ShirjanKhadka1/investment-analyst-toolkit

One command, no keys, no installs beyond Python 3.10+, a result in under 60 seconds:
bash
python -m toolkit.terminal demo.csv
demo.csv is synthetic sample data, not real market data: a seeded
random walk (seed 5, start 1000, daily drift +0.12%, volatility
1.2%), 80 weekday bars 2026-07-20 to 2026-11-06, NEPSE-style prices
in NPR, symbol DEMO. Generated 2026-11-07; the provenance is in the
file header, and the worked example below repeats it.
A free terminal that turns any daily-bars CSV into indicators, setups, honest Time-Machine odds and a risk gate — with every number reproducible.
> NEPSE, native: sourced frictions, effective-dated rules, the Time Machine and a zero-LLM rules mode — NEPSE is the tested, default market, and NEPSE data is CSV-only by design.
> World markets, supported: US, India, UK, crypto and FX via sourced market profiles, SEC EDGAR fundamentals and keyless public connectors — labelled tested or experimental in docs/global-evidence.md.
> Research and education only — not investment advice, never a buy/sell/hold recommendation. Every example uses clearly-labelled demo data unless a source is cited. The track-record badge reads n=0 honestly: no research call has been logged and settled yet (see docs/track-record.md), and it stays at n=0 until a real call settles.
## Install — first result in under 60 seconds
bash
git clone https://github.com/ShirjanKhadka1/investment-analyst-toolkit.git
cd investment-analyst-toolkit
python -m toolkit.terminal demo.csv
No pip install step: the toolkit is standard-library only, so there is nothing to install beyond Python 3.10+ itself. No API key, no account, no network. Timed 2026-11-07 on this machine: a clean clone plus the command above took 2.3 seconds end to end. (Your machine and network will differ; the point is that the only wait is the clone.)
## Watch it run (10 seconds)
The GIF above is rendered from a real run of the command (11 frames, one per four output lines; the full output is in docs/demo-transcript.md). The same output as an animated SVG: docs/demo.svg. Both are generated by scripts/generate_demo_assets.py — never hand-typed, and tests/test_demo_assets.py fails if the transcript ever drifts from a live run.
## Worked example — real numbers from demo.csv
demo.csv is synthetic example data (80 weekday bars, 2026-07-20 to 2026-11-06, symbol DEMO, NEPSE-style prices in NPR; provenance in the file header). One command prints four sections; these are the numbers it prints, pinned by tests/test_terminal.py and tests/test_demo_assets.py:
Indicator snapshot — last close 1284.91, above SMA20 (1166.08) and EMA20 (1186.73); RSI(14) 88.2; MACD 12/26/9 53.32, signal 41.82; Bollinger 20/2 %B 0.91; ATR(14) 16.79.
Screening — 2 of 5 named setups match: breakout_20d (close 1284.91 above the 20-day high 1274.72) and bull_flag (+12.07% run, last-5 range 3.24%). A match means the past closes satisfy a stated rule — not a buy/sell signal.
Time-Machine odds — hypothetical plan entry 1284.91, stop 1259.72, target 1335.29 (reward:risk 2.00): target first 57.5%, stop first 39.0%, neither 3.5% over 10,000 seeded replays; historical replay 68.9% target-first over 61 past windows; average realised R 0.79. Simulation, not a prediction.