How heraclitus0/cognize works
Cognize is a programmable cognition engine: a system supplies evidence, optionally through multimodal perception, and an epistemic state updates belief, drift, memory, and rupture status. Policies customize those updates; meta-policy selection and bounded evolution adapt policy behavior. Graphs coordinate multiple states and can apply programmable influence. The excerpts show the kernel’s policy machinery, perception encoders, and graph stepping, but omit substantial implementation details; documented capabilities are included without inferring unshown internal calls.
Main entry points: epistemic.py Epistemic state · network.py Epistemic graph · perception.py Perception fusion · meta_learning.py Evolution bounds
Components 14
Evidence input
- Perception fusion
cognize/perception.py - Custom encoders
Epistemic kernel
- Epistemic state
cognize/epistemic.py - State telemetry
cognize/epistemic.py
Policy adaptation
- Policy presets
cognize/policies.py - Meta-policy manager
cognize/epistemic.py - Policy memory
cognize/epistemic.py - Shadow evaluation
cognize/epistemic.py - Evolution bounds
cognize/meta_learning.py
Graph coordination
- Epistemic graph
cognize/network.py - Programmable edges
cognize/network.py - Cascade traces
cognize/network.py
Other components
- Host system
- Application consumer
Connections 15
- Host system → Perception fusion: supplies inputs
README.md - Host system → Epistemic state: supplies evidence
README.md - Perception fusion → Custom encoders: calls encoders
cognize/perception.py - Perception fusion → Epistemic state: provides vector
README.md - Epistemic state → Policy presets: uses injected policies
README.md - Epistemic state → Meta-policy manager: selects policies
cognize/epistemic.py - Meta-policy manager → Shadow evaluation: evaluates candidates
cognize/epistemic.py - Meta-policy manager → Policy memory: records rewards
cognize/epistemic.py - Evolution bounds → Meta-policy manager: sets evolution bounds
cognize/meta_learning.py - Epistemic graph → Epistemic state: steps graph nodes
README.md - Epistemic graph → Programmable edges: applies edge strategies
README.md - Epistemic graph → Cascade traces: records cascades
README.md - Epistemic state → State telemetry: exports state history
README.md - Epistemic state → Application consumer: returns state results
README.md - Epistemic graph → Application consumer: returns graph results
README.md