FLUX Negative Space Engine

"What's left when you eliminate everything else IS the truth." — Sherlock Holmes

The space BETWEEN agents IS the computation.

Core Insight

FLUX(A,B) = L_composed - L_A - L_B

Where:

• L_A = Agent A's Laplacian (its internal constraint structure)
• L_B = Agent B's Laplacian (its internal constraint structure)
• L_composed = Their combined Laplacian (with coupling)

The residual — what exists in the composed system but not in either individual — is the collaborative intelligence. It only exists in the negative space between agents.

Architecture

Core Engine (core.py)

• FluxNegativeSpace.compose() — Compute FLUX between two agents
• FluxNegativeSpace.multi_compose() — FLUX across 3+ agents
• Returns eigenvalues, collaborative modes, flux energy, structure detection

Casting Call Engine (casting_call.py)

• Tasks are Laplacians (task structure)
• Agent resumes are spectral fingerprints
• Best match = best FLUX energy with the task
• Agent affinity = FLUX between any two agents

Peer Review Engine (peer_review.py)

• Agent outputs are Laplacians
• Conservation ratio = how expected the output is
• Rayleigh quotient = the review score
• Low conservation = anomaly detected

Decomposition Engine (decomposition.py)

• Complex tasks → Fiedler partition → subtasks
• Each subtask's FLUX with available agents is computed
• Optimal assignment = max total FLUX energy
• Greedy with diversity bonus

Spreadsheet-as-Graph Engine (spreadsheet.py)

• Rows × columns = bipartite graph
• Cell values = edge weights
• Column correlations = structural Laplacian
• Conservation = data integrity
• Anomaly detection = conservation drop under perturbation

Installation

pip install numpy

Usage

from flux_negative_space import FluxNegativeSpace

flux = FluxNegativeSpace()
result = flux.compose(agent_A_laplacian, agent_B_laplacian, coupling_strength=0.1)

print(f"Collaborative modes: {result.collaborative_modes}")
print(f"FLUX energy: {result.flux_energy}")
print(f"Structured: {result.is_structured}")

Run Demo

python demo.py

Dependencies

• Built on FLUX Constraint Language
• Uses SpectralMessage for agent Laplacian construction
• numpy for linear algebra

License

MIT

Part of the SuperInstance OpenConstruct ecosystem.