gagara-boost-tsclient

TypeScript client for gagara-boost — a LightGBM training and prediction service.

Installation

npm install @gagara/gagara-boost-tsclient

Quick Start

import { GagaraBoostClient } from '@gagara/gagara-boost-tsclient';

const client = new GagaraBoostClient({
  baseUrl: 'https://gagara-boost.example.com',
  token: process.env.GAGARA_TOKEN,
});

// Health check
const ok = await client.health();

// Create a workspace
const workspace = await client.createWorkspace('my-workspace');

// Upload a dataset (Parquet)
const file = await fetch('/data/train.parquet').then((r) => r.arrayBuffer());
const upload = await client.uploadDataset(file, {
  workspaceId: workspace.id,
  alias: 'train-data',
  filename: 'train.parquet',
});

// Create a row set
const rowSet = await client.createRowSet({
  workspace_id: workspace.id,
  name: 'train-slice',
  base_dataset_id: upload.dataset_id,
  predicate: null,
});

API Overview

GagaraBoostClient

const client = new GagaraBoostClient({
  baseUrl: string,
  token?: string,
  fetch?: typeof fetch,
  timeout?: number,
});

Workspaces

• listWorkspaces()
• createWorkspace(name | { name })
• getWorkspace(id)
• renameWorkspace(id, name)
• deleteWorkspace(id)

Datasets

• listDatasets(workspaceId?)
• getDataset(id)
• uploadDataset(file, { workspaceId?, alias?, filename?, contentType? })
• downloadDataset(id)
• deleteDataset(id)
• updateDatasetAlias(id, alias)
• getDatasetMeta(id)
• refreshDatasetMeta(id)
• getDatasetSchema(id)
• listDatasetColumns(datasetId)
• createDatasetColumn(datasetId, payload)
• updateDatasetColumn(datasetId, columnId, payload)

Row Sets

• listRowSets(workspaceId?)
• getRowSet(id)
• createRowSet(payload)
• updateRowSet(id, payload)
• deleteRowSet(id)
• getRowSetSchema(id)
• getRowSetMeta(id, { force? })
• getRowSetSample(id)

Column Sets

• listColumnSets({ workspaceId?, datasetId? })
• getColumnSet(id)
• createColumnSet(payload)
• cloneColumnSet(id)
• updateColumnSet(id, payload)
• deleteColumnSet(id)

Training Param Sets

• listTrainingParamSets(workspaceId?)
• getTrainingParamSet(id)
• createTrainingParamSet(payload)
• updateTrainingParamSet(id, payload)
• deleteTrainingParamSet(id)

Models

• listModels({ workspaceId?, datasetId? })
• getModel(id)
• renameModel(id, name)
• deleteModel(id)

Actions

• train(request)
• calculateOptimalParamSet(request)
• predict(modelId, request)
• predictWithFreeParameter(modelId, request)

Error Handling

import { GagaraBoostError } from '@gagara/gagara-boost-tsclient';

try {
  await client.getWorkspace('missing');
} catch (err) {
  if (err instanceof GagaraBoostError) {
    console.error(err.status, err.message, err.body);
  }
}

Notes

• Gagara Boost expects a bearer token in the Authorization header for all requests.
• Datasets must be uploaded in Parquet format.

License

Apache-2.0