By the numbers
The current state of Meshfleet — releases, tests, performance, and the v1.0 perf gates we promised. All numbers come from the live GitHub repo and the installed mesh.
Releases
9
total releases
8
v0.8.x releases
7
blog posts
10
features shipped
Testing
181
tests passing
3 × 3
CI matrix (OS × Node)
17
test files
v1.0 performance gates
The v1.0 plan committed to two perf gates. Both verified in BENCHMARKS.md.
5.5ms
p50
Spawn bookkeeping (registerAgentInLedger)
Target: <100ms ✓
no drops
verified
10k messages on a single fleet
All 10,000 messages persisted to the ledger
Routing performance
How fast route_work runs against various fleet sizes.
| Operation | p50 | p95 | p99 |
|---|---|---|---|
| routeWork(roster=10) | 0.01ms | 0.03ms | 0.09ms |
| routeWork(roster=100) | 0.06ms | 0.12ms | 0.17ms |
| routeWork(roster=1000) | 0.70ms | 0.91ms | 1.14ms |
| sendMessage (warm) | 0.16ms | 0.31ms | 0.38ms |
| saveData (1k agents, 10k msgs) | 5.59ms | 6.63ms | 8.26ms |
| loadData (1k agents, 10k msgs) | 5.63ms | 6.62ms | 7.24ms |
Numbers from BENCHMARKS.md, generated by npx tsx benchmark/bench.ts. Re-run any time.
The v0.8.x series
v0.8.0 Hardening push — retry, recovery, schema versioning, route_work top_n
v0.8.1 Skill taxonomy (hierarchical matching)
v0.8.2 Routing feedback loop (Wilson-style)
v0.8.3 Synonym expansion (30+ dev terms)
v0.8.4 Performance benchmarks
v0.8.5 Template versioning
v0.8.6 Template sharing (export/import JSON)
v0.8.7 Fleet dashboard TUI
Stay in the loop
Subscribe via RSS, watch the changelog, or star the repo for release notifications.