</CodeBlock>

## Using OpenAI SDK

The `OpenAIEmbeddings` class can also use the OpenAI API on Azure to generate embeddings for a given text. By default it strips new line characters from the text, as recommended by OpenAI, but you can disable this by passing `stripNewLines: false` to the constructor.

For example, if your Azure instance is hosted under `https://{MY_INSTANCE_NAME}.openai.azure.com/openai/deployments/{DEPLOYMENT_NAME}`, you
could initialize your instance like this:

<IntegrationInstallTooltip></IntegrationInstallTooltip>

```bash npm2yarn
npm install @langchain/openai
```

```typescript
import { OpenAIEmbeddings } from "@langchain/openai";

const embeddings = new OpenAIEmbeddings({
azureOpenAIApiKey: "YOUR-API-KEY", // In Node.js defaults to process.env.AZURE_OPENAI_API_KEY
azureOpenAIApiVersion: "YOUR-API-VERSION", // In Node.js defaults to process.env.AZURE_OPENAI_API_VERSION
azureOpenAIApiInstanceName: "{MY_INSTANCE_NAME}", // In Node.js defaults to process.env.AZURE_OPENAI_API_INSTANCE_NAME
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}", // In Node.js defaults to process.env.AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME
});
```

If you'd like to initialize using environment variable defaults, the `process.env.AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME`
will be used first, then `process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME`. This can be useful if you're using these embeddings
with another Azure OpenAI model.

If your instance is hosted under a domain other than the default `openai.azure.com`, you'll need to use the alternate `AZURE_OPENAI_BASE_PATH` environment variable.
For example, here's how you would connect to the domain `https://westeurope.api.microsoft.com/openai/deployments/{DEPLOYMENT_NAME}`:

```typescript
import { OpenAIEmbeddings } from "@langchain/openai";

const embeddings = new OpenAIEmbeddings({
azureOpenAIApiKey: "YOUR-API-KEY",
azureOpenAIApiVersion: "YOUR-API-VERSION",
azureOpenAIApiDeploymentName: "{DEPLOYMENT_NAME}",
azureOpenAIBasePath:
"https://westeurope.api.microsoft.com/openai/deployments", // In Node.js defaults to process.env.AZURE_OPENAI_BASE_PATH
});
```
18 changes: 18 additions & 0 deletions examples/src/llms/azure_openai-chat.ts
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
import { AzureChatOpenAI } from "@langchain/azure-openai";

export const run = async () => {
const model = new AzureChatOpenAI({
modelName: "gpt-4",
prefixMessages: [
{
role: "system",
content: "You are a helpful assistant that answers in pirate language",
},
],
maxTokens: 50,
});
const res = await model.invoke(
"What would be a good company name for a company that makes colorful socks?"
);
console.log({ res });
};
36 changes: 32 additions & 4 deletions libs/langchain-azure-openai/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ You can do so by adding appropriate fields to your project's `package.json` like
"name": "your-project",
"version": "0.0.0",
"dependencies": {
"@langchain/openai": "^0.0.9",
"@langchain/azure-openai": "^0.0.4",
"langchain": "0.0.207"
},
"resolutions": {
Expand Down Expand Up @@ -56,7 +56,6 @@ Then initialize the model and make the calls:
import { AzureChatOpenAI } from "@langchain/azure-openai";

const model = new AzureChatOpenAI({
modelName: "gpt-4-1106-preview",
// Note that the following are optional, and will default to the values below
// if not provided.
azureOpenAIEndpoint: process.env.AZURE_OPENAI_API_ENDPOINT,
Expand All @@ -72,7 +71,6 @@ const response = await model.invoke(new HumanMessage("Hello world!"));
import { AzureChatOpenAI } from "@langchain/azure-openai";

const model = new AzureChatOpenAI({
modelName: "gpt-4-1106-preview",
// Note that the following are optional, and will default to the values below
// if not provided.
azureOpenAIEndpoint: process.env.AZURE_OPENAI_API_ENDPOINT,
Expand All @@ -94,11 +92,41 @@ const embeddings = new AzureOpenAIEmbeddings({
// if not provided.
azureOpenAIEndpoint: process.env.AZURE_OPENAI_API_ENDPOINT,
azureOpenAIApiKey: process.env.AZURE_OPENAI_API_KEY,
azureOpenAIApiDeploymentName: process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME,
azureOpenAIApiDeploymentName: process.env.AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME,
});
const res = await embeddings.embedQuery("Hello world");
```

## Using Azure managed identity

If you're using [Azure Managed Identity](https://learn.microsoft.com/azure/ai-services/openai/how-to/managed-identity), you can also pass the credentials directly to the constructor:

```typescript
import { DefaultAzureCredential } from "@azure/identity";
import { AzureOpenAI } from "@langchain/azure-openai";

const credentials = new DefaultAzureCredential();

const model = new AzureOpenAI({
credentials,
azureOpenAIEndpoint: process.env.AZURE_OPENAI_API_ENDPOINT,
azureOpenAIApiDeploymentName: process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME,
});
```

## Compatibility with OpenAI API

This library is provides compatibility with the OpenAI API. You can use an API key from OpenAI's developer portal like in the example below:

```typescript
import { AzureOpenAI, OpenAIKeyCredential } from "@langchain/azure-openai";

const model = new AzureOpenAI({
modelName: "gpt-3.5-turbo",
credentials: new OpenAIKeyCredential("<your_openai_api_key>"),
});
```

## Development

To develop the Azure OpenAI package, you'll need to follow these instructions:
Expand Down
Loading

0 comments on commit c694a5d

Please sign in to comment.