v0.6.0-ai-dev-governance.md

Astaire Benchmark Snapshot — v0.6.0

Captured: 2026-04-20 Dataset: ai-dev-governance dogfood workspace via .astaire/memory_palace.db

Command:

UV_CACHE_DIR=/tmp/uvcache uv run --project astaire python -m benchmarks.bench_context --db .astaire/memory_palace.db --root .

Results:

# Astaire Performance Benchmarks

## Tag query latency
- Astaire Ms: 3.03
- Raw Fs Ms: 403.12
- Speedup: 133.1
- Astaire Results: 4
- Raw Fs Files: 8

## Token savings (scoped context, budget=4000)
- Astaire Tokens: 2125
- Raw Fs Tokens: 5035
- Savings Pct: 57.8
- Astaire Ms: 9.99
- Raw Fs Ms: 192.32
- Raw Fs Files: 2

## Token savings (L0 vs all documents)
- L0 Tokens: 563
- All Doc Tokens: 89900
- Savings Pct: 99.4
- Compression Ratio: 159:1

## Assembly latency (full collection)
- Astaire Ms: 21.9
- Raw Fs Ms: 194.88
- Speedup: 8.9
- Astaire Tokens: 7815
- Raw Fs Tokens: 52147
- Raw Fs Files: 46

Current graphify-backed release state in the same workspace:

• 5 active claims
• 1 relationship
• live graphify.query route at L0
• fresh no-op graphify verification visible in recent activity