Chenin

Turn Génie 2000 (G2K) gamma-spectrometry reports into structured data and an activity-vs-depth synthesis for a sediment core, ready for age-depth modelling. Built for the EDYTEM-PTAL research group (CNRS).

Chenin does three things:

1. Extract — parse a raw G2K .txt report into six clean tables (peaks, nuclide activities, detection limits, …).
2. Synthesise — combine several reports into one table, one row per sample, with the nuclide activities and uncertainties you care about, at their depth in the core.
3. Visualise & export — browse everything in a web app and export to CSV, Parquet or SERAC.

You point Chenin at a folder of reports and describe the core layers — which report is which slice of sediment, and at what depth. Cores are usually cut at a constant step with the slice number in the file name, so Chenin can generate the depths for you and let you correct them. Which nuclides land in the table comes from a synthesis template, and a sensible lab default ships with Chenin — so in the common case you write nothing at all.

Requirements

• Python ≥ 3.14
• uv (package/environment manager)

Install

As a tool (for end users)

uv tool install git+https://github.com/Ezacuak/chenin.git

This puts a chenin command on your PATH. Update later with uv tool upgrade chenin.

From a clone (for development)

git clone https://github.com/Ezacuak/chenin.git
cd chenin
uv sync

Then prefix commands with uv run (e.g. uv run chenin ...).

Quick start

# 1. Inspect a single report
chenin extract data/NOIR24-01/NOI_S_1.txt          # all sections to stdout
chenin extract data/NOIR24-01/NOI_S_1.txt -s s3    # just section 3
chenin extract data/NOIR24-01/NOI_S_1.txt -o out/  # each section to a CSV

# 2. Build a synthesis from a folder of reports (uses the packaged default template)
chenin synthesis data/NOIR24-01/ --thickness 0.5                    # print the table
chenin synthesis data/NOIR24-01/ --thickness 0.5 -o synthesis.csv   # export it
chenin synthesis data/NOIR24-01/ --template my_template.csv         # custom columns

# 3. Launch the web app (recommended for day-to-day use)
chenin app

chenin app opens the Streamlit interface in your browser: upload the reports on the Load Data page, describe the core layers on the Synthesis page, then browse the extracted sections and the synthesis, with a sediment-core view and per-nuclide depth profiles.

The bare form chenin report.txt still works as a shortcut for chenin extract report.txt.

Core layers

A G2K report says what a sample contains, never where it came from. The layer geometry — depth top, depth bottom, dry bulk density — has to come from you.

Because cores are cut at a constant step and the slice number is usually written into the file name, Chenin generates the depths from a slicing rule:

chenin synthesis data/NOIR24-01/ --start 0 --thickness 0.5

By default the slice number is read from the end of the file name (NOI_S_13 → 13), so a core with unmeasured slices keeps its true depths — if slice 12 was never counted, slice 13 still sits at 6.0 cm instead of sliding up to fill the hole. Pass --sequential to stack the reports contiguously instead. In the app the same rule prefills the Core layers table, where every cell stays editable for cores that are not regular.

The synthesis template

Which nuclides land in the synthesis comes from a synthesis template: a compact table whose header row names the output columns and whose single method row says how to obtain each one — gamma peaks (NUCLIDE@energy, ;-separated for a weighted mean) or an = formula over other columns:

PB-210,       RA-226,                                          PB-Exc,             K-40
PB-210@46.54, PB-214@295.21; PB-214@351.92; BI-214@609.31,     =[PB-210] - [RA-226], K-40@1460.82

The standard EDYTEM-PTAL template ships with Chenin and is applied automatically. Pass --template my_template.csv (or edit and download one in the app) to override it.

Each measured column yields an Activite …/Incertitude … pair, prefixed by the layer geometry: Echantillon, Profondeur, Epaisseur, DBD.

Documentation

Full documentation is in docs/. The user guide is also rendered inside the app, under Documentation in the sidebar — the markdown files are the single source for both.

Start here	For
Overview	What Chenin does, the workflow, the vocabulary
Install & Update	Getting it running
Formulas	Derived columns and uncertainty propagation
Command Line	Every subcommand and flag
Troubleshooting	When something goes wrong
Architecture	Contributors: package layout and data flows

Project layout

src/chenin/
├── g2k_parser/     # parsing library: G2K report -> {section: DataFrame}
├── synthesis/      # build model, slicing rule, report loading, synthesis builder
├── ui/             # Streamlit app (pages, components, shared state)
└── cli.py          # `chenin` command (extract / synthesis / app)
docs/               # user guide (shipped + rendered in-app) and developer guide
tests/              # pytest suite

Development

uv sync                     # install, including dev dependencies
uv run pytest               # run the test suite
uvx ruff check src/ tests/  # lint
uv run chenin app           # the app, against your working tree

See Contributing for the uv workflow, the documentation house style and the conventions worth not breaking, and Parsing G2K reports for the report format itself.