Skip to content

README.md

Latest commit

 

History

History
106 lines (82 loc) · 4.84 KB

README.md

File metadata and controls

106 lines (82 loc) · 4.84 KB
Raw

Examples

Worked examples of using anomalyx on real data. These live outside the Cargo workspace (they shell out to the installed anomalyx binary), so they don't affect the build or the gates.

stock_anomalies.py

Fetches a stock's daily history from Yahoo Finance, enriches it with daily-return and intraday-range columns, runs anomalyx scan, and prints the anomalous trading days — mapping each finding's handle back to a calendar date. It's a compact demonstration of consuming the tq1 contract: it parses the dense JSON envelope (the dictionary + dense finding rows), not pretty text.

pip install yfinance          # one-time
cargo install anomalyx        # or set $ANOMALYX to the binary path

# Anomalous trading days within one ticker (point / multivariate / collective):
python3 examples/stock_anomalies.py NVDA --period 2y

# Only the strongest, with false-discovery-rate control:
python3 examples/stock_anomalies.py NVDA --period 2y --fdr 0.01 --min-severity high

# Distributional drift of one ticker's behavior against another:
python3 examples/stock_anomalies.py NVDA --period 1y --baseline AMD

Any extra flags are passed straight through to anomalyx scan (e.g. --top 20, --no-column-roles). The exit code mirrors anomalyx: 0 clean, 1 anomalies found, 2 error.

On real NVDA history this surfaces, for example, the 2025‑01‑27 DeepSeek selloff (top volume + the single largest multivariate outlier), the April‑2025 tariff volatility, and the second‑half‑2025 price regime shift (coll.cusum) — and in --baseline mode, that NVDA's volume and volatility distributions differ sharply from a peer's.

synergy_market.py

Pairs anomalyx with agent-calc — another contract-first CLI, an exact math kernel — on the live market. Two typed-JSON contracts chained: anomalyx is descriptive (which days/regimes broke the pattern, assumption-free); agent-calc is exact (what those findings mean as deterministic statistics).

pip install yfinance
cargo install anomalyx                                  # or set $ANOMALYX
(cd ../agent-calc && cargo build --release)             # then point $AGENT_CALC at it
export AGENT_CALC=../agent-calc/target/release/agent-calc

python3 examples/synergy_market.py
python3 examples/synergy_market.py --market SPY --period 2y --fdr 0.01
python3 examples/synergy_market.py --tickers SPY,NVDA,TSLA --top 12

anomalyx finds the anomalous days and the price regime shift (point.modz / mv.mahalanobis / coll.cusum); the detector output then feeds agent-calc, which computes the exact return distribution (describe_sample — note the fat-tailed kurtosis), the worst day's tail probability under a fitted Gaussian (normal_cdf — often "1-in-millions", i.e. the model is what's broken), a two-sample t-test on the returns either side of the CUSUM break (two_sample_t — is the regime shift a real change in mean return, or only in trajectory?), and exact Pearson r of each basket name to the market. The punchline is that both halves emit machine-readable contracts, so findings flow into the math kernel with no prose and no float drift.

journal_anomalies.py

Finds anomalies in the systemd journal (Linux + systemd). Pipes journalctl -o json to anomalyx on stdin (so it content-sniffs as journal, not plain JSON) and maps each finding back to its timestamp / unit / message.

python3 examples/journal_anomalies.py --lines 20000
python3 examples/journal_anomalies.py --since "2 hours ago" --top 20

# Distributional drift between two windows (which units / priorities shifted):
python3 examples/journal_anomalies.py --since "1 hour ago" \
        --baseline-since "3 hours ago" --baseline-until "1 hour ago"

Single-window finds per-unit content anomalies (e.g. CPU‑usage spikes); the --baseline-since mode runs dist.chi2 over _SYSTEMD_UNIT / PRIORITY to flag units that appeared or whose share changed. Column roles keep journald's many id / counter / timestamp fields out of the way automatically.

polymarket_anomalies.py

Pulls a prediction market's price history from Polymarket's public APIs (read-only, no key), enriches it with the per‑step probability change, and finds the information shocks — sharp probability jumps (point / mv) and sustained regime shifts in the odds (coll.cusum).

python3 examples/polymarket_anomalies.py                 # top market by volume
python3 examples/polymarket_anomalies.py "bitcoin"       # first match by question/slug
python3 examples/polymarket_anomalies.py "fed" --top 15  # search first, then scan flags

Pass any search term before scan flags (the term is an optional positional).

Maps each finding back to its UTC timestamp; the timestamp column is auto-classified a sequence and skipped, so the findings are about the odds, not the clock.