feat: outer loop Phase 1 — population, mutations, similarity, serialization

[akashgit]

Summary

Phase 1 of the outer loop evolutionary swarm search system. Implements the foundational layer: Pydantic models, graph mutation operators, population/archive management, novelty filtering, and workflow serialization.

Components

• factory/outer_loop/models.py — Strict Pydantic v2 models: MutationType, MutationRecord, Individual, HyperparameterRecord, SwarmConfig, OuterLoopState, GenerationSummary
• factory/outer_loop/mutations.py — 6 structured graph mutation operators (insert_node, remove_node, redirect_edge, parallelize, serialize, mutate_params) with MutationStrategy protocol and WeightedRandomStrategy default implementation. Includes validate_and_repair for post-mutation graph integrity.
• factory/outer_loop/population.py — Population (collection management, tournament selection) and MAPElitesArchive (4D quality-diversity grid with Pareto front computation)
• factory/outer_loop/similarity.py — structural_hash (SHA-256 canonical form), graph_edit_distance, compute_features (depth/fork_degree/agent_count/gate_count), NoveltyFilter
• factory/workflow/primitives.py — Added Workflow.to_dict() and Workflow.from_dict() for round-trip serialization with _type discriminators

Tests

85 tests across 5 test files covering all components with round-trip serialization, edge cases, and integration with all builtin workflows.

Target Branch

feat/outer-loop-evolution