JPI to FlySto CSV Converter

Converts J.P. Instruments (JPI) EDM engine-monitor .JPI files into CSVs compatible with FlySto.net.

Tested against several EDM-830 downloads (one clean fixture of 16 flights / 26k+ samples, plus four real-world files exercising the firmware's repeat-compression and malformed-record edge cases).

Install

python3 -m venv .venv
.venv/bin/pip install -e ".[dev]"

(Python 3.9 or newer.)

Use it

Web UI (recommended for non-technical users)

.venv/bin/flask --app jpiconverter.web.app run

Then open http://127.0.0.1:5000, drop in a .JPI file, and download the CSV.

Command line

# single file
.venv/bin/jpi2csv path/to/file.JPI

# explicit output path
.venv/bin/jpi2csv path/to/file.JPI -o out.csv

# batch — convert every .JPI in a folder
.venv/bin/jpi2csv path/to/folder/

Tests

.venv/bin/pytest

What you get

One CSV per uploaded .JPI, with all flights concatenated. The CSV uses the FlySto V1 template (39 columns, CRLF line endings):

• INDEX is monotonic across all flights (no per-flight reset).
• Row 2 holds the tach summary across the entire file: Engine - Tach Start = ..., Tach End = ..., Tach Duration = ... (duration is the sum of per-flight engine-on time, not the wall-clock span between flights).
• GPS (LAT/LNG/ALT/SPD) is decoded when the EDM had a GPS feed at flight time — typically 95–97% of rows on flights with a GPS lock at engine start. Rows without a lock have those cells blank.
• Optional columns the source does not log (T1, FF2, FP, AMP, RFL, LFL) are left blank, except USD2 which emits the literal NA to match the template.

Known limitations

Partial decode on malformed records. Newer EDM-830 firmware sometimes emits records the parser can't fully interpret (repeat-compression sentinels and decodeflags-mismatch records). The converter recovers as much of the affected flight as possible, logs a warning, and moves on. In practice this affects roughly one flight in fifteen, and only truncates the tail of that single flight.

Deploying

For a hosted (multi-user) version:

gunicorn jpiconverter.web.app:app

…on any Python-friendly host (Render, Fly.io, Railway, PythonAnywhere). This repository also ships with a vercel.json and api/index.py shim so it deploys to Vercel as a serverless Python function with no extra configuration.

Attribution

• 2sec/python-edm — the binary parser this project's decoder is adapted from.
• openenginedata.org — public browser-side converter whose source documented the GPS field-index encoding used here.
• wannamak/edmtools — JPI metric reference (Apache-2.0).

License

MIT — see LICENSE.

Layout

src/jpiconverter/
├── parser/       # vendored from 2sec/python-edm + attribution
├── decode.py     # adapter: .JPI bytes -> list[Flight]
├── geo.py        # decimal degrees -> hemisphere-DMS string
├── flysto.py     # list[Flight] -> FlySto V1 CSV bytes
├── cli.py        # `jpi2csv` CLI
└── web/          # Flask app + upload page

tests/
├── fixtures/U260220.JPI   # real EDM-830 download
├── test_decode.py
├── test_geo.py
└── test_flysto.py