Skip to content

Polymathix OS v1.0.2: Cross-Domain Physical Invariants & Cryptographic Scientific Provenance Ledger

Choose a tag to compare

View all tags
@sohel3shaikh sohel3shaikh released this 23 Oct 10:26
· 1 commit to master since this release
v1.0.2
1ea3783

🚀 Polymathix OS v1.0.2 — Cross-Domain Correctness & Scientific Reproducibility

Polymathix OS v1.0.2 introduces a mathematically grounded Cross-Domain Physical Invariants Verification Engine and a Cryptographic Scientific Provenance Ledger, directly addressing the fundamental challenges of scientific correctness across disparate domains and deterministic reproducibility across diverse LLMs and local Ollama model versions.

🌟 What's New in v1.0.2

1. 🛡️ Cross-Domain Physical Invariants & Correctness Verifier (polymathix.core.invariants)

How do you define "correctness" across domains as fundamentally different as protein docking and crystallographic lattices?
In physical sciences, correctness cannot be a single heuristic. Polymathix formalizes correctness through domain-specific conservation laws and physical invariant suites evaluated during the VERIFY stage:

  • Structural Biology & Biomolecular Mechanics:
    • Pauli Repulsion Steric Clash Exclusion: Enforces non-bonded interatomic thresholds ($d \ge 0.70 imes (r_{\text{vdW},i} + r_{\text{vdW},j})$).
    • Hydrogen-Bond Potential Well Bounds: Strict Donor-Acceptor distance ($2.4\text{ \AA} \le d \le 3.6\text{ \AA}$) and angular criteria.
    • Coulombic Salt-Bridge Pairing Radius: Cation-anion interaction cutoff within aqueous Debye-Hückel screening limits ($d \le 4.0\text{ \AA}$).
    • Peptide Backbone Metric Continuity: Trans-peptide Cα-Cα spacing bounds ($2.5\text{ \AA} \le d \le 4.5\text{ \AA}$, centered at $3.8\text{ \AA}$).
  • Solid-State Chemistry & Inorganic Crystallography:
    • Unit Cell Formal Electroneutrality: Formal oxidation state summation over unit cell stoichiometry strictly equals zero (e.g. in $\text{SrTiO}_3$, $\text{Sr}^{2+} + \text{Ti}^{4+} - 3\times\text{O}^{2-} = 0.00$).
    • Shannon-Prewitt Interatomic Radius Bound: Verifies interatomic spacing against coordinated ionic/covalent radii sums without unphysical lattice collapse.
    • Space Group Symmetry Invariance: Validates fractional coordinate consistency under space group symmetry operations (e.g., $Pm\bar{3}m$ No. 221).
    • Volumetric Mass Density Concordance: Unit cell theoretical mass density concordance (e.g. $5.12\text{ g/cm}^3$ for $\text{SrTiO}_3$, $2.26\text{ g/cm}^3$ for Graphene/SWCNT).
  • Cheminformatics & Molecular Mechanics:
    • Kekulé Octet Valence Saturation: Zero hypervalent atom centers.
    • Lipinski Rule-of-Five Compliance: Automated bounds screening for molecular weight and lipophilicity ($\text{LogP}$).
    • Force-Field Residual Energy Gradient: RMS force gradient convergence below stationary potential thresholds.
  • Automated Correctness Dossier: Generates a structured verification dossier with confidence scoring ($0.0 \to 1.0$) and explicit numerical residual tracking.

2. 🔐 Cryptographic Scientific Provenance Ledger & Deterministic Replay Engine (polymathix.core.provenance)

How do you guarantee reproducibility when users attach different LLMs or update local Ollama models?
Polymathix v1.0.2 introduces a 3-Pillar Reproducibility Architecture:

  1. Separation of Intent vs Computation: The LLM functions strictly as an intent-to-DSL compiler (producing canonical SMILES, space group symbols, stoichiometry, and seed specifications). The LLM is never permitted to invent raw floating-point coordinates or energy trajectories.
  2. Deterministic Physics Kernels: All physical simulations, force-field minimizations (MMFF94 / OpenMM), and coordinate math are executed by seed-locked, bit-reproducible scientific kernels.
  3. Cryptographic Run Manifest (run_provenance): Every run records an immutable ledger containing:
    • Model Identity & Checksum: Provider name, model identifier, model SHA256 digest, sampling temperature ($T=0.0$ greedy argmax), and PRNG seed (default $42$).
    • Prompt Fingerprint: SHA256 hash of system and user prompts.
    • Canonical Intermediate Representation (IR): Formula, atom count, space group, and SHA256 coordinate manifold fingerprint.
    • Deterministic Replay Engine: Allows any researcher with a run manifest to replay and verify the discovery workflow bit-for-bit without calling an LLM.

3. 🧪 Comprehensive Test Suite (17 / 17 Tests Passing)

Expanded automated verification suite covering both new modules:

python scripts/test_all.py
# [TEST] 16. Cross-Domain Physical Invariants & Correctness Verifier... --> PASSED
# [TEST] 17. Cryptographic Run Provenance & Deterministic Replay Engine... --> PASSED
# VERIFICATION SUMMARY: 17 PASSED, 0 FAILED

🎬 Quick Start (Run Locally in 3 Steps)

# 1. Clone the repository
git clone https://github.com/sohel3shaikh/Polymathix.git
cd Polymathix

# 2. Install dependencies
pip install -r requirements.txt

# 3. Launch Polymathix OS
python -m streamlit run app.py

Open your browser at http://localhost:8501.

📦 Release Assets

  • Polymathix-Master-Release.zip: Complete offline standalone release archive (v1.0.2).
  • Polymathix_Showcase_Demo1.mp4: Authentic 3-minute video walkthrough and real UI demonstration.
  • Polymathix_Showcase_Demo.mp4: High-definition video showcase asset.
Assets 5
Loading