trueline-mcp 2.4.4 → 2.4.5

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -100,6 +100,29 @@ disk.
100
100
 
101
101
  No more silent corruption. No more ambiguous string matches.
102
102
 
103
+ ### Benchmarks
104
+
105
+ Measured on real project files (`src/streaming-edit.ts`, 529 lines),
106
+ comparing total bytes through the context window (call + result, ÷4 ≈
107
+ tokens):
108
+
109
+ | Workflow | Built-in | Trueline | Saved |
110
+ |-------------------------|----------|-----------|-------|
111
+ | Navigate & understand | 22 094 | 3 609 | 84% |
112
+ | Explore then edit | 22 729 | 8 515 | 63% |
113
+ | Search & fix | 22 731 | 812 | 96% |
114
+ | Multi-region read | 22 094 | 2 720 | 88% |
115
+ | Multi-file exploration | 39 296 | 1 761 | 96% |
116
+ | **Total** |**128 944**| **17 417**| **86%** |
117
+
118
+ The search-and-fix workflow saves the most: a single `trueline_search`
119
+ call replaces grep + full-file read + old-string echo, cutting **96%**
120
+ of token usage. Even the explore-then-edit workflow — which includes an
121
+ exploratory read, a targeted re-read, and an edit — still saves **63%**
122
+ over the built-in equivalent.
123
+
124
+ Run the benchmark yourself: `bun run benchmarks/token-benchmark.ts`
125
+
103
126
  ## Design
104
127
 
105
128
  See [DESIGN.md](DESIGN.md) for the protocol specification, hash