·                               
██    ·           ██            ██            ██                                            ██      
██▒░              ██▒░          ██▒░  ·       ██▒░                               ✧          ██▒░    
██▒░               ▒▒░          ██▒░          ██▒░        ████████          ██████           ▒▒░    
██▒░             ·          ·   ██▒░          ██▒░        ████████▒░        ██████▒░                
████████ ·        ██        ██████████    ██████████      ██▒▒▒▒██▒░         ▒▒▒▒▒██        ██      
████████▒░        ██▒░      ██████████▒░  ██████████▒░    ██▒░  ██▒░              ██▒░      ██▒░    
██      ██        ██▒░       ▒▒▒██▒▒▒▒▒░   ▒▒▒██▒▒▒▒▒░·   ██▒░   ▒▒░        ████████▒░      ██▒░    
██      ██▒░      ██▒░          ██▒░          ██▒░        ██▒░              ████████▒░      ██▒░    
██      ██▒░      ██▒░          ██▒░          ██▒░ ·      ██▒░            ██      ██▒░      ██▒░    
██      ██▒░      ██▒░          ██▒░          ██▒░        ██▒░            ██      ██▒░      ██▒░    
████████ ▒▒░      ██▒░          ████          ████        ██▒░      ██     ▒████████▒░      ██▒░    
████████▒░        ██▒░          ████▒░        ████▒░·     ██▒░      ██▒░    ████████▒░      ██▒░    
 ▒▒▒▒▒▒▒▒░         ▒▒░           ▒▒▒▒░         ▒▒▒▒░       ▒▒░       ▒▒░    ·▒▒▒▒▒▒▒▒░       ▒▒░    
                                                                                                    
     ·      ·                                                                                       
                                                                                                    
                                              bittr.ai                                              

tess-vetter

Domain-first Python library for TESS transit detection + vetting (array-in/array-out).

This package follows a "domain-first" design: array-in/array-out astronomy algorithms with optional platform helpers. The core api/, compute/, and validation/ modules are pure functions; the platform/ module provides opt-in I/O and network clients when needed.

Package structure:

• Pure domain logic (no I/O, no network): api/, compute/, validation/, transit/, recovery/, activity/
• Opt-in infrastructure (network clients, caching, disk I/O): platform/

The platform/ module is entirely optional and only used when explicitly imported.

What's in here

• Vetting pipeline: default preset runs 15 checks (V01-V12, V13, V15, V11b); opt-in extended preset adds metrics-only diagnostics (V16-V21), for 21 total checks available via profiles (tess_vetter.api.vet)
• Transit detection: TLS/LS periodograms, multi-planet search, candidate merging (tess_vetter.api.periodogram)
• Pixel diagnostics: centroid shift, difference images, WCS-aware localization, aperture dependence (tess_vetter.api.pixel)
• Transit recovery: detrend + stack + trapezoid fitting for active stars (tess_vetter.api.recovery)
• FPP (optional): TRICERATOPS+ support with a vendored copy under src/tess_vetter/ext/ (tess_vetter.api.fpp)
• Code mode (experimental): operation-catalog and MCP tooling under tess_vetter.code_mode; this surface is experimental and may change
• Citations: many public API entry points carry machine-readable literature references (see REFERENCES.md and tess_vetter.api.references)

Installation

Requires Python 3.11–3.12.

Minimal install (basic vetting)

pip install tess-vetter

CLI entrypoints (equivalent):

• btv ...
• tess-vetter ...
• python -m tess_vetter ...

Preflight your environment before demos/runs:

btv doctor --profile vet

With transit detection (TLS)

pip install 'tess-vetter[tls]'

With MCMC fitting

pip install 'tess-vetter[fit]'

With physical transit model fitting (batman)

pip install 'tess-vetter[batman]'

With advanced detrending (wotan)

pip install 'tess-vetter[wotan]'

With limb darkening coefficients (ldtk)

pip install 'tess-vetter[ldtk]'

With external vetter integration (exovetter)

pip install 'tess-vetter[exovetter]'

With false positive probability (TRICERATOPS)

pip install 'tess-vetter[triceratops]'

With Apple Silicon acceleration (MLX, macOS only)

pip install 'tess-vetter[mlx]'

Full install

pip install 'tess-vetter[all]'

Development

Using uv (recommended for this repo; uses uv.lock):

uv sync --all-extras --group dev

Using pip:

python -m pip install -e ".[all]"

License note for optional extras: The [triceratops] extra includes pytransit, which is GPL-2.0 licensed. Installing this extra changes the effective license of your environment. The [ldtk] extra is also GPL-2.0. The core package remains BSD-3-Clause.

Quickstart

The recommended import alias follows patterns from astropy (import astropy.units as u):

import tess_vetter.api as btv

Agent-first walkthroughs and CLI-first examples are in docs/quickstart.rst.

