Ψ PsiDataViz

Organize, parse, convert, and visualize experimental & computational scientific data — from any public link, no account or API key required.

---

Every lab accumulates a sprawl of instrument and computational data in dozens of formats, scattered across folders and cloud drives. PsiDataViz points at a public data location — a GitHub or Codeberg repo, a Google Drive folder, or a Box folder — scans it, makes sense of the formats, and lets you filter, overlay, compare, and convert datasets in your browser.

Hosting data for PsiDataViz? Use GitHub, Google Drive, Codeberg, or Box (all scan from a plain public link). Dropbox and Proton Drive aren't scannable — see docs/data-sources.

The name: sci and psi are homophones, and Ψ is a science-coded Greek letter — so the project, library, and app share the Psi prefix and the Ψ mark.

• psidata — a framework-agnostic, pip-installable Python library: the universal data model, a confidence-scored reader registry, source connectors, and format conversion. Works standalone in scripts, marimo, or Jupyter. “ΨData — Scientific Data.”
• PsiDataViz — the web app (FastAPI + React) on top of it. “ΨDataViz — Scientific Data Visualization.”

Highlights

• Point-and-scan, keyless. Public GitHub repos, Google Drive folders, Codeberg repos, and Box folders scan with no credentials. Files that share a base name across formats collapse into one dataset.
• Readers today: DSC (TA Trios), NMR (JCAMP-DX incl. ASDF & Nanalysis NMReady FID→spectrum, .tsv, 2D TopSpin totxt, zipped Bruker & SpinSolve datasets), FTIR (Bruker .dpt, Bruker OPUS .0, JCAMP, PerkinElmer .asc), Raman, XRD (1D .xy, PANalytical .csv/.xrdml/.udf, .dat/.asc; 2D detector images .edf/.img/.mccd/.tif/.h5 via FabIO, shown as heatmaps — and azimuthally integrated to a 1D pattern when the header carries a calibration), UV-Vis (.txt + Thorlabs .csv), TGA (TA Instruments .txt → weight % vs temperature), Brillouin (multichannel-scaler .asc), Acoustic (interferometry FFT Spectrum.csv inside the .zip), microscopy (SEM / TEM / optical .tif/.jpg/.png shown as the real micrograph, plus Gatan TEM .dm3/.dm4 via ncempy), electrochemistry (Gamry .DTA cyclic voltammetry + Chemotion JCAMP-DX .jdx → Potential × Current), mass spec & SIMS (ChemSpectra JCAMP-DX MASS SPECTRUM → m/z × abundance; standard MS and secondary-ion MS grouped separately), dielectric (broadband ε′/ε″ vs frequency on a log axis, per temperature), HPLC (chromatogram CSV + Agilent ChemStation DAD .D runs inside .tar.bz2, with an interactive wavelength slider over the time×wavelength matrix), circular dichroism (.dcs), mechanical / generic spreadsheets (.xlsx/.xls — finds the data table across sheets), and computed IR/Raman spectra from Gaussian/ORCA/Psi4 frequency jobs — .log/.out via cclib (Gaussian/ORCA/Q-Chem/NWChem/Psi4), or pre-extracted _ir.txt/_raman.txt — for overlay on experiment. (More formats are the active focus; see the roadmap.)
• QUICK tab — one source: scan → filter by technique → overlay → compare formats → convert/download.
• DATA tab — a multi-source workspace: add several public sources, then filter and overlay datasets across all of them, with an editable metadata panel (sample / instrument / conditions + tags, pre-filled from the parse) under each loaded dataset.
• 3D structure viewer — molecular & crystal structures (.xyz/.mol/.sdf/.pdb/.cif, plus Gaussian/ORCA input geometries .gjf/.inp) and a computational job's optimized geometry render in an interactive 3Dmol.js viewer beside the data, so a Gaussian/ORCA .log shows its IR/Raman spectra and its molecule — with animated vibrational normal modes (pick a frequency and watch the atoms move).
• Molecule viewer everywhere — type a SMILES or a compound name (resolved via RDKit + PubChem) to see its 3D structure, and the viewer auto-shows the compound when a dataset's name implies one (e.g. a CBD Raman spectrum displays cannabidiol). Mass spectra render as stick/peak plots.
• ΨDataSound — for acoustic .wav recordings, play the sound in the browser and toggle between the waveform (time vs amplitude) and its FFT spectrum.
• Convert any dataset to CSDM, HDF5, CSV, Parquet, Feather, or per-signal CSV (zip).
• Single-image deploy. One container serves the React UI and the API together.

Quickstart

Run with Docker (the whole app)

git clone https://github.com/jyarger/PsiDataViz.git
cd PsiDataViz
docker build -t psidataviz . && docker run --rm -p 8070:8000 psidataviz
# open http://localhost:8070

Develop locally

Prerequisites: uv (Python) and Node 20+.

# Python library + backend deps, incl. all reader extras (cclib, fabio, h5py, brukeropusreader, …)
uv sync --all-packages --all-extras

# Backend API on :8000 (PYTHONPATH form is the most portable across environments)
PYTHONPATH="packages/psidata/src:apps/backend/src" uv run python -m psidata_backend.main

# Frontend dev server on :5173 (in another terminal), proxying /api -> :8000
cd apps/frontend && npm install && npm run dev

Use the library on its own

from psidata import read, Candidate
ds = read(Candidate(filename="spectrum.dx", content=open("spectrum.dx", "rb").read()))
print(ds.technique, ds.signals[0].x.label, ds.signals[0].npoints)

Repository layout (uv workspace monorepo)

PsiDataViz/
├── packages/psidata/        # ★ the standalone, pip-installable science library
│   └── src/psidata/         #   model · registry · readers/ · sources/ · convert/ · export
├── apps/
│   ├── backend/             # FastAPI service (scan, records, dataset, catalog, convert, compare)
│   ├── frontend/            # React + TypeScript + Plotly UI
│   └── psidataviz-dash/     # interim Dash app (being retired)
├── docs/                    # ROADMAP, adding-a-reader, deploy
└── Dockerfile, docker-compose.yml, deploy/

Documentation

Full docs live in docs/ — a wiki-style index covering the roadmap, architecture, data sources, adding a reader, design decisions, and deployment.

Contributing

PsiDataViz is open source under Apache-2.0 — contributions, forks, and reuse are welcome. The most valuable contributions right now are new readers for scientific data formats we don't yet cover. See CONTRIBUTING.md and docs/adding-a-reader.md, and open a data-format request for any format that fails to parse.

License

Copyright © 2026 Jeffery L. Yarger (Yarger Lab). Licensed under the Apache License 2.0 — free to use, modify, and distribute, including commercially.