For the complete documentation index, see llms.txt. This page is also available as Markdown.

Product updates

New exchange coverage, channels, clients and API changes

Data Coverage

Hyperliquid fast order book normalization

Hyperliquid order book normalization now uses the 5-level fastBook feed for real-time streaming and historical data recorded from 2026-06-17 onward. This applies to the Node.js client and the normalized book_snapshot_5 and book_snapshot_25 CSV datasets. From that date, both datasets are derived from snapshots currently observed around every 500ms. Earlier historical data continues to use the 20-level l2Book feed.

Available in Node.js client tardis-dev 18.1.0 or newer.

See Hyperliquid coverage.

Data Coverage

Gemini WebSocket v3 market data

Gemini raw historical feeds now use Gemini WebSocket API v3 from 2026-07-24 UTC onward. Captured v3 data includes public trades on trade, 100ms incremental level 2 order book updates on the Tardis depth channel, and native top-of-book updates on bookTicker.

Gemini v3 raw payloads use lowercase symbols such as btcusd. For backward compatibility with Gemini v2, Tardis symbol IDs and normalized messages remain uppercase, while Gemini filters match symbols case-insensitively so existing filters such as BTCUSD continue to work. Historical data recorded before the switch remains available in the legacy v2 format.

See Gemini coverage.

Data Coverage

Kraken WebSocket v2 market data

Kraken raw historical feeds now use Kraken WebSocket API v2 from 2026-07-10 UTC onward. Captured v2 data includes public trades, 1000-level order books, BBO-triggered tickers, and instrument reference data including tokenized assets.

Kraken v2 also changes exchange-native symbol IDs for pairs containing Bitcoin or Dogecoin: v1 XBT and XDG become v2 BTC and DOGE. For example, use XBT/USD before the switch and BTC/USD from 2026-07-10 onward. This applies to every pair containing either currency.

Kraken v2 real-time streaming and data normalization require Node.js client tardis-dev 16.6.3 or newer, or Tardis Machine 16.6.2 or newer.

See Kraken coverage.

Data Coverage

MEXC Spot and Futures market data

High frequency historical market data is now available for MEXC Spot and MEXC Futures exchanges.

Data normalization for both exchanges is available in Node.js client tardis-dev and Tardis Machine 16.6.0 or newer.

Billing

Account-level data transfer limits

Data transfer limits now apply at the account level for CSV dataset downloads and raw data replay API responses.

See API rate limits.

Data Coverage

Bybit pre-launch derivatives data

Market data for Bybit Derivatives pre-launch instruments is now available once Bybit moves an instrument into the ContinuousTrading phase.

These symbols are included in Bybit Derivatives historical and real-time feeds for supported channels, including trades, order book data, tickers, liquidations, and book ticker data. The Instruments Metadata API marks them with underlyingType: pre_market, so clients can distinguish pre-launch contracts from normal crypto-native contracts. Bybit auction-only pre-launch phases remain excluded until continuous trading starts.

See Bybit Derivatives coverage.

Data Coverage

Hyperliquid fast order book feed

Hyperliquid raw historical feeds now include the fastBook channel from 2026-06-17 onward. It is Hyperliquid l2Book with fast: true, exposed as a separate raw channel with 5-level order book snapshots currently observed around every 500ms.

See Hyperliquid coverage.

API

Faster exchange and instrument metadata APIs

Exchange details and Instruments Metadata API responses are now faster and more reliable for large exchanges and broad filters, including options-heavy venues and filters for futures, perpetuals, and options.

Metadata responses now return the full available instrument universe more consistently, including historical or inactive symbols that previously could be omitted on exchanges with very large symbol lists. This makes symbol discovery, availability checks, and automated instrument selection more predictable across supported exchanges.

Metadata refreshes are also more resilient to temporary upstream issues, so newly listed, delisted, and recovered symbols should appear more consistently. The /v1/exchanges/:exchange response no longer includes the legacy zero-valued datasets.stats placeholder; use datasets.symbols for CSV dataset availability.

Data Coverage

Polymarket market data

Historical Polymarket market data is now available for CLOB order books, trades, top-of-book, market lifecycle events, and sports results.

Polymarket replay and data normalization support is available in Node.js client tardis-dev 16.5.0 and Tardis Machine 16.5.0.

See Polymarket coverage.

Data Coverage

OKX order book and Tokyo migration

OKX Spot, Futures, Swap, and Options raw order book feeds now use the public books channel instead of the colocation-only books-l2-tbt channel from 2026-05-21 UTC onward.

The OKX recorders were moved from Hong Kong to Tokyo after OKX announced its Hong Kong to Tokyo trading server migration and updated its colocation policy to support DMM/DMT clients only. We were asked to switch from colocation domains to the public network domains www.okx.com and ws.okx.com. A migration data gap affected OKX feeds on 2026-05-21 from 00:00:00 UTC to 05:16:19 UTC.

Replaying OKX Spot, Futures, Swap, and Options order book data recorded from 2026-05-21 onward requires Node.js client tardis-dev 16.4.2 or newer, or Tardis Machine 16.4.2 or newer.

Data Coverage

Bullish market data

Historical Bullish market data is now available, including order book updates, trades, top-of-book updates, tickers, index prices, and perpetual and options data.

