pgx
PGX — Lightweight Postgres for AI agents. Written in Rust. Thousands of isolated databases in one runtime. 🐘
Translation pending
An archive of the public Starlink ephemerides, hashed and anchored in Bitcoin, with a daily measure of what the public satellite catalogue can see.
DataPython · ShellGitHub @Kira-RyanOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Loading the full README. The saved excerpt is shown below.
Every eight hours, SpaceX publishes where each Starlink satellite is going, with its
uncertainty — the only public feed of its kind. Each set is replaced by the next, and no
continuous public archive of them exists. Ephemera keeps the history, and publishes every day how
well the public catalogue can actually see.Read the full README ↗About every eight hours, SpaceX posts a new public manifest of files saying where each Starlink satellite will be for the next two to three days, with its uncertainty. When the next manifest lands, that set is gone, and no continuous public archive of those sets exists. Ephemera keeps the history, and publishes every day how well the public catalogue can actually see.
The files SpaceX's public manifest lists at api.starlink.com/public-files/ephemerides/ give 48 to
72 hours of position and velocity with full covariance at 60-second steps, about three times a day.
Space-Track stopped mirroring those files on 28 July 2025, and the server keeps only the current
set. No continuous public archive of them exists; the only public sample is one week (25 Nov to
1 Dec 2024) inside the SpaceTrack-TimeSeries dataset (arXiv 2506.13034). This project exists because
SpaceX publishes the feed openly.
Ephemera archives every file each manifest lists under a witnessed hash chain, and computes from the archive a daily, browser-checkable scoreboard: the distance between the public catalogue's prediction and the operator's own, by element age and altitude shell, and the fraction of the constellation the public catalogue can see. Planned, not built: the self-consistency of the published covariance by lead time, and an honestly defined census of trajectory changes. Kelso keeps the elements. McDowell keeps the objects. This keeps the record of how well the world can see.
Status (8 Oct 2026). The whole pipeline runs on one rented server in the EU, since 3 Oct 2026: it pulls every manifest, roots each one in a Merkle tree stamped with OpenTimestamps, captures the manifest and ten root-chosen files into the Wayback Machine and re-hashes the copies, ships the raw files to cold storage, scores each cycle against the public catalogue, and publishes the site every four hours. Not yet in place: a second poller, the self-consistency series, the census, a restore drill. Progress is tracked in
PLAN.md. ReadDOCS/concept.mdfirst, thenDOCS/decisions.md, thenprobes/README.md, thenarchive/README.md.
A median of about 11,130 files per manifest (8,568 to 18,812 over the 120 manifests to 8 Oct
2026, from the published ledger), about 2 MB each, three manifests a day, and no continuous public
archive. The first measurements were taken 30 and 31 August 2026;
the canonical measurements table, with dates and methods, is probes/p1_cycle_pull/README.md;
re-measure before quoting.
| Path | What |
|---|---|
DOCS/ | Concept, decision log, claims register — read before building |
probes/ | Cheap experiments that retire the big unknowns before anything is built on them |
archive/ | Poller, watcher and witness: hashing, Merkle roots, OpenTimestamps, Wayback |
infra/ | Account guards (AWS and Cloudflare) and guarded deploy scripts |
tools/ | The claims lint that enforces the register's wording in make test |
web/ | The static site (build-time computation only, no server) |
score/ | Clean-room SGP4 scoring against the operator's published trajectories; self-consistency and the census are planned |
data/ | Planned: derived products only; raw files live on cold object storage, never here |
licences/ | Planned: per-source licence and terms audit |
Everything in this repository is re-implemented from open libraries (python-sgp4 / Skyfield /
Orekit; hashlib Merkle trees) on personal infrastructure with an account guard. No code from any
employer or client engagement is used, adapted, or consulted. The rule is recorded as D02 and
restated in CLAUDE.md because coding agents build most of this.
MIT for this repository's code. Derived datasets are intended for CC BY 4.0 (D06, proposed). The
raw SpaceX files are to be mirrored as published, under terms SpaceX has not stated; the position
taken is D06 and the status of the request to SpaceX (sent 31 Aug 2026; no answer by 30 Sep 2026,
the thirty days D06 allowed, so the files are held as published and come down on request) is in
DOCS/claims-register.md. Space-Track general perturbations will be redistributed only under
Space-Track's blanket approval for basic SSA data, with citation.