Skip to content

Releases

Rediscover 0.1.0 — real physics, checkable evidence

Rediscover 0.1.0 — real physics, checkable evidence Latest
Latest

Choose a tag to compare

View all tags
@shyamolkonwar shyamolkonwar released this 06 Nov 18:46
v0.1.0
13c2e5f

Rediscover 0.1.0

An open-source physics workbench and MCP server for finding known gravitational-wave signals in real public detector strain.

Included

  • Official GWOSC data access, versioned inputs and checksum cache.
  • Finite nonspinning IMRPhenomD matched-filter search with PyCBC.
  • Data quality checks, coarse detector-frame mass estimates and cross-detector timing.
  • Offline HTML reports with raw/conditioned strain, Q-transform and SNR plots.
  • CLI and nine stdio MCP tools.
  • Optional bounded Azure/Responses agent runner with evidence validation.
  • Actual baseline, gpt-6-sol and gpt-5.5 results: six demonstration tasks each.
  • MIT license, methodology, tests, contribution and security documentation.

Verified

43 tests passed on the Azure Linux VM, including real H1/L1 data and an actual MCP client/server exchange. The dependency audit reported no known vulnerabilities. A clean Python 3.12 virtual environment installed the built wheel and successfully executed the full GW150914 report workflow. All testing and package builds took place remotely.

Install

uv tool install --python 3.12 'git+https://github.com/shyamolkonwar/rediscover.git@v0.1.0'
rediscover run GW150914-v3 --catalog

Alternatively download the wheel below and install it inside a Python 3.12 virtual environment. This release is on GitHub, not PyPI; the existing PyPI rediscover package is unrelated.

Assets

  • Python wheel and source distribution.
  • gw150914.html: real, self-contained report; download and open in your browser.
  • gw150914-evidence.zip: full run inputs, evidence records and checksum manifest.

Scientific limits

This is a known-event educational reproduction, not a new discovery or calibrated detection-significance calculation. Masses are coarse detector-frame bank points, not posterior credible intervals. The small demonstration suite does not rank models. The timestamp-rounding correction made before the release run is disclosed in examples/verified/RESULTS.md.

MCP registry and PyPI listings are not published. Use the documented GitHub installation and stdio MCP configuration.

Assets 6
Loading