A Node library designed to facilitate accessing Scribe's Private Documents (MI) API.
npm install @scribelabsai/mi
The constructor expects an environment object:
const EnvironmentSchema = object({
API_URL: string(),
IDENTITY_POOL_ID: string(),
USER_POOL_ID: string(),
CLIENT_ID: string(),
REGION: string(),
});
The API_URL
is "mi.scribelabs.ai/v1"
.
The REGION
is "eu-west-2"
.
Contact Scribe to obtain other details required for authentication.
import { ScribeMIClient } from '@scribelabsai/mi';
const client = new ScribeMIClient({
API_URL: 'mi.scribelabs.ai/v1',
REGION: 'eu-west-2',
IDENTITY_POOL_ID: 'Contact Scribe for authentication details',
USER_POOL_ID: 'Contact Scribe for authentication details',
CLIENT_ID: 'Contact Scribe for authentication details',
});
Authentication is handled by Scribe's Auth library, without the need for you to call that library directly.
// Authenticate with username / password
await client.authenticate({ username: 'myUsername', password: 'myPassword' });
// OR with refresh token
await client.authenticate({ refreshToken: 'myRefreshToken' });
The MI client will try to automatically re-authenticate with your refresh token, if you try to make an API call after credentials have expired.
const jobid: string = await client.submitTask(myFileBuffer, {
filetype: 'pdf',
filename: 'example-co-2023-q1.pdf',
companyname: 'Example Co Ltd',
});
The filetype
parameter is required: it should match the file's extension / MIME type.
Other parameters are optional:
filename
is recommended: it should be the name of the uploaded file. It appears in API responses and the web UI.companyname
can optionally be included for company Financials data: it should be the legal name of the company this document describes, so that documents relating to the same company can be collated.
The returned jobid
can be used to find information about the task status via getTask
, or via the web UI.
Fetch details of an individual task:
const task: MITask = await client.getTask(jobid);
console.log(task.status);
Or list all tasks:
const tasks: MITask[] = await client.listTasks();
After documents have been processed by Scribe, the task status (which can be seen via getTask
/ listTasks
) is "SUCCESS"
. At this point, you can export the model:
const task = await client.getTask(jobid);
// Use fetchModel
const model: MIModel = await client.fetchModel(task);
// Alternatively, fetch the model directly from its URL
return task.modelUrl;
In either case, note that the model is accessed via a pre-signed URL, which is only valid for a limited time after calling getTask
/ listTasks
.
When using Scribe to process fund data, multiple models can be consolidated for export in a single file:
const tasks = await client.listTasks();
const tasksToCollate = tasks.filter(
(task) => task.status === 'SUCCESS' && task.originalFilename.startsWith('Fund_1')
);
const collatedModel = await client.consolidateTasks(tasksToCollate);
const task = await client.getTask(jobid);
await client.deleteTask(task);
Deletion is irreversible.
After a successful deletion, the file, any output model, and any other file derived from the input are deleted permanently from Scribe's servers.
Documentation for the underlying REST API may also be useful, although we recommend accessing the API via this library or our Python SDK.