Build with TuluAI

01 / GETTING STARTED

Start with the public API

Public catalog reads do not require an API key. The production command uses the latest published build. To test changes in this workspace before publishing, run the preview command from Replit Shell.

curl https://tuluai.app/api/gena/datasets

curl "https://${REPLIT_DEV_DOMAIN}:5000/api/gena/datasets"

{
  "datasets": [
    {
      "id": "tulueng100-reference",
      "title": "TuluEng100",
      "language": "Tulu",
      "license": "..."
    }
  ]
}

02 / GENA DATA API

Datasets, available to build with

Gena makes public datasets discoverable through a small JSON API. Public reads need no credentials. Contributors can publish datasets directly after confirming their rights and public-release intent.

Publish a dataset

Send the current signed-in account token in the authorization header. The request must include a license, confirmation that you have the right to share the data, and confirmation that you want it published publicly.

curl -X POST https://tuluai.app/api/gena/datasets \
  -H 'Authorization: Bearer <signed-in-account-token>' \
  -H 'Content-Type: application/json' \
  -d '{
    "title": "Example collection",
    "description": "Short description of the data and its collection.",
    "language": "English",
    "license": "CC BY 4.0",
    "fileName": "examples.jsonl",
    "content": "{\"text\":\"Sample row\"}\n",
    "rightsConfirmed": true,
    "publicReleaseConfirmed": true
  }'

03 / OFFICIAL CLIENTS

Use Gena from your project

Both clients default to https://tuluai.app. Listing and downloading public datasets do not need an API key. Submitting data requires a current signed-in user token.

npm install @tuluai/gena-datasets

import { GenaDatasetsClient } from '@tuluai/gena-datasets'

const gena = new GenaDatasetsClient()
const datasets = await gena.listDatasets({ language: 'Tulu' })
const dataset = await gena.getDataset(datasets[0].id)

python -m pip install gena-datasets

from gena_datasets import GenaDatasetsClient

gena = GenaDatasetsClient()
datasets = gena.list_datasets(language="Tulu")
dataset = gena.get_dataset(datasets[0]["id"])

See the Gena library guide for complete install steps, read and download examples, dataset submission, and error handling.

04 / DEVELOPER API KEYS

Authenticate from your server

Create and revoke keys in the developer dashboard. A key is shown only once. Store it in your server-side secret manager and send it as a bearer token. Do not include it in browser or mobile code.

The versioned /api/v1 endpoints below require a key with the matching scope. For automated key management, GET and POST /api/developers/keysplus DELETE /api/developers/keys/{id} use your signed-in TuluAI account token.

curl -X POST https://tuluai.app/api/v1/chat \
  -H 'Authorization: Bearer <your-api-key>' \
  -H 'Content-Type: application/json' \
  -d '{"message":"Hello"}'

05 / INTEGRATIONS

Runtime and publishing credentials

Runtime credentials are managed by TuluAI. Developers should use scoped TuluAI API keys rather than provider credentials. Never put server keys in source control, client code, or public documentation.

06 / PRODUCT API ROUTES

Server-side model routes

Model credentials stay on TuluAI servers. Product clients call TuluAI routes rather than sending service credentials from the browser.

Product routes validate their own request shape and rate limits. Treat them as application endpoints, not as a promise of a stable third-party API contract.

07 / GENA NOTEBOOKS

Bring your own compute

Gena stores and organizes data. Notebook code runs in the browser, on the user’s local Gena Runner, or on a Jupyter service the user connects. The application server is not a notebook execution host.

• Install the desktop Runner from Gena Notebooks for local Python execution.
• Jupyter connections use the user’s own server URL and access token. Keep that token private and do not commit it.
• Notebook code runs with the connected computer’s file and network permissions. Review code before running it.

08 / SECURITY

Keep credentials and data safe

• Keep model-service, session, registry, and deployment secrets server-side in Replit Secrets or an ignored local environment file.
• Variables prefixed with NEXT_PUBLIC_ are intended for browser configuration. Never use that prefix for a private provider secret.
• Firebase web API configuration is public by design. Authentication and Firebase rules enforce access; the API key alone is not authorization.
• Use the Gena dataset clients for public data. Publishing requires a signed-in account, rights confirmation, and public-release confirmation.
• Gena processing and model training require explicit consent for the selected content. Check current provider pricing before any paid operation.