learndl-base-modules

name	learndl-base-modules
description	Routes agents to learndl base packages under `src/proj` (paths, DB, calendar, logging, Proj facade) and `src/func` (tensor/factor ops, numpy/pandas helpers). Use when editing or importing `proj` or `func`, configuring environments, logging, trade calendars, or shared numerics—not `src/res` strategy/factor pipelines or `src/data` loaders.

learndl base libraries (proj / func)

Use this file as a route map only; behavior and APIs live in source and module docstrings.

Router table

Need	Open first
Proj singleton, cwd, instance registry	src/proj/env/variable/proj.py, src/proj/env/variable/ins.py
Paths, data roots, config roots	src/proj/env/path.py, src/proj/env/machine.py
Logging, LogFile, file logging	src/proj/log/logger.py, src/proj/log/logfile.py, src/proj/env/variable/files.py
Trade calendar, offsets, business days	src/proj/calendar/calendar.py, trade_date.py, basic.py
data interface	src/proj/db/interface.py, io/saver.py, io/loader.py
Default configs (factor, model, trading, …)	src/proj/env/constant/conf/
HTTP, device, email, SQLite, misc	matching modules under src/proj/util/
Process pools, proxy/cache callers	src/proj/util/proxy/
Script locks, scheduling, CLI helpers	src/proj/util/script/
Factor tensors, rolling, neutralize, rank	src/func/tensor.py
Index align, fillna, index_merge	src/func/basic.py
NumPy/pandas transforms, OLS, cov	src/func/transform.py
Torch metrics, IC, weighted corr	src/func/metric.py
Symmetric orthogonalization	src/func/linalg.py
Package exports	src/proj/__init__.py, src/func/__init__.py

src/proj layout

• env/: MACHINE, PATH, CONST, Proj (machine and directory conventions).
• log/: Logger, LogFile; prefer src.proj.Logger in app code.
• cal/: CALENDAR, TradeDate, etc.—start here for session/backtest dates.
• db/: low-level access and mappers; not the same as src/data vendor loaders.
• core/: Silence, Duration, singleton, small shared abstractions.
• util/: general utilities—not src/func.
• util/func/: engineering helpers (parallel, disk, shell)—not the math package src/func.

src/func layout

• tensor.py: largest module; TsRoller, process_factor, neutralize_*, rank, corrwith, ts_*.
• basic.py: DIV_TOL, allna, index/intersect_* helpers.
• transform.py: NumPy/pandas/statsmodels-oriented.
• metric.py: torch-oriented metrics, ic / rankic variants.
• linalg.py: symmetric_orth (torch and numpy).

Easy to confuse

• src/func: shared quant math (from func import ... / import func.tensor).
• src/proj/util/func: project utilities; unrelated despite the name.

Outside this skill

• Pipelines and business logic: mostly src/data, src/res/.
• Longer operator docs: resources/instructions/ (e.g. uv_instruction.md).

How to use

1. Pick a row from the router table, open the file(s), read the module docstring at the top.
2. This copy lives under .cursor/skills/ so Cursor can discover it as a project skill.