Sources

In Python V2, history.source(...) both declares a source and reads its history. There is no separate --source, TOML [sources], or script.sources declaration.

Selectors

<symbol>@<source>@<exchange>
<symbol>@<source>@<exchange>@<provider>

Add source options after the final segment:

candles = history.source("btc@candles@binancef:timeframe=60")
book = history.source("btc@orderbook@bulkf:depth=20", 0)
oi = history.source("btc@oi@hyperliquidf@mmt:timeframe=30")

Core futures use the base asset, such as btc. Hyperliquid HIP-3 uses {dex}:{coin}, such as xyz:tsla. Spot uses the exact pair, such as hype/usdc. Outcome symbols keep their side, such as 1009:0.

xyz_direct = history.source("xyz:tsla@candles@hyperliquidf:timeframe=60")
xyz_mmt = history.source("xyz:tsla@candles@hyperliquidf@mmt:timeframe=60")

Selector lookup is ASCII case-insensitive, so XYZ:TSLA and xyz:tsla resolve to the same declared feed. Market Lab keeps hyperliquidf as the canonical exchange and handles the provider-specific DEX route internally.

MMT selectors include both the indexed exchange and mmt:

candles = history.source("btc@candles@binancef@mmt:timeframe=60")
trades = history.source("btc@trades@binancef@mmt")

Declare selectors with literal strings or module-level string constants. Market Lab inspects them before the job starts so every subscription is deterministic. F-strings and other dynamic lookups may read only feeds declared elsewhere in the script.

Runtime access

With no index, history.source returns the retained list. Index 0 returns the newest record:

CANDLES = "btc@candles@binancef:timeframe=60"


def on_data(ctx, history):
    records = history.source(CANDLES)
    latest = history.source(CANDLES, 0)

Source options are part of the declaration, but Market Lab exposes a normalized identity on ctx:

ctx.source          # btc@candles@binancef
ctx.source_type     # candles
ctx.provider
ctx.exchange
ctx.symbol
ctx.source_configs

The market record remains in history; it is not copied onto ctx.

Availability

Provider and market type determine whether a source is valid. Spot and outcome markets do not provide perpetual-only OI.

Multiple sources

Declare each feed where it is read:

def on_data(ctx, history):
    btc = history.source("btc@candles@binancef:timeframe=900", 0)
    zec = history.source("zec@candles@binancef:timeframe=900", 0)
    if btc is None or zec is None:
        return

Then backtest without repeating the selectors:

mlab script backtest pairs.py \
  --from 2026-07-15 \
  --to 2026-07-16

Each selector gets an independent history buffer. See History.

Source references:

• Candles
• Orderbook
• Trades
• Volume Delta
• Open Interest
• Volumes