v5.2.0 — Token Economics, Dedup, Mode Profiles

What's New

Three features inspired by claude-mem scavenge:

💡 Token Economics Display

Search results now show compression value:

💡 Read ~42 tokens (saved 78% vs ~191 discovery tokens)

⚡ Content-Hash Dedup

SHA256 prevents duplicate memories. Same content within 60s? Blocked automatically.

🎛️ Mode Profiles

Switch behavior for different workflows:

• memory-tool mode dev — 15 results, full tracking
• memory-tool mode ops — 5 results, errors & learnings only
• memory-tool mode research — 20 results, deep exploration
• memory-tool mode monitor — 3 results, minimal output

Stats

• 510 tests passing
• 17 modules (cli, config, database, embedding, export, fsrs, graph, importance, memory_ops, relations, runs, snapshots, sync, dream, corrections, display, utils)
• Zero core dependencies

Install

pip install ai-iq==5.2.0
# With vector search:
pip install ai-iq[full]==5.2.0

Full Changelog: v5.1.0...v5.2.0