Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

EVM and Solana query filters now accept a single value or a list of values (OR-match), matching the existing Hyperliquid get_fills behavior. Applies to get_transactions, get_logs, get_transfers, get_traces, get_state_diffs (EVM) and get_instructions, get_transactions, get_balances, get_token_balances, get_rewards, get_logs (Solana).
prefetch_workers parameter on QueryCursor / with_progress() to configure the number of concurrent prefetch shard workers (clamped to 16).
sqd_py[fast] extra pulling in uvloop (auto-enabled on import when installed; not available on Windows).

Prefetch is now enabled by default (2 workers) whenever the end of the query range is known, falling back to a single serial worker otherwise. The shards parameter is kept as a deprecated alias for prefetch_workers.
Streaming JSON parsing switched from ujson to orjson.

Expanded Dataset / EvmDataset with 72 EVM chains served by the SQD portal (Arbitrum, Base, Optimism, Polygon, zkSync, and many more), plus Dataset.HYPERLIQUID for the HyperEVM chain.

Hyperliquid (HyperCore fills) dataset support via SQD(dataset="hyperliquid-fills") and get_fills(...). Filters (coin, user, builder, dir, fee_token, cloid) accept a single value or a list (OR-match).
examples/hyperliquid_fills.py and examples/hyperliquid_candles.py (OHLCV aggregation derived from fills).

Retried prefetch range streaming on transient aiohttp client errors to avoid crashing on truncated responses
Disabled tqdm progress bars when stderr is not a TTY (or TERM=dumb) to avoid duplicated output in unsupported terminals, with a plain-text fallback
Added ETA, rate, and latest block details to the plain-text progress fallback
Computed plain-text rates from time deltas to avoid 0.00/s on fast block streams
Show sub-minute elapsed time with fractional seconds in plain-text logs
Show sub-second elapsed time with millisecond precision in plain-text logs
Added day-level formatting for long ETAs
Replaced multi-shard parallel fetching with a two-range prefetch pipeline
Clamped shards to a two-range prefetch window for historical catchup
Made prefetch opt-in (default shards=1) and marked it experimental

Fixed progress bar disconnect when switching from parallel catchup to live mode: previously, the progress bar would switch to "Live" mode immediately when parallel workers finished fetching, even though the consumer was still processing queued blocks. Now, the switch only happens after all catchup blocks have been consumed

Fixed graceful shutdown in parallel mode: previously, all queued blocks would still be sent to the consumer during shutdown. Now, when shutdown is requested (Ctrl+C / SIGTERM), block forwarding stops immediately

Added default None values to all optional parameters in EVMQuery request builder methods (add_transactions_request, add_logs_request) to make them easier to call directly