End-to-end lifecycle simulation test suite

[CalebisGross]

Summary

Simulate the full mnemonic user journey from fresh install through 3 months of usage. This is a Go integration test binary at cmd/lifecycle-test/ that exercises all 8 cognitive agents, encoding pipeline, consolidation decay, dreaming cross-pollination, abstraction hierarchy, and retrieval quality under realistic scale.

Why

We have unit tests, an HTTP benchmark (cmd/benchmark/), and an IR quality benchmark (cmd/benchmark-quality/), but nothing validates the full user lifecycle — the thing our users actually experience. This fills that gap.

Approach

• Go integration test binary (not Playwright — mnemonic is a CLI/daemon/MCP system, not a web app)
• Real SQLite, stub LLM with optional --llm flag for real LLM testing
• Time simulation via timestamp backdating + direct agent RunOnce() calls
• 8 phases: Install → First Memories → Ingestion → Daily Use → Consolidation → Dreaming → Growth → Long-term Health
• Deterministic with fixed random seed for reproducible results

Phases

Phase	Simulates	Key Systems
1. Install & Setup	Day 0 fresh install	DB init, schema, config
2. First Memories	Day 1 first MCP use	Encoding, FTS5, episoding
3. Project Ingestion	Day 2 ingest_project	Bulk ingest, dedup
4. Daily Use	Days 3-14	Associations, Hebbian learning, feedback
5. Consolidation	Week 2	Decay, merge, patterns, signal/noise
6. Dreaming & Abstraction	Week 3-4	Cross-project links, insights, principles
7. Growth	Month 1-3 (500-1000 memories)	Retrieval quality under load, scaling
8. Long-term Health	Month 3	Archive, decay curves, regression test

Sub-issues

Tracked as individual issues in the Lifecycle Simulation milestone.

Verification

make build
./bin/mnemonic lifecycle-test                    # Run all phases
./bin/mnemonic lifecycle-test --phase=daily      # Run single phase
./bin/mnemonic lifecycle-test --report markdown   # Markdown report
./bin/mnemonic lifecycle-test --llm              # Real LLM mode