Sairo v3.0.0 — AI Storage Intelligence + Petabyte-Scale Performance

[ashwathstephen]

We're excited to release Sairo v3.0.0 — our biggest update yet.

What's New

AI Storage Intelligence (MCP)

Sairo now includes an optional MCP server that lets AI assistants like Claude, Cursor, and Windsurf analyze your entire storage infrastructure using natural language.

Instead of clicking through dashboards, just ask:

• "What's eating all the space?"
• "How much is this costing me?"
• "Find all parquet files modified this week"
• "Are there any duplicate files?"
• "Run a full storage audit"

26 tools, 4 guided workflows, zero configuration. The AI picks the right tools automatically.

Setup guide →

Petabyte-Scale Performance

• Folder listing: 191,231x faster — 311ms → 0.002ms on 2M objects
• SQLite PRAGMA tuning — 64MB cache, 256MB mmap, in-memory temp store
• Crawl workers: 6→12 buckets, 4→16 prefix threads
• 10K batch inserts — 5x fewer SQLite commits
• Async FTS rebuild — Search stays available during index rebuilds
• Sub-prefix splitting — Auto-parallelizes single-prefix buckets with 500K+ objects

Tested on Production

All changes benchmarked on real production data: 557K objects, 167 TB, running on Leaseweb StorageGRID.

Upgrade

docker pull stephenjr002/sairo:v3.0.0

Or update your Helm chart:

helm upgrade sairo oci://registry-1.docker.io/stephenjr002/sairo-helm

Full changelog: CHANGELOG.md

---

Questions? Feature requests? Drop them in Q&A or Ideas.