Tessera

Tessera is an open-source, telescope-agnostic engine for gravitational-wave follow-up. It turns an LVK HEALPix sky map into ranked host galaxies and feasible, validated observing schedules for your telescopes. Every result comes with a reproducibility manifest.

GW170817 planned with the public GWTC-1 sky map, a GLADE+ excerpt, and an example 1 m telescope profile. NGC 4993 ranks first, and its field is observed on the sixth visit, 10.7 h after merger.

What Tessera does

1. Ranks host galaxies transparently. Every score component is shown: sky density p_sky, the distance term p_dist, the astrophysical weight w, the credible level, the distance z-score and data-quality flags. S is a ranking score, not a host probability. The opt-in host-posterior strategy gives normalised probabilities under a stated generative model.
2. Reports what the catalog cannot explain. The probability budget shows how much is cataloged (P_cat), how much is uncataloged, how much lies outside the footprint, and how much lies beyond reliable depth.
3. Builds schedules that can actually be executed. Hard constraints come from astroplan: twilight, airmass, horizon mask, Moon, hour-angle limits and operational windows. Slews and filter changes are modelled per axis. A deterministic greedy optimiser records why each visit was chosen, and an independent validator rechecks every visit at 10 s resolution. Infeasible schedules cannot be exported.
4. Makes runs reproducible. Each run records a plan_key, input checksums, the catalog snapshot ID, profile hashes, the full config, and software and IERS versions. tessera rerun and tessera compare diff any two runs.

Screenshots

Start a plan: upload a sky map (or give a GraceDB URL), pick a catalog, preset and telescopes.

REST API: versioned /v1 endpoints with OpenAPI docs.

Ranked host galaxies, with every score component and flag (downloadable as CSV, ECSV, JSON, Parquet or VOTable):

A validated schedule, with the decision behind each visit and a per-visit feasibility check. The schedule stays proposed until a person approves one exact version. It exports to generic JSON/CSV, ACP, LCO, Treasure Map, or a Rubin ToO package:

Quick start

git clone https://github.com/bibinthomas123/Tessera.git && cd Tessera
python -m venv .venv && .venv/bin/pip install -e ".[dev]"

# 1. build a versioned catalog snapshot (here: the GLADE+ excerpt shipped for tests)
.venv/bin/tessera ingest-catalog glade+ tests/data/glade_plus_gw170817.tsv \
    --out catalogs/glade-excerpt --upstream-version 2022-10-31

# 2. plan GW170817 for an example 1 m telescope
.venv/bin/tessera plan --skymap tests/data/GW170817_skymap.fits.gz --catalog catalogs/glade-excerpt \
    --profile example-1m --preset bns --nights 1 --t-start 2017-08-17T22:00:00 --out out/gw170817

# 3. re-check feasibility independently, and export for an ACP-controlled telescope
.venv/bin/tessera validate-schedule "out/gw170817/schedules/example-1m@1.json"
.venv/bin/tessera export "out/gw170817/schedules/example-1m@1.json" --adapter acp

As a library, inside an existing pipeline:

from tessera import plan, PlanConfig

r = plan("bayestar.multiorder.fits", "catalogs/gladeplus-2022.10", ["example-1m"],
         PlanConfig(preset="bns"), t_start="2017-08-17T22:00:00")
r.galaxies[:10]            # ranked galaxies (astropy Table)
r.budget["p_cat"]          # share of probability the catalog can explain
r.schedules["example-1m@1"]["visits"]
r.write("out/")            # all outputs + manifest.json

Running the service

cp .env.example .env      # then set POSTGRES_PASSWORD (e.g. openssl rand -hex 24)
docker compose up -d      # api, worker, alert listener, PostgreSQL
docker compose exec api tessera keys create --name ops --scopes read,plan,approve

The UI is at http://localhost:8000/ui/ and the API docs at /v1/docs. For a multi-user deployment behind an institutional gateway, there is a Helm chart in deploy/helm/tessera, and the service supports OIDC and gateway authentication; see docs/deployment.md. Live LVK alerts arrive from GCN Kafka, SCiMMA or GraceDB. New sky maps are treated as revisions of the same event, and retractions withdraw schedules automatically.

Validation

tessera benchmark regenerates docs/benchmark-report.md:

There are 100 tests, covering service integration, alert revision handling and a GW170817 golden regression. CI runs them on Python 3.10, 3.11 and 3.12 and builds the Docker image.

Installing

The PyPI distribution name is tessera-gw, because the name tessera on PyPI belongs to an unrelated project. The Python package and the command are both tessera. Optional extras: [service], [postgres], [alerts], [rubin], [validation].

Data and caveats

• Test data is public. It includes the GWTC-1 GW170817 sky map (LIGO DCC P1800381), the GraceDB BAYESTAR maps for S190425z and S200105ae, and a GLADE+ excerpt from VizieR VII/291 (Dalya et al. 2022; please cite it). The AT2017gfo light curve is derived from Villar et al. 2017, Table 3.
• Telescope profiles shipped here are illustrative or approximate, not facility-validated.
• Science model 1.2.0. The light curve, all reference densities (B-band, K-band, and stellar mass on the same GAMA/Chabrier scale as GLADE+ masses) and the kilonova diversity σ_M = 1.0 mag are sourced from the literature; see docs/science.md.
• Adapters.
  • The Rubin adapter is checked against rubin-scheduler.
  • LCO is checked against the open-source Observation Portal and a dated snapshot of LCO's public instrument list (instrument types, telescope classes, filters, modes).
  • Treasure Map is checked against its official client.
  • ACP is unverified.
  • Each facility must confirm the format before submission is enabled. See docs/adapters.md.

Further reading: science model · API · validation protocol · contributing · roadmap. Licensed under BSD-3-Clause.