Single candidate vetting

import numpy as np
from tess_vetter.api import Candidate, Ephemeris, LightCurve, vet_candidate

lc = LightCurve(time=time, flux=flux, flux_err=flux_err)
candidate = Candidate(
    ephemeris=Ephemeris(period_days=3.5, t0_btjd=1850.0, duration_hours=2.5),
    depth_ppm=500,
)

bundle = vet_candidate(lc, candidate, network=False)

# Results are structured: status, metrics, flags, citations
for r in bundle.results:
    print(f"{r.id} ({r.name}): {r.status}")
    print(f"  metrics: {r.metrics}")
    print(f"  flags: {r.flags}")
    print(f"  citations: {r.citations}")

Batch vetting (multiple candidates, one light curve)

from tess_vetter.api import vet_many

candidates = [
    Candidate(ephemeris=Ephemeris(period_days=3.5, t0_btjd=1850.0, duration_hours=2.5)),
    Candidate(ephemeris=Ephemeris(period_days=7.0, t0_btjd=1852.0, duration_hours=3.0)),
]

bundles, summary = vet_many(lc, candidates, network=False)

# summary is a list of dicts with stable columns for sorting/filtering
for row in summary:
    print(f"P={row['period_days']:.2f}d: {row['n_ok']} ok, {row['n_skipped']} skipped")

Pipeline introspection

from tess_vetter.api import list_checks, describe_checks

# List all registered checks with their requirements
checks = list_checks()
for c in checks:
    req = c["requirements"]
    print(f"{c['id']}: {c['name']} (tier={c['tier']}, needs_network={req['needs_network']})")

# Human-readable summary
print(describe_checks())

Running a subset of checks

from tess_vetter.api import vet_candidate

# Run only LC-only checks (V01-V05)
bundle = vet_candidate(lc, candidate, checks=["V01", "V02", "V03", "V04", "V05"])

Network behavior

Catalog-backed checks are always opt-in. You must pass network=True (and provide metadata like RA/Dec and TIC ID) to enable external queries; otherwise those checks return skipped results.

CLI FPP workflow (agent-safe)

For repeatable FPP runs, use a two-step flow:

1. btv fpp-prepare to stage light curves/runtime artifacts and emit a manifest.
2. btv fpp --prepare-manifest ... (or btv fpp-run --prepare-manifest ...) to compute from the staged manifest.

Example:

btv fpp-prepare --toi "TOI-5807.01" --network-ok --cache-dir outputs/cache -o outputs/fpp/toi_5807.prepare.json
btv fpp --prepare-manifest outputs/fpp/toi_5807.prepare.json --no-network -o outputs/fpp/toi_5807.fast.json

Notes:

• --prepare-manifest on btv fpp uses the same prepared-manifest compute path as btv fpp-run.
• prepared-manifest mode now fails fast on missing staged artifacts by default (--require-prepared).
• use --allow-missing-prepared only for debugging/recovery workflows.
• In prepared mode, do not mix direct candidate/staging flags (for example --tic-id, --period-days, --cache-dir) with --prepare-manifest.

Citations

Many public API entry points and vetting checks include a list of literature references in their results. The full reference list lives in REFERENCES.md (and is also available programmatically via tess_vetter.api.references).

Code map

• src/tess_vetter/api/: stable host-facing facade (recommended import surface)
• src/tess_vetter/compute/: core array algorithms (detrending, periodograms, scoring)
• src/tess_vetter/validation/: check implementations and aggregation primitives
• src/tess_vetter/pixel/: pixel-level and WCS-aware diagnostics
• src/tess_vetter/recovery/, src/tess_vetter/transit/, src/tess_vetter/activity/: domain modules
• src/tess_vetter/platform/io/, src/tess_vetter/platform/catalogs/: optional I/O and catalog helpers
• src/tess_vetter/ext/triceratops_plus_vendor/: vendored TRICERATOPS+ (see THIRD_PARTY_NOTICES.md)

Platform Support

• macOS / Linux: First-class support. All features work as expected.
• Windows: Best-effort support. Some platform-specific features may have limitations:
  • Cache file locking uses fcntl (Unix-only); graceful fallback on Windows.
  • Network timeouts use SIGALRM which may not work on all platforms.

Security

Cache files (light curves, TRICERATOPS results) use pickle serialization for performance. Ensure cache directories have appropriate permissions in shared or multi-user environments.

Development

Run development commands in the project environment with uv run:

uv run pytest
uv run ruff check .
uv run mypy src/tess_vetter

Docs

User-facing docs are built with Sphinx from docs/:

uv run sphinx-build -b html -W docs docs/_build/html

For agent onboarding, start with docs/quickstart.rst. For API usage recipes and signatures, see docs/api.rst.

Internal working notes live in working_docs/ and are not part of the stable API.

License

BSD-3-Clause. See LICENSE for details.