Quickstart¶
Get from zero to a validated Telemachus dataset in a few minutes.
Install¶
Python 3.10+ :
This ships the library (import telemachus), the tele CLI, and every
adapter (AEGIS, PVS, STRIDE). Schema validation is built-in — no ajv
or external tool needed.
Try it — 5-minute demo¶
The easiest way to see Telemachus in action is the AEGIS demo notebook. It downloads a real open dataset from Zenodo, loads it, and plots one trip. Runnable in Colab in one click.
Read a dataset¶
import telemachus as tele
df = tele.read("path/to/manifest.yaml") # or directly to a .parquet
print(f"{len(df):,} rows, {df['trip_id'].nunique()} trips, "
f"profile = {tele.sensor_profile(df)}")
tele.read() returns a regular pandas DataFrame with SI units and
UTC-aware timestamps — no wrapper, no hidden state.
Validate a dataset¶
Or from Python:
report = tele.validate(df, profile="full") # core | imu | full
print(report)
# ValidationReport(PASS, profile=full, level=basic, errors=0, warnings=0)
Convert an open dataset¶
tele convert aegis /path/to/aegis/csvs -o datasets/aegis/
tele convert stride /path/to/stride/road_data -o datasets/stride/ --category driving
tele convert pvs /path/to/pvs/trips -o datasets/pvs/ --placement dashboard
Next steps¶
- Validating files — strict and lenient modes
- Reading Telemachus data — Python, DuckDB, pandas
- Writing an adapter — convert vendor X → Telemachus
- Manifest FAQ — what SPEC-02 changes
- Concepts — record format and profiles