Experience Engine v0.1

A reusable infrastructure component for experience lifecycle management. Accumulates structured Experience Units with intent, outcome, and confidence — so the second attempt at any task is faster and more accurate than the first.

Core Loop

Intake → Process → Store → Retrieve → Apply → Feedback → Evolve

Structure

experience_engine/
├── core/           # Experience dataclass + schema versioning
├── ingestion/      # Intake entry point
├── processing/     # normalize → summarize → tag pipeline
├── storage/        # JSON-based persistence (memory.json)
├── retrieval/      # keyword / domain / tag / avoid filtering
├── feedback/       # correct / wrong → confidence + avoid flag
├── utils/          # id, time helpers
├── api.py          # Unified ExperienceAPI
└── docs/           # Conceptual + implementation specs

Usage

from api import ExperienceAPI

# Record an experience
exp = ExperienceAPI.add(
    raw="ADC error: adc1_config_width not found after IDF5.2 to 6.0 migration",
    domain="engineering",
    intent="migrate IDF5.2 to 6.0",
    outcome="success"
)

# Retrieve
results = ExperienceAPI.query(keyword="IDF", domain="engineering")

# Apply feedback
ExperienceAPI.feedback(exp.id, "correct")

Design Principles

• Not a knowledge base — a decision infrastructure
• Not a log — every unit has intent, outcome, and confidence
• Negative experiences matter — outcome: failed + feedback wrong → avoid: True
• Confidence is earned — feedback shifts it up or down over time

V0.2 Roadmap

• Semantic / vector retrieval
• Scope + decay scoring
• Experience composition (combine multiple units to solve complex problems)
• Auto skill generation from accumulated patterns

---

We do not store data. We accumulate experience.