Context API

Some Context features are available via an API. This allows developers to build customer applications using Context. We provide a TypeScript/JavaScript SDK for Node.js. Our API functionality is currently limited to the bot querying flow. We will adding support for other features and languages in the near future.

If you have a specific use case that you'd like to see supported, please contact us.

💡

This is the first version of our API. Please reach out to us if you have any questions or feedback.

Getting Started

Install the Context NPM package (opens in a new tab):

npm install --save @context-labs/sdk

In your application, import the Context SDK and instantiate a new Context client.

import { 
  ContextSDK, 
  ContextSDKError, 
  type QueryResponse, 
  type Query 
} from '@context-labs/sdk';
 
const context = new ContextSDK({}); // note: you must pass an empty object here for now

💡

You must pass an empty object to the constructor.

Query

Querying a bot is the process of asking a bot a question and receiving a response based on the sources that are connected to the bot.

To query a bot, call the query method on the client and provide an object containing the botId, query, and callbacks for the onData, onComplete, and onError parameters:

const main = async () => {
  const { controller }: QueryResponse = await context.query({
    botId: 'bJuJvrIyk',
    query: 'How can I get started with Solana?',
    onData: (data: string) => {
      console.log(data);
    },
    onComplete: (query: Query) => {
      console.log(query);
    },
    onError: (error: ContextSDKError) => {
      console.error(error);
    },
  });
};
 
main();

Read the description of each parameter below:

botId: The ID of the bot you wish to query. This field is available in the bot's URL. For example, the bot ID for the bot at https://app.usecontext.io/dashboard/bots/details?botId=bJuJvrIyk (opens in a new tab) is bJuJvrIyk.

query: The question you would like ask the bot as a string value.

onData: This optional callback is called for every chunk of data returned by Context. The data is returned as a string value. Use this callback to stream data in real-time to your application.

onComplete: This optional callback is called when data is done streaming. The value returned is a Query object. Use this callback to determine when the query is complete and inspect the query object for additional information.

onError: This optional callback is called when an error is encountered. A ContextSDKError value is passed, containing more details about the nature of the error. Use this callback to handle errors.

Setting Feedback

It is possible to set user feedback on a query. This is useful for letting administrators know if a query was helpful or not, and training the bot to return better results in the future.

The setQueryFeedback method accepts an object containing a string queryId field and a boolean helpful field indicating whether the query was helpful or not. This method returns a Query object.

const makeRequest = () =>
  new Promise<Query>((resolve, reject) => {
    context.query({
      botId: 'bJuJvrIyk',
      query: 'How can I get started with Solana?',
      onComplete: (query: Query) => {
        resolve(query);
      },
      onError: (error: ContextSDKError) => {
        console.error(error);
        reject(error);
      },
    });
  });
 
const main = async () => {
  const { _id: queryId } = await makeRequest();
  const query: Query = await context.setQueryFeedback({
    queryId,
    helpful: true,
  });
  console.log(query.helpful) // true
};
 
main();

Read the description of each parameter below:

query: The question you would like ask the bot as a string value.

onData: This optional callback is called for every chunk of data returned by Context. The data is returned as a string value. Use this callback to stream data in real-time to your application.

Searching

Searching a bot is the process of using semantic search (opens in a new tab) to find the the most releveant content from the sources that are connected to the bot.

To search a bot, call the search method on the client and provide an object containing the botId and query.

const main = async () => {
  const { snippets }: SemanticSearchResponse = await context.search({
    botId: 'bJuJvrIyk',
    query: 'How can I get started with Solana?',
  });
 
  console.log(snippets);
};
 
main();

Read the description of each parameter below:

query: The value you would like to search for as a string value.

Usage