SGLSETI

SGLSETI is an open-source Python package that generates reproducible sky targets for searches for solar gravitational lens (SGL) technosignatures. It turns a curated stellar-system hypothesis, an observer, and past or future epochs into telescope-usable SGL target regions — with versioned geometry models, explicit uncertainty, and content-hashed provenance on every row.

One stateless, epoch-neutral engine supports the core workflows:

• Archival target generation — calculate where an SGL relay hypothesis would have appeared at historical epochs (instantaneous, or as first-class observation intervals with start/mid/stop samples) so the positions can be cross-referenced with an archive by an external tool.
• Commensal target generation — calculate current or future positions, visibility, and simple circular-field pointings for use alongside another observing program.
• Beam-crossing search — scan past or future time intervals for the moments an observer passes closest to a target's hypothesized relay beam axis (inbound uplink or outbound downlink), reporting impact-parameter events and assumed-beam-radius windows (sglseti crossings, ADR-0003) — plus cheap point/interval queries for joining a known schedule.
• Archival survey toolkit (Python API) — the continuous locus evaluator and tolerance-guaranteed adaptive sampler, swept envelopes over observation intervals, covered relay-distance refinement against a caller-supplied footprint test, seeded Monte Carlo uncertainty propagation (sky regions, crossing distributions, side-of-axis stability), and deterministic chunked execution with bounded-memory streaming export for archive-scale runs.

Modeled physics, explicitly versioned

• Geometry: the reviewed tusay2022_eq5_7_v1 role model (Tusay et al. 2022 eq. 5–7) and the sun_star_axis_v1 crossing-axis contract, with every approximation declared per row and quantified in the accuracy budget.
• Target motion: pluggable provider families — linear astrometry, catalog acceleration solutions, two-body orbits for resolved binaries and barycenters (validated against the published Alpha Cen AB and Sirius AB solutions), and externally generated checksummed ephemerides (the only family that models planet endpoints).
• Observers: Earth center, terrestrial sites, Solar-System bodies from the pinned planetary ephemeris, spacecraft from checksummed tabular ephemerides or SPICE kernels, and programmatic observers for testing.

Product boundary

SGLSETI ends at the generation and export of target products (ECSV, JSON, CSV, VOTable, DS9 regions, and a provenance manifest). It does not query archives, ingest observations, track historical search coverage, manage candidates, or schedule telescopes. Those belong to external systems that consume SGLSETI output.

Status

v1.1, survey-ready: the full survey-driven roadmap (notes/improvements.md) is implemented and verified — 480+ offline tests, including frozen fixtures for two published SGL searches and microarcsecond-level cross-validation against JPL Horizons. Identities (calculation IDs, content hashes) are at their frozen greenfield baseline for the first archival consumers. Remaining open items are deliberately gated or demand-driven (see the note).

Documentation

• Quick start — all workflows, CLI and Python
• Target registry — schema, provider families, provenance and covariance blocks
• Conventions and product reference — frames, time scales, role epochs, every output column
• Accuracy budget — measured floors and declared approximations by model, observer type, epoch span, and provider
• Resources — pinned ephemeris kernels and IERS, offline
• Limitations — model validity and boundaries
• Benchmarks — archive-scale performance measurements
• Science specification and ADRs — the reviewed model contracts

Installation

Requires Python 3.11+. With uv:

uv sync          # create the virtual environment and install dependencies
uv run sglseti --help

Or with pip:

python -m pip install -e .
sglseti --help

Usage

sglseti generate \
  --targets examples/targets.yaml \
  --request examples/historical.yaml \
  --output-dir build/archive-targets

or from Python:

from sglseti import generate_loci, load_request, load_target_registry, write_products

registry = load_target_registry("examples/targets.yaml")
result = generate_loci(load_request("examples/historical.yaml"), registry)
write_products(result, "build/archive-targets", generated_utc="2026-08-18T00:00:00+00:00")

Every output row keeps its caller-supplied epoch_id join key, its coordinate frame/origin/epoch semantics, model identity, target- and observer-provider identity, resource checksums, and validity status; a manifest.json separates the deterministic science identity from run circumstance. A taste of the survey-facing API:

from astropy.time import Time
from sglseti import (
    AstropyEphemeris,
    Observer,
    RelayRange,
    Role,
    Tusay2022Eq57V1,
    adaptive_locus,
    covered_z_intervals,
)

locus = adaptive_locus(  # polyline within 0.5" of the
    target=registry["barnard"],  # continuous locus, guaranteed
    role=Role.RX,
    observation_time=Time("2015-09-05T07:41:00", scale="utc"),
    observer=Observer.earth_center(),
    relay_range=RelayRange(550.0, 2500.0),
    tolerance_arcsec=0.5,
    ephemeris=AstropyEphemeris(),
    model=Tusay2022Eq57V1(),
)
# ...then refine your instrument footprint into covered z intervals with
# covered_z_intervals(contains=your_footprint_test, ...)

See the quick start for observation intervals, uncertainty propagation, crossings, and chunked archive-scale execution.

Ephemeris resources

Calculations never touch the network. Astropy's built-in analytic ephemeris works out of the box (its relay-pointing floor is ≤ 0.02 mas — see the accuracy budget); citable products should pin the canonical kernel JPL DE440s (ADR-0002), fetched once with the only network-using command:

sglseti fetch ephemeris de440s --output-dir resources/
# prints the verified sha256 and the request snippet:
#   ephemeris:
#     adapter: jpl_file
#     path: resources/de440s.bsp

sglseti fetch iers --output-dir resources/ optionally downloads the IERS-A Earth-orientation table as a pinned, checksum-identified file for a request's iers block (sub-arcsecond effect on apparent/site products; never required — without it, astropy's bundled tables are used and identified in provenance). The jpl_file ephemeris adapter needs the jpl extra (uv sync --extra jpl or pip install 'sglseti[jpl]'); the SPICE spacecraft observer needs the spice extra (spiceypy).

Development

uv sync                 # installs the dev dependency group
uv run pytest           # unit / integration / regression tests (offline)
uv run ruff check .
uv run mypy

Default tests never require network access. See CONTRIBUTING.md.

License

MIT — see LICENSE.