A lightweight, self-hostable web database for GRB afterglows and other transients.
Status: active development — not a finished product. AJST is a personal project, built and maintained by one person on a personal computer through vibe coding powered by the Kimi-K3 model. Development time and model quota are limited, so expect rough edges and gradual progress.
AJST was inspired by a number of earlier statistical/catalog works in the transient community:
Its core goals:
- Open — code (this repo, MIT) and data (the separate AJST-Data repo, CC BY 4.0) are both public.
- Lightweight — a Flask + PostgreSQL backend and a build-free vanilla-JS frontend; no Docker, no Node toolchain, no external CDN at runtime.
- Trustworthy data — data lives in its own repository with documented provenance per source and a stated quality-review plan (see below).
- Deployable on a personal computer — clone,
pip install, point at a local PostgreSQL, done. - Useful tools included — see the feature list below.
The data is not in this repository. It lives in the separate AJST-Data repository. Importing it is entirely optional: AJST runs fine with an empty database, and you may populate it with your own data instead.
The current dataset was batch-collected from publicly available papers and GCN circulars, and also inherits data from a research project:
The data has NOT yet been manually reviewed entry by entry. The author plans to spend roughly one year auditing the quality of every entry. Until then, do not use this data directly for serious scientific research.
Web UI (all local, no build step):
- Home / overview — entry point with quick statistics
- Transient list — filtering, export, one-click Galactic-extinction correction
- Detail page — light-curve plotting, data table editing, tags, aliases
- Compare — overlay light curves of multiple transients
- Digitizer — extract data points from published figure screenshots
(calibration → manual/color-based point picking → CSV export or direct
database write). The algorithm is inspired by the MIT-licensed
graph-digitizerproject; it is an original reimplementation, written as an alternative to the AGPL-licensed WebPlotDigitizer. - Filters management — photometric filter definitions (admin-editable)
- Afterglow fitting — submit and monitor
VegasAfterglow MCMC fits
from the detail page (optional dependency); a preset-combination engine
(
vegas_unified) with curated priors, selectable jet structure / circumburst medium / host extinction, and joint physical constraints (e.g. FS+RS pair plus a standalone FS, two-component jet) - Host galaxies — per-source host coordinates, redshift (spectroscopic or photometric) and multi-band photometry (AB/Vega/ST mag systems), plus a built-in pcigale SED-fitting tab: fixed-z or photometric-z runs with selectable bands, results table and best-model SED plot, and one-click write-back of the adopted parameters
- GCN tool — browse GCN circulars with per-source info cards and photometry entry
- Light-curve upload — batch CSV import with column mapping
- Statistical relations — Amati / Yonetoku / Ghirlanda / lag–luminosity / variability–luminosity / Ep–α
- Statistics — overview, redshift distribution, band coverage, host-galaxy coverage and M*/SFR distributions
- Admin panel — user management (
/admin)
APIs and integrations:
- STDweb ingest API —
POST /api/ingest/photometry(Bearer-token authenticated) is designed for seamless connection with the STDpipe/STDweb photometry pipeline; an AI agent can assist with the deployment and wiring on the STDweb side. - REST API for transients, light curves, filters, tags, statistics, export,
extinction, relations, fitting, spectra, GCN, and admin (see
docs/TECHNICAL.md).
- Galactic extinction — computed with
dustmaps using the CSFD dust map
(Liu et al. 2023) and the P92 extinction law (Pei 1992, Rv = 3.1) via
dust_extinction. Requires a
one-time
dustmaps.csfd.fetch()download. - Filter definitions —
filters.json; the Spitzer/IRAC entries are taken from the SVO Filter Profile Service; filter transmission curves for the host-galaxy SED fits are also retrieved from SVO FPS. - Host-galaxy SED fitting — pcigale (Boquien et al. 2019; v2025.0
tested), run as a subprocess with results rendered by
matplotlib. - All-sky map — embedded Aladin Lite 3.8.2 (CDS).
- Afterglow fitting — VegasAfterglow (v2.0.6 tested), with
cornerandmatplotlibfor posterior plots. - Spectral line markers — the line lists on the spectra page are taken verbatim from the object pages of the Transient Name Server (wis-tns.org).
- Digitizer — algorithm inspired by the MIT-licensed
graph-digitizer; an original reimplementation as a self-hosted alternative to WebPlotDigitizer (AGPL). - Frontend vendor libraries (bundled locally, no CDN): Bootstrap 5.3.3, bootstrap-icons 1.11.3, Chart.js 4.4.7 (all MIT).
- Data sources — see the per-directory READMEs in AJST-Data; if you use a subset of the data, please cite the original works listed there.
- A considerable number of features are not yet developed or integrated.
- Some existing features are relatively rough and may be removed in future versions.
- For some planned features the author has not yet formed a mature design; they will be refined step by step.
- AJST is deliberately lightweight: it aims to cover the commonly needed functionality, but it has an explicit scope boundary and will not grow without limit.
Community participation is very welcome:
- Open an issue for suggestions, bug reports, or data-quality notes.
- Open a pull request if you'd like to co-develop.
Requirements: Python ≥ 3.10, PostgreSQL ≥ 14.
git clone https://github.com/Cooper-J2000/AJST.git
cd AJST
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
# optional, only for afterglow fitting:
# pip install "VegasAfterglow[mcmc]" corner matplotlib
# create the database (as a PostgreSQL superuser or with createdb rights)
createdb ajst_catalog
# set the admin password (REQUIRED — otherwise a random one is generated
# at every startup and printed nowhere)
export AJST_CATALOG_PASSWORD='choose-a-strong-password'
# optional: pull the data repository into ./catadata and import it
git clone https://github.com/Cooper-J2000/AJST-Data.git catadata
cd backend && python3 etl.py && cd ..
# run
./backend/start.sh # listens on 0.0.0.0:5000Without catadata/, the app still starts with an empty database; you can add
your own transients through the UI or the APIs.
| Variable | Default | Purpose |
|---|---|---|
DATABASE_URL |
postgresql+psycopg2:///ajst_catalog |
PostgreSQL DSN |
AJST_CATALOG_PASSWORD |
random per startup | initial admin password |
AJST_INGEST_TOKEN |
unset (ingest API disabled) | Bearer token for /api/ingest/* |
AJST_DATA_DIR |
<repo>/catadata |
data directory location |
AJST_PYTHON |
python3 |
interpreter used by start.sh |
PORT |
5000 |
listen port |
The GCN tool can download/update the circular archive itself, or you can fetch the full archive manually:
bash scripts/fetch_gcn_archive.shbackend/ Flask app, SQLAlchemy models, ETL, fitting engines, routes
frontend/ build-free SPA (vanilla JS ESM) + bundled vendor libraries
scripts/ maintenance scripts (GCN archive fetch, ...)
docs/ TECHNICAL.md — full technical documentation (Chinese)
catadata/ (git-ignored) clone of AJST-Data, or your own data
Code: MIT. Data: CC BY 4.0 in the separate
AJST-Data repository; subsets
under external/ remain subject to the terms and citation requirements of
their original sources.