Bullish data normalization is available in Node.js client tardis-dev 16.4.0 and Tardis Machine 16.4.1.

See Bullish coverage.

API

Optimized filtered raw data replay API requests

Filtered raw data replay API requests can now use the optional sliceSize parameter to fetch up to 10 consecutive minutes per request. Use it to reduce the number of API requests when replaying historical data. Omit sliceSize to keep the existing one-minute behavior.

Responses include x-slice-size with the slice size used and x-suggested-slice-size as a recommendation for future requests with the same filter applied. The recommendation balances fewer requests against response size.

The Node.js client tardis-dev 16.3.1, Python client tardis-dev 4.2.0, and Tardis Machine 16.2.1 use this automatically for filtered historical replay, so library users get fewer API requests without changing replay code.

Data Coverage

Bitget V3 market data

Bitget Spot and Bitget Futures raw historical feeds now use Bitget WebSocket API V3. Order book collection now uses the books full-depth channel.

Bitget Futures also includes the V3 liquidation channel. Replaying or normalizing Bitget data recorded from 2026-04-28 onward requires Node.js client tardis-dev 16.2.1 or newer, or Tardis Machine 16.1.2 or newer, because older clients do not support the new V3 raw message formats.

Clients

Find instrument symbols

You can now find exchange-specific instrument symbols from normalized market filters across multiple exchanges. Use this when you know the market you want, such as active BTC/USDT linear perpetuals, but do not want to hardcode each exchange's symbol format.

The new Node.js findInstrumentSymbols() helper and Python find_instrument_symbols() helpers use Instruments Metadata API filters and return symbols grouped by exchange.

Available in Node.js client tardis-dev 16.2.0 and Python client tardis-dev 4.1.0.

API

Instruments metadata min notional and underlying type

Instruments metadata now includes minNotional when an exchange exposes a minimum order value or notional separately from minimum order quantity.

It also includes underlyingType for non-option instruments, using native for crypto-native instruments and exchange-sourced classifications such as equity, commodity, fx, index, fixed_income, or pre_market when available. The instruments filter endpoint accepts underlyingType, so you can query categories such as commodity perpetuals on supported exchanges.

See Instruments Metadata API.

Data Coverage

Lighter market data

Historical Lighter market data is now available, including order book updates, trades, tickers, and perpetual and spot market stats.

Data normalization for Lighter is available in Node.js client tardis-dev 16.1.1 and Tardis Machine 16.1.0.

See Lighter coverage.

Data Coverage

Binance index constituents

You can now replay historical Binance USDS-M Futures index price constituents through the generated indexConstituents channel.

Binance's REST endpoint exposes the current index constituents. We record those payloads into the minute-by-minute historical feed, so you can inspect how each index's constituent exchanges, symbols, prices and weights changed over time.

See Binance USDS-M Futures coverage.

Data Coverage

Binance premium index data

You can now replay Binance USDS-M Futures premium index snapshots through the generated premiumIndex channel.

This adds access to Binance's REST premium index payloads in the same minute-by-minute historical feed format as our other channels, including interestRate, which is not available in the WebSocket markPrice stream.

See Binance USDS-M Futures coverage.

Clients

zstd historical replay

Historical replay in the Node.js and Python clients now prefers zstd for data feed responses and local cache files.

The result is smaller replay payloads by default, while keeping gzip available when you need gzip-compressed slices.

Clients

Unified Python client v3

The Python client is now a unified tardis-dev client for historical replay, CSV dataset downloads, cache management and exchange metadata.

It exposes a simpler top-level API for the most common workflows:

  • replay() for historical market data replay

  • download_datasets() for CSV dataset downloads

  • clear_cache() for local cache cleanup

  • get_exchange_details() for exchange metadata

replay() also accepts Python datetime values for from_date and to_date. Naive datetimes are treated as UTC.

Python client quickstart · Migration notice

Clients

ESM Node.js client

The Node.js client is now ESM-only, with simplified package exports and examples that use native ES module syntax.

Node.js client quickstart

Clients

Tardis Machine Docker images

Tardis Machine Docker images are now published for both x64 and ARM64 architectures, making local replay and streaming easier to run on Apple Silicon machines, ARM servers and standard x64 environments.

Tardis Machine quickstart

Clients

Binance Futures WebSocket endpoints

Our backend services, Tardis Machine, and the Node.js client now handle Binance Futures' split public WebSocket endpoints, keeping real-time streaming aligned with Binance's exchange-side API changes.

This shipped in tardis-dev 14.3.0 and tardis-machine 14.3.0.

See Binance's Important WebSocket Change Notice.

Clients

Gate.io Futures decimal sizes

Tardis Machine and the Node.js client now handle Gate.io Futures decimal-size metadata via X-Gate-Size-Decimal, so normalized sizes stay correct for affected contracts.

Data Coverage

Binance funding info

Generated fundingInfo channels are now available for Binance USDS-M Futures and Binance COIN Futures.

These channels expose Binance funding cap, floor and interval updates in the historical data feed when Binance publishes adjusted funding information.

Binance USDS-M Futures · Binance COIN Futures

Data Coverage

Binance insurance balance

The generated insuranceBalance channel is now available for Binance USDS-M Futures, with snapshots from Binance's insurance fund balance endpoint.

Last updated

Was this helpful?