February 6, 2026

Streaming API responses

API responses now stream data in chunks instead of buffering the entire result before sending. For large query results, this reduces time to first byte and lets clients start processing data while the query is still running.

The streaming works transparently, so no changes to your API calls are required.

Auto-promote deployments

The /v1/deploy API now accepts an auto_promote parameter. When set to true, the deployment automatically promotes to live and deletes the previous deployment once data is ready. No need to poll for status and make a separate promote call.

curl -X POST "https://api.tinybird.co/v1/deploy?auto_promote=true" \
  -H "Authorization: Bearer $TOKEN"

Faster deployment promotions

Deployment promotions are now faster. Instead of recreating all Data Sinks during promotion, Tinybird now detects which sinks changed and only updates those. Sinks that haven't changed are left untouched. When multiple sinks need updating, they're processed in parallel.

IMPORT_FORMAT inference for S3 and GCS

The IMPORT_FORMAT property for S3 and GCS Data Sources is now optional. If you don't specify it, the format is automatically inferred from the file extension (.csv, .ndjson, .parquet, etc.). Requires CLI 3.0.2.

Bug fixes and improvements

  • [All]: Invalid JSON paths in Data Source imports (e.g., $.alerts instead of $.alerts[:]) now return a 400 Bad Request error with a helpful message, instead of a 500 Internal Server Error.
  • [Forward]: The "Open in Explorations" menu item has been updated to "Open in Playgrounds".

Deprecation notice

Starting February 22, 2026, the endpoint charts feature will no longer be supported, which means that calls to /v0/pipes/*/endpoint/charts/* will return an HTTP 404 Not Found response.

From the blog

How we built Playgrounds: A technical deep-dive into the architecture behind Playgrounds. Multi-node SQL queries, workspace sharing, split-screen for better context, and AI-assisted SQL with diff preview.

How Fever made real-time partner reporting reliable under load with Tinybird: A live entertainment platform processing 300TB per month replaced a CDC pipeline from Postgres using MSK and ClickHouse® with Tinybird. The result: stable latency under load, automatic PR-based schema changes, and quarantine for bad records.

TypeScript SDK (preview)

We're working on a TypeScript SDK that makes Tinybird native to your codebase. Data Sources and Pipes defined in TypeScript, fully typed, with CLI commands to let AI coding agents build, test, and deploy.

Define your schema with full type inference:

import { defineDatasource, t, engine } from "@tinybirdco/sdk";

export const pageViews = defineDatasource("page_views", {
  schema: {
    timestamp: t.dateTime(),
    pathname: t.string(),
    session_id: t.string(),
    country: t.string().lowCardinality().nullable(),
  },
  engine: engine.mergeTree({
    sortingKey: ["pathname", "timestamp"],
  }),
});

Ingest with type-safe payloads:

await tinybird.ingest.pageViews({
  timestamp: new Date(),
  pathname: "/home",
  session_id: "abc123",
  country: "US",
});

Install with npm install @tinybirdco/sdk. Check the GitHub repo to follow progress.

View previous update