name

generic-agent

description

Self-evolving agent memory via context information density maximization. Crystallizes completed tasks into reusable Skills (L3), maintains layered memory (L0–L4), and compresses context to preserve decision-relevant signal. Hybrid model: project-scoped (.pi-memory/ in repo) + global (~/.pi/agent/skills/generic-agent/memory/). Project skills shadow global skills. Auto-triggered after reflect_capture to sync learnings, and before tasks to recall relevant skills. Use when building on prior work, iterating on similar tasks, preserving session learnings, or managing long-running sessions.

GenericAgent — Self-Evolving Memory for pi

Core Principle

Context information density maximization. Performance is determined not by context length, but by how much decision-relevant information is maintained within a finite budget.

Auto-Trigger (Two-Way Integration with reflect)

After reflect_capture fires:

reflect_capture (auto) → generic-agent triggers → sync_from_reflect.py
    → Converts reflect captures → L3 Skills (patterns) / L2 Facts (process/knowledge)
    → Skills indexed and immediately available for future recall

Before any task:

recall.py --query "<user request>"
    → Searches project + global L3 Skills
    → Searches L2 Facts
    → Searches L4 Sessions
    → Searches reflect's captures (bonus)
    → Returns highest-density context for the current task

Memory Layers (L0–L4) × Hybrid Scope

Layer	Global	Project	Purpose
L0	`memory/.l0_meta/`	`<repo>/.pi-memory/.l0_meta/`	Agent meta-rules
L1	`memory/.l1_index/`	`<repo>/.pi-memory/.l1_index/`	Fast-routing index
L2	`memory/.l2_facts/`	`<repo>/.pi-memory/.l2_facts/`	Stable knowledge
L3	`memory/.l3_skills/`	`<repo>/.pi-memory/.l3_skills/`	Reusable Skills
L4	`memory/.l4_sessions/`	`<repo>/.pi-memory/.l4_sessions/`	Session archives

Quick Commands

# Recall before starting a task (do this first!)
python memory/recall.py --query "<what you need>"

# Sync reflect captures to skills (after reflect_capture fires)
python memory/sync_from_reflect.py

# Evolve after completing a task (manual backup if reflect missed something)
python memory/evolve.py --session <id> --task "<description>"

# Compress long sessions
python memory/compress_context.py --session <id> --budget 6000

# List all skills
python memory/recall.py --list-skills

# Force specific layer
python memory/recall.py --query "<q>" --global-only
python memory/recall.py --query "<q>" --project-only

Reflect → Skill Mapping

reflect type	→ generic-agent layer	Example
`type=pattern`	L3 Skill	Error fix SOPs
`type=process`	L2 Fact	Campaign lifecycle
`type=knowledge`	L2 Fact	Domain insights

Layer Detection

--project <path> — explicit project path
--global — force global layer
Auto-detect .pi-memory/ in cwd → project layer
Fall back to global

Key Principle

After reflect_capture auto-runs, immediately call:

python memory/sync_from_reflect.py --dry-run  # preview what will sync
python memory/sync_from_reflect.py             # actually sync

This converts every reflect capture into a searchable, reusable Skill. The more you use pi, the smarter it gets — automatically.