Quant Buffet API

Sandbox rules

Allowed imports, builtins, limits, and error types.

Draft-preview and batch lab runs execute your code in `backtest/sandbox_runner.py`. Source is parsed with `ast`, stripped of bootstrap boilerplate, then `exec`'d with a restricted namespace.

Allowed import roots

backtest.* — data, engine, metrics, templates, universes
numpy / np, pandas / pd
math, json, typing, collections, dataclasses
functools, itertools, datetime, re, statistics, decimal
__future__

Blocked

Standard library modules outside the allow-list (os, sys, pathlib, subprocess, …).
Third-party packages (requests, sklearn, matplotlib, …).
Calls to eval, exec, compile, open, input, breakpoint.
async / await syntax.
Source files larger than 80 KB.

Note. Generated strategy files may include a sys.path bootstrap; backtest/lab_sanitize.py removes it before validation. Do not rely on that in hand-written lab code.

Safe builtins

A subset of Python builtins is injected: abs, min, max, sum, len, range, enumerate, zip, map, filter, sorted, container types, round, print, isinstance, hasattr, type, and common exceptions.

Runtime limits

Limit	Value
Max symbols in `ASSETS`	15
Min price history	30 rows after load
Min equity curve length	20 points after run
Equity chart points returned	90 (downsampled)
Default backtest start (payload)	2000-01-01 or lab UI date

Common error types

Type	Typical cause
`SyntaxError`	Invalid Python; line number returned when available.
`ContractError`	Missing `make_on_day` or wrong return shape.
`ValueError`	Bad import, whitelist symbol, empty ASSETS, signal never ready.
`ImportError`	Blocked module at runtime.
`RuntimeError`	No price data loaded for any symbol.