agentic-flow 2.0.1-alpha.18 → 2.0.1-alpha.19
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/CHANGELOG.md +38 -0
- package/dist/.tsbuildinfo +1 -1
- package/dist/intelligence/EmbeddingService.d.ts +182 -0
- package/dist/intelligence/EmbeddingService.d.ts.map +1 -1
- package/dist/intelligence/EmbeddingService.js +879 -0
- package/dist/intelligence/EmbeddingService.js.map +1 -1
- package/package.json +1 -1
- package/wasm/reasoningbank/reasoningbank_wasm_bg.js +2 -2
- package/wasm/reasoningbank/reasoningbank_wasm_bg.wasm +0 -0
package/CHANGELOG.md
CHANGED
|
@@ -2,6 +2,44 @@
|
|
|
2
2
|
|
|
3
3
|
All notable changes to this project will be documented in this file.
|
|
4
4
|
|
|
5
|
+
## [2.0.1-alpha.19] - 2025-12-31
|
|
6
|
+
|
|
7
|
+
### Added
|
|
8
|
+
- **Cache Pretrain**: Pre-populate embedding cache for faster cold starts
|
|
9
|
+
- `pretrain(sources)` - Pretrain from files, globs, or text arrays
|
|
10
|
+
- `pretrainCodePatterns()` - 45 common programming patterns (~3ms)
|
|
11
|
+
- `pretrainFromRepo(path)` - Crawl and cache entire repositories
|
|
12
|
+
|
|
13
|
+
- **AI-Enhanced Pretrain** (ruvector integration):
|
|
14
|
+
- `pretrainWithAI()` - Uses attention mechanisms for intelligent caching
|
|
15
|
+
- `pretrainIntelligent()` - 5-stage analysis (patterns, AST, git, deps, chunks)
|
|
16
|
+
- `pretrainIncremental()` - Only cache changed files (git diff)
|
|
17
|
+
- `pretrainSemantic()` - Code-aware chunking (functions, classes, interfaces)
|
|
18
|
+
- `prefetchForContext()` - Context-aware prefetch based on current task
|
|
19
|
+
|
|
20
|
+
- **Attention-Guided Caching**:
|
|
21
|
+
- `HyperbolicAttention` - Hierarchical code structure (AST trees)
|
|
22
|
+
- `MoEAttention` - Expert routing (frontend/backend/data/test)
|
|
23
|
+
- `GraphRoPeAttention` - Dependency graph understanding
|
|
24
|
+
- `FastGRNN` - Pattern prediction for smart prefetch
|
|
25
|
+
|
|
26
|
+
- **Smart Pretraining Features**:
|
|
27
|
+
- Semantic chunking by code boundaries (not fixed size)
|
|
28
|
+
- Priority-based caching of frequently accessed patterns
|
|
29
|
+
- Background pretraining (non-blocking)
|
|
30
|
+
- Git history analysis (cache hot files first)
|
|
31
|
+
- Dependency graph caching
|
|
32
|
+
|
|
33
|
+
### Performance (pretrain)
|
|
34
|
+
| Operation | Time | Notes |
|
|
35
|
+
|-----------|------|-------|
|
|
36
|
+
| Code patterns | ~3ms | 45 patterns, one-time |
|
|
37
|
+
| Custom texts | ~0.3ms | 5 texts batch |
|
|
38
|
+
| Cached query | ~0.01ms | Memory cache hit |
|
|
39
|
+
| Incremental | ~50ms | Git-based, 20 commits |
|
|
40
|
+
| AI pretrain | ~200ms | Full 5-stage analysis |
|
|
41
|
+
| Context prefetch | ~5ms | Task-aware prediