Repository navigation
🍯 TokenJar v1.1.0 — Official Release
🍯 TokenJar v1.1.0: AI Token Optimizer — Slicing Cache, Hybrid Fuzzy Search & Auto-Sync
TokenJar v1.1.0 delivers major architectural upgrades focused on fine-grained inspection caching, typo-tolerant symbol resolution, a modernized CLI command suite, and a zero-maintenance update pipeline:
🚀 Key Improvements
- ⚡ Slicing Session Cache: Targeted line slices (
start_line/end_line) and symbol reads are now cached with normalized fragment keys (#L1-30), providing 99.8% token savings and <1ms latency on repeated reads. - 🛡️ Auto-Paging Ceiling (80-line cap): Prevents runaway file reads from blowing up context windows or dumping
output.txtto disk by providing seamless continuation links. - 🎯 Hybrid Fuzzy Symbol Search: Combines Tree-sitter AST queries with Levenshtein & Trigram Dice similarity to automatically suggest
💡 Did you mean: ...?when typos occur (zero overhead on exact matches). - 🔄 Post-Update Auto-Sync (
tokenjar update): Upgrades binary, IDE configurations, slash commands, and in-placeAGENTS.mdrules across all active projects in a single step. - 🧹 Zero-Trace Surgical Uninstall (
tokenjar uninstall): Purges all configs, schemas, and caches without ever touching user code or custom rules. - 💡 Recommended 3-Step Workflow: Guides users to install into a permanent folder (
tokenjar install) and activate rules directly via terminal (tokenjar on).
💻 Modernized CLI Commands & Migration Guide
TokenJar v1.1.0 streamlines the developer CLI with clean, intuitive commands (all legacy command names remain supported as backward-compatible aliases):
| Action | New Command (v1.1.0) | Replaces (Legacy) | Details |
|---|---|---|---|
| 🟢 Project Activation | tokenjar on |
tokenjar init-rules |
Injects TokenJar optimization rules into repository AGENTS.md |
| ⚪ Project Deactivation | tokenjar off |
tokenjar init-rules --clean |
Cleanly removes TokenJar optimization rules from project |
| 🌐 Global IDE MCP Setup | tokenjar on --global / tokenjar enable |
tokenjar install-mcp |
Configures MCP across all detected IDEs (Cursor, Claude, AGY, Windsurf) |
| 🔌 Global MCP Removal | tokenjar off --global / tokenjar disable |
tokenjar uninstall-mcp |
Safely removes MCP server from all IDE configurations |
| 🧹 Cache & Metrics Reset | tokenjar clean |
tokenjar cache-clear, reset-stats |
Reset metrics (--stats) and clear L2 SQLite cache (--cache) |
| 🔄 Engine Self-Update | tokenjar update |
manual re-download | Auto-upgrades binary + auto-syncs IDEs, slash commands & project rules |
tokenjar uninstall |
manual file cleanup | Complete removal: reverts IDE configs, cleans rules, purges PATH & cache | |
| 🎛️ Output Optimization | tokenjar output on / off |
manual config | Toggle compact surgical diffs and zero-truncation quality mandate |
🧪 Benchmarks & Stress Test Metrics
- 50-Step Real-Life Developer Scenario:
- 50/50 steps passed (100% success rate) in 2.17 seconds (43.4 ms/step).
- Raw Tokens Processed: 150,060 tokens -> Delivered to Model: 4,641 tokens.
- Net Token Reduction: 96.9% (145,419 tokens saved in a single session).
- Est. Financial Savings: ~$0.44 USD per session (~$8.80 / 1,000 steps).
- Quality & Accuracy: 100.0% (Zero assertion failures, Zero data loss).
- 100-Step Extreme Stress Test:
- SQLite WAL Multi-threaded Contention: 407 ops/sec across 12 worker threads.
- AST Skeletons: 0.10ms – 0.42ms per file.
- Terminal Stream Pruning: 4.40ms on 1,500-line logs.
For installation and full documentation, visit the TokenJar Repository.
Full Changelog: v1.0.3...v1.1.0
Full Changelog: v1.0.3...v1.1.0
Full Changelog: v1.0.3...v1.1.0
Full Changelog: v1.0.3...v1.1.0