sigmap 3.4.0 → 3.5.1

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.
Files changed (18) hide show
  1. package/AGENTS.md +610 -17
  2. package/CHANGELOG.md +35 -0
  3. package/README.md +18 -14
  4. package/gen-context.js +561 -8
  5. package/package.json +2 -1
  6. package/packages/adapters/llm-full.js +25 -0
  7. package/packages/cli/package.json +1 -1
  8. package/packages/core/index.js +10 -0
  9. package/packages/core/package.json +1 -1
  10. package/src/eval/analyzer.js +3 -0
  11. package/src/extractors/generic.js +26 -0
  12. package/src/extractors/patterns.js +135 -0
  13. package/src/extractors/python_dataclass.js +77 -0
  14. package/src/extractors/typescript_react.js +60 -0
  15. package/src/extractors/vue_sfc.js +99 -0
  16. package/src/format/llm-txt.js +28 -0
  17. package/src/format/llms-txt.js +70 -0
  18. package/src/mcp/server.js +1 -1
package/README.md CHANGED
@@ -19,6 +19,8 @@
19
19
  npx sigmap # 10 seconds. zero config. your AI never reads the wrong file again.
20
20
  ```
21
21
 
22
+ > Latest: **v3.5.0** adds Phase C/D intelligence expansion with framework-specialized extractors and cross-module pattern detection.
23
+
22
24
  <div align="center">
23
25
  <img src="demo.gif" alt="SigMap demo — reducing 80K tokens to 4K in under 10 seconds" width="760" />
24
26
  </div>
@@ -48,15 +50,15 @@ npx sigmap # 10 seconds. zero config. your AI never reads the wrong file again
48
50
 
49
51
  <div align="center">
50
52
  <img src="docs/comparison-chart.svg" alt="SigMap benchmark — before vs after across 3 RAG quality metrics" width="700" />
51
- <br/><sub><a href="https://manojmallick.github.io/sigmap/guide/task-benchmark.html">80 tasks · 16 real repos · no LLM API · <strong>full methodology →</strong></a></sub>
53
+ <br/><sub><a href="https://manojmallick.github.io/sigmap/guide/task-benchmark.html">90 tasks · 18 real repos · no LLM API · <strong>full methodology →</strong></a></sub>
52
54
  </div>
53
55
 
54
56
  | | Without SigMap | With SigMap |
55
57
  |---|:---:|:---:|
56
58
  | Task success | 10% | **59%** |
57
- | Prompts per task | 2.84 | **1.54** |
59
+ | Prompts per task | 2.84 | **1.78** |
58
60
  | Tokens per session | ~80,000 | **~2,000** |
59
- | Right file found | 13.7% | **87.5%** |
61
+ | Right file found | 13.6% | **84.4%** |
60
62
  | Hallucination risk | 92% | **0%** |
61
63
 
62
64
  </details>
@@ -124,24 +126,26 @@ Reproduced with `node scripts/run-benchmark.mjs` on public repos:
124
126
 
125
127
  | Repo | Language | Raw tokens | After SigMap | Reduction |
126
128
  |------|----------|------------|--------------|-----------|
127
- | express | JavaScript | 15.5K | 201 | **98.7%** |
128
- | flask | Python | 84.8K | 3.4K | **96.0%** |
129
- | gin | Go | 172.8K | 5.7K | **96.7%** |
130
- | spring-petclinic | Java | 77.0K | 634 | **99.2%** |
129
+ | express | JavaScript | 70.6K | 911 | **98.7%** |
130
+ | flask | Python | 147.9K | 6.7K | **95.4%** |
131
+ | gin | Go | 216.4K | 6.0K | **97.2%** |
132
+ | spring-petclinic | Java | 97.9K | 3.4K | **96.5%** |
131
133
  | rails | Ruby | 1.5M | 7.1K | **99.5%** |
132
- | axios | TypeScript | 31.7K | 1.5K | **95.2%** |
133
- | rust-analyzer | Rust | 3.5M | 5.9K | **99.8%** |
134
+ | axios | TypeScript | 105.7K | 6.1K | **94.3%** |
135
+ | rust-analyzer | Rust | 3.5M | 6.3K | **99.8%** |
134
136
  | abseil-cpp | C++ | 2.3M | 6.3K | **99.7%** |
135
- | serilog | C# | 113.7K | 5.8K | **94.9%** |
136
- | riverpod | Dart | 682.7K | 6.5K | **99.0%** |
137
+ | serilog | C# | 195.5K | 6.9K | **96.4%** |
138
+ | riverpod | Dart | 747.2K | 6.5K | **99.1%** |
137
139
  | okhttp | Kotlin | 31.3K | 1.4K | **95.5%** |
138
140
  | laravel | PHP | 1.7M | 7.2K | **99.6%** |
139
141
  | akka | Scala | 790.5K | 7.1K | **99.1%** |
140
- | vapor | Swift | 171.2K | 6.4K | **96.3%** |
141
- | vue-core | Vue | 404.2K | 8.8K | **97.8%** |
142
+ | vapor | Swift | 171.4K | 6.4K | **96.2%** |
143
+ | vue-core | Vue | 414.4K | 8.6K | **97.9%** |
142
144
  | svelte | Svelte | 438.2K | 8.0K | **98.2%** |
145
+ | fastify | JavaScript | 54.4K | 2.6K | **95.3%** |
146
+ | fastapi | Python | 178.4K | 5.2K | **97.1%** |
143
147
 
144
- **Average: 99.3% reduction across 16 languages.** See [`benchmarks/reports/token-reduction.md`](benchmarks/reports/token-reduction.md) or reproduce with `node scripts/run-benchmark.mjs`.
148
+ **Average: 97.5% reduction across 18 repos (16 languages).** See [`benchmarks/reports/token-reduction.md`](benchmarks/reports/token-reduction.md) or reproduce with `node scripts/run-benchmark.mjs`.
145
149
 
146
150
  ---
147
151