EDAT
Evolution, learning, and pattern capture
EDAT (Evolutionary Data and Training) is the learning layer. Every error, insight, successful pattern, and architectural decision flows through EDAT. It surfaces reusable patterns to agents via NoqNoq query, preventing the organism from making the same mistake twice.
Pattern Capture
When an agent solves a problem, the pattern is captured and stored. Next time any agent faces the same problem, the solution is already available.
Error Learning
Errors are not just fixed — they are recorded with root cause, fix, and prevention pattern. The organism learns from every failure.
MCP Server
Exposed as a Model Context Protocol server. Claude and other LLMs query EDAT patterns directly via MCP tool calls.
Cross-Session
Patterns persist across all sessions and all agent instances. Knowledge gained by one Claude session is available to all future sessions.
Validation Layer
EDAT validates that new code and decisions are consistent with known patterns. It warns when an agent is about to repeat a documented anti-pattern.
qSlice Integration
Patterns are stored in qSlice-DB using the Five-Tree architecture. EDAT queries use qSlice graph traversal to find related patterns.