QuantCraft documentation reference

QuantCraft ships two complementary kinds of documentation: in-app topics you read inside the IDE, and strategy reference Markdown bundled with the product for deeper API detail. Together they describe how to write strategies, run backtests, and use the IDE. These web docs mirror many of those topics — start with Introduction or Getting started.

---

Where to read docs in the app

1. Open QuantCraft from the main app.
2. Expand the right sidebar (if it is collapsed).
3. Open the Documentation tab (alongside QuantCraft AI).

You land on a searchable index of topics. Type in the search box to filter by title, summary, or embedded keywords (for example on_finish, PaperAccount, chart_indicator_series).

Tap a topic card to open the full article. Use Back to return to the index.

---

Topic index (in-app)

Topics are grouped into two sections, matching the Documentation sidebar.

Getting started

Reference

Any card in the index that does not open a full article yet is labeled Soon in the UI.

---

How the in-app docs relate to your workflow

Use this quick map when you are doing something in the IDE:

The in-app articles are written for strategy authors and IDE users. Server operator setup for QuantCloud is not covered in these public docs.

---

Strategy imports

Write every import as quantcraft.*, with suffix-free module names (for example quantcraft.inputs, not quantcraft.inputs_module). The older ide.* and quantcraft.*_module paths still resolve, as does quantcraft.backtest.* — the package's former name before it was renamed to quantcraft.process.* — so existing strategies keep working unchanged — but do not write new code against them.

Shorter re-exports also work: quantcraft.indicator_series, quantcraft.account. For Fama-French, prefer from quantcraft import fama_french and read fama_french.qc inside callbacks.

Fama-French row access: dataset["rows"] is a YYYYMM → values dict, not a list. Use .row(...) or .rows_chronological(...); latest_row(...) only in forward runs (in a backtest it is look-ahead — see Fama-French factors). Do not use rows[0] / rows[-1].

Do not use internal packages such as quantcraft.b2.* (the data-download layer behind the price and fundamentals loaders) or quantcraft.ai_assistant.* in strategy scripts — those are internal helpers, not the strategy API. Do not confuse quantcraft.process.ohlcv (strategy OHLCV API) with quantcraft.b2.ohlcv (internal data layer).

Cross-sectional helpers

quantcraft.cross_sectional has vectorized helpers for comparing many symbols at once (numpy): stack_field, latest_return, latest_sma, latest_std, latest_rsi, rank, zscore, demean, winsorize, plus ColumnarBarArena.

Forward / bulk runs only as written. In a backtest, rt.qc_ohlcv_by_symbol holds each symbol's whole loaded range, so "latest" there means the last bar of the test — look-ahead. In a backtest, collect each symbol's closes from its bar as the run walks and feed those arrays to the helpers instead.

Module handles vs imported names

quantcraft.process.runtime and quantcraft.fama_french are module aliases, so attribute access reads the live per-session value:

Everything else re-exports names by value, so from ... import <name> binds a snapshot taken at import time.

For the handles the engine assigns per run, the engine rebinds those names in the root algo file after loading it, so the from ... import form is safe there — for the names it rebinds:

• Backtest / Debug Test: account, chart_indicators, get_bar, qc_ohlcv, qc_ohlcv_by_symbol, qc_ohlcv_by_tf, qc_fundamentals, qc_fama_french, qc_calendar.
• Forward runs: only account, qc_fundamentals and qc_fama_french — so for chart_indicators and the OHLCV handles use rt.<name> there.

A split helper module is different: it stays cached in sys.modules across runs in the same process, so an imported name keeps whichever session's handle was current at first import — stale for the next run (forward / bulk). In helper modules, import the module and use attribute access:

from quantcraft.inputs import params is the exception: params is one live object whose values change per run, so the from form is safe everywhere — see Input parameters.

---

Bundled strategy reference (Markdown)

For line-level module behavior, callback contracts, and engine details, the product also includes Markdown under the quantcraft-dev bundle (conceptually: “QuantCraft strategy reference”). A typical table of contents looks like this:

Those files are aimed at strategy authors who need exact names and contracts. Advanced topics such as BACKTEST_MODULE (WebSocket backtest entry) and IDE_BACKTEST_RUNNER (subprocess runner) are included in the bundle for contributors but are not primary user paths. QuantCraft AI in the IDE is preloaded with this material so answers stay aligned with the shipped modules.

---

Other help surfaces in QuantCraft

These are not the Documentation tab but are part of the same “reference ecosystem” for users:

• QuantCraft AI — Ask natural-language questions; answers can cite the same dev corpus. Connect a model under AI → API Providers (your own key, or QuantCraft Premium / hosted options).
• Import and export — Move .py (and related) files between disk and the workspace.
• Test results — Equity chart, metrics, trades, and HTML export after a backtest.

---

Search tips

• Prefer short tokens that appear in code: on_finish, params, chart_indicator_series, coerce_chart_time, pane_id.
• If nothing matches, clear the search — the index lists every topic by section.
• Cross-topic ideas (for example “fundamentals + on_bar”) often need two articles: open Fundamentals for data shape, then Strategy lifecycle for when it is available.

---

Document history

In-app Documentation is maintained alongside the strategy runtime. When the Python modules or backtest payload change, the in-app topics and bundled Markdown are updated together so examples and names stay in sync. Strategy import examples are written with suffix-free quantcraft.* paths throughout; the older ide.*_module and quantcraft.*_module paths are documented only as a reading aid for existing scripts.