NodeJS client for SOCO.ai.
npm install soco.aiimport { SocoClient, Config } from 'soco.ai';
const config: Config = {
adminApiKey: '...', // needed for Admin APIs
clientId: '...', // needed for APIs & Parse Doc API
queryApiKey: '...',// needed for Query APIs
};
const client = new SocoClient(config);Using:
import { QueryRequest } from 'soco.ai';
const request: QueryRequest = {
uid: 'ID of the end user',
query: {
query: 'who is Jimmy?',
n_best: 3,
func_type: 'qa_faq_combine'
}
};
const answers = await client.query(request);qa- question and answer is finding the answer directly in the textfaq- frequently asked questions is finding the most similar questionslm- language model, is a task to predict the next work. For example, the output of lm(“nice to meet”) will be “you”qa_faq- returnsqaandfaqin one request as two separate arraysqa_faq_combine- returnsqaandfaqtogether in one array, sorted by score
It is possible to pass waitUntilOpFinished to SocoClient functions, that call APIs that return an operation ID,
which will make the function wait until the operation is finished or failed.
The default value of waitUntilOpFinished is false. If waitUntilOpFinished set to true,
const waitUntilOpFinished = true;
const response = await client.addDocs(request, waitUntilOpFinished);The function that waits for operation to be finished is available as well.
const operationId = '...';
const status = client.waitUntilFinished(operationId);Using:
Using:
import { AddDocsRequest } from 'soco.ai';
const request: AddDocsRequest = {
data: [
{
data: [{ text: 'there was bear called Jimmy', type: 'content' }],
meta: { doc_id: '123' },
doc_type: 'soco_doc'
}
],
auto_index: true
};
const response = await client.addDocs(request, true);Using:
- https://docs.soco.ai/soco-api/data-process/json-files-crud-operation/add-documents
- https://docs.soco.ai/soco-api/data-process/file-type#soco_faq
const filePath = 'test/e2e/fixtures/dataset-1'; // dir with CSV files
const response = await client.addFAQsFromCSVs(filePath, {}, true, true);Using:
- https://docs.soco.ai/soco-api/data-process/parse-raw-files
- https://docs.soco.ai/soco-api/data-process/json-files-crud-operation/add-documents
import { ParseDocRequest } from 'soco.ai';
const parseRequest: ParseDocRequest = {
client_id: 'a client ID',
file_type: 'url',
file_url: 'http://qiiip.org/Adobe Acrobat Reader FAQs.pdf',
lang: 'en',
};
const response = await client.addUnstructuredDoc(parseRequest, undefined, true, true);Using:
import { ReadDocsRequest } from 'soco.ai';
const request: ReadDocsRequest = {
skip: 0,
limit: 10
};
const response = await client.readDocs(request);Using:
import { UpdateDocsRequest } from 'soco.ai';
const request: UpdateDocsRequest = {
data: [
{
_id: '...',
data: [{ text: 'ok, lets talk about something else', type: 'content' }],
meta: { someMeta: 'mata info is mandatory' },
doc_type: 'soco_doc'
}
]
};
const response = await client.updateDocs(request, true);Using:
If data is left empty or undefined, it removes all the documents.
import { DeleteDocsRequest } from 'soco.ai';
export const request: DeleteDocsRequest = {
doc_ids: undefined,
auto_index: false,
};
const response = await client.deleteDocs(request, true);Using:
If data is left empty or undefined, it removes all the documents.
import { AggregateDocsRequest } from 'soco.ai';
const request: AggregateDocsRequest = { field: '$meta.doc_id' };
const response = await client.aggregateDocs(request);Using:
import { ReindexRequest } from 'soco.ai';
const request: ReindexRequest = {
params: { lm: {}, qa: {}, kw: {}, qq: {}, tuple: {} },
};
const answers = await client.reindex(request);Using:
import { RefreshRequest } from 'soco.ai';
const request: RefreshRequest = {
params: { lm: {}, qa: {}, kw: {}, qq: {}, tuple: {} }
};
const answers = await client.refresh(request);Using:
const operationId = response.op_id;
const status = await client.getOperationStatus(operationId);Using:
const operations = await client.readOperations(0, 10, -1);When API fails to return the result, it SocoClient functions throw an error with all the details.
CSV files with FAQ need to be in this structure. There can/should be multiple questions leading to a single answer to improve matching accuracy.
question ,answer
how are you doing,I am doing fine
how is it going ,I am doing fine
how is life ,It is fine
how old are you ,I am 42 years old
what is your age ,I am 42 years old
import { SocoClient, Config, ReindexRequest, ParseDocRequest } from "soco.ai";
const pdfFile = "http://path to your pdf.pdf"; // there is no OCR, plain text needs to be in the PDF
const csvFolder = "/Users/path to your folder with CSV files";
async function loadData() {
// create config with secrets
const config: Config = {
adminApiKey: "copy paste the key here from https://app.soco.ai",
clientId: "copy paste the key here from https://app.soco.ai",
queryApiKey: "copy paste the key here from https://app.soco.ai"
};
// create SOCO client
const client = new SocoClient(config);
// delete all the docs from SOCO (if needed, to ensure clean start)
await client.deleteDocs({ doc_ids: undefined, auto_index: true }, true);
// refresh all indexes - it can take a lot of time, better to reindex only what needs to be reindexed
const request: ReindexRequest = {
params: { lm: {}, qa: {}, kw: {}, qq: {}, tuple: {} }
};
await client.reindex(request, true);
// add FAQs from CSV files
await client.addFAQsFromCSVs(csvFolder, {}, true, true);
// add content of PDF file (unstructured data)
const parseRequest: ParseDocRequest = {
client_id: config.clientId,
file_type: "url",
file_url: pdfFile,
lang: "en"
};
await client.addUnstructuredDoc(parseRequest, {}, true, true);
// refresh all indexes
await client.reindex(request, true);
}
loadData()
.then(_ => console.log("Done"))
.catch(error => console.error(error));