This will help you getting started with IBM watsonx.ai chat models. For detailed documentation of all IBM watsonx.ai features and configurations head to the IBM watsonx.ai.

​
Overview

​
Integration details

ClassPackageLocalSerializablePY supportPackage downloadsPackage latest
ChatWatsonx@langchain/community❌✅✅NPM - DownloadsNPM - Version

​
Model features

Tool callingStructured outputJSON modeImage inputAudio inputVideo inputToken-level streamingToken usageLogprobs
✅✅✅✅❌❌✅✅❌

​
Setup

To access IBM watsonx.ai models you’ll need to create a/an IBM watsonx.ai account, get an API key, and install the @langchain/community integration package.

​
Credentials

Head to IBM Cloud to sign up to IBM watsonx.ai and generate an API key or provide any other authentication form as presented below.

​
IAM authentication

export WATSONX_AI_AUTH_TYPE=iam
export WATSONX_AI_APIKEY=<YOUR-APIKEY>

​
Bearer token authentication

export WATSONX_AI_AUTH_TYPE=bearertoken
export WATSONX_AI_BEARER_TOKEN=<YOUR-BEARER-TOKEN>

​
IBM watsonx.ai software authentication

export WATSONX_AI_AUTH_TYPE=cp4d
export WATSONX_AI_USERNAME=<YOUR_USERNAME>
export WATSONX_AI_PASSWORD=<YOUR_PASSWORD>
export WATSONX_AI_URL=<URL>
Once these are places in your enviromental variables and object is initialized authentication will proceed automatically. Authentication can also be accomplished by passing these values as parameters to a new instance.

​
IAM authentication

import { WatsonxLLM } from "@langchain/community/llms/ibm";

const props = {
  version: "YYYY-MM-DD",
  serviceUrl: "<SERVICE_URL>",
  projectId: "<PROJECT_ID>",
  watsonxAIAuthType: "iam",
  watsonxAIApikey: "<YOUR-APIKEY>",
};
const instance = new WatsonxLLM(props);

​
Bearer token authentication

import { WatsonxLLM } from "@langchain/community/llms/ibm";

const props = {
  version: "YYYY-MM-DD",
  serviceUrl: "<SERVICE_URL>",
  projectId: "<PROJECT_ID>",
  watsonxAIAuthType: "bearertoken",
  watsonxAIBearerToken: "<YOUR-BEARERTOKEN>",
};
const instance = new WatsonxLLM(props);

​
IBM watsonx.ai software authentication

import { WatsonxLLM } from "@langchain/community/llms/ibm";

const props = {
  version: "YYYY-MM-DD",
  serviceUrl: "<SERVICE_URL>",
  projectId: "<PROJECT_ID>",
  watsonxAIAuthType: "cp4d",
  watsonxAIUsername: "<YOUR-USERNAME>",
  watsonxAIPassword: "<YOUR-PASSWORD>",
  watsonxAIUrl: "<url>",
};
const instance = new WatsonxLLM(props);
If you want to get automated tracing of your model calls you can also set your LangSmith API key by uncommenting below: 
# export LANGSMITH_TRACING="true"
# export LANGSMITH_API_KEY="your-api-key"

​
Installation

The LangChain IBM watsonx.ai integration lives in the @langchain/community package: 
import IntegrationInstallTooltip from "@mdx_components/integration_install_tooltip.mdx";
<IntegrationInstallTooltip></IntegrationInstallTooltip>

<Npm2Yarn>
  @langchain/community @langchain/core
</Npm2Yarn>

​
Instantiation

Now we can instantiate our model object and generate chat completions: 
import { ChatWatsonx } from "@langchain/community/chat_models/ibm";
const props = {
  maxTokens: 200,
  temperature: 0.5
};

const instance = new ChatWatsonx({
  version: "YYYY-MM-DD",
  serviceUrl: process.env.API_URL,
  projectId: "<PROJECT_ID>",
  // spaceId: "<SPACE_ID>",
  // idOrName: "<DEPLOYMENT_ID>",
  model: "<MODEL_ID>",
  ...props
});
Note:
  • You must provide spaceId, projectId or idOrName(deployment id) unless you use lighweight engine which works without specifying either (refer to watsonx.ai docs)
  • Depending on the region of your provisioned service instance, use correct serviceUrl.

​
Invocation

const aiMsg = await instance.invoke([{
  role: "system",
  content: "You are a helpful assistant that translates English to French. Translate the user sentence.",
},
{
  role: "user",
  content: "I love programming."
}]);
console.log(aiMsg)

AIMessage {
  "id": "chat-c5341b2062dc42f091e5ae2558e905e3",
  "content": " J'adore la programmation.",
  "additional_kwargs": {
    "tool_calls": []
  },
  "response_metadata": {
    "tokenUsage": {
      "completion_tokens": 10,
      "prompt_tokens": 28,
      "total_tokens": 38
    },
    "finish_reason": "stop"
  },
  "tool_calls": [],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "input_tokens": 28,
    "output_tokens": 10,
    "total_tokens": 38
  }
}

console.log(aiMsg.content)

 J'adore la programmation.

​
Chaining

We can chain our model with a prompt template like so: 
import { ChatPromptTemplate } from "@langchain/core/prompts";

const prompt = ChatPromptTemplate.fromMessages(
  [
    [
      "system",
      "You are a helpful assistant that translates {input_language} to {output_language}.",
    ],
    ["human", "{input}"],
  ]
)
const chain = prompt.pipe(instance);
await chain.invoke(
    {
      input_language: "English",
      output_language: "German",
      input: "I love programming.",
    }
  )

AIMessage {
  "id": "chat-c5c2c08d3c984254acc48225c39c6a08",
  "content": " Ich liebe Programmieren.",
  "additional_kwargs": {
    "tool_calls": []
  },
  "response_metadata": {
    "tokenUsage": {
      "completion_tokens": 8,
      "prompt_tokens": 22,
      "total_tokens": 30
    },
    "finish_reason": "stop"
  },
  "tool_calls": [],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "input_tokens": 22,
    "output_tokens": 8,
    "total_tokens": 30
  }
}

​
Streaming the Model output

import { HumanMessage, SystemMessage } from "@langchain/core/messages";

const messages = [
    new SystemMessage('You are a helpful assistant which telling short-info about provided topic.'),
    new HumanMessage("moon")
]
const stream = await instance.stream(messages);
for await(const chunk of stream){
    console.log(chunk)
}

 The
 Moon
 is
 Earth
'
s
 only
 natural
 satellite
 and

​
Tool calling

import { tool } from "@langchain/core/tools";
import { z } from "zod";

const calculatorSchema = z.object({
    operation: z
      .enum(["add", "subtract", "multiply", "divide"])
      .describe("The type of operation to execute."),
    number1: z.number().describe("The first number to operate on."),
    number2: z.number().describe("The second number to operate on."),
  });
  
const calculatorTool = tool(
async ({ operation, number1, number2 }) => {
    if (operation === "add") {
    return `${number1 + number2}`;
    } else if (operation === "subtract") {
    return `${number1 - number2}`;
    } else if (operation === "multiply") {
    return `${number1 * number2}`;
    } else if (operation === "divide") {
    return `${number1 / number2}`;
    } else {
    throw new Error("Invalid operation.");
    }
},
{
    name: "calculator",
    description: "Can perform mathematical operations.",
    schema: calculatorSchema,
}
);

const instanceWithTools = instance.bindTools([calculatorTool]);

const res = await instanceWithTools.invoke("What is 3 * 12");
console.log(res)

AIMessage {
  "id": "chat-d2214d0bdb794483a213b3211cf0d819",
  "content": "",
  "additional_kwargs": {
    "tool_calls": [
      {
        "id": "chatcmpl-tool-257f3d39532141b89178c2120f81f0cb",
        "type": "function",
        "function": "[Object]"
      }
    ]
  },
  "response_metadata": {
    "tokenUsage": {
      "completion_tokens": 38,
      "prompt_tokens": 177,
      "total_tokens": 215
    },
    "finish_reason": "tool_calls"
  },
  "tool_calls": [
    {
      "name": "calculator",
      "args": {
        "number1": 3,
        "number2": 12,
        "operation": "multiply"
      },
      "type": "tool_call",
      "id": "chatcmpl-tool-257f3d39532141b89178c2120f81f0cb"
    }
  ],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "input_tokens": 177,
    "output_tokens": 38,
    "total_tokens": 215
  }
}

​
API reference

For detailed documentation of all IBM watsonx.ai features and configurations head to the API reference: API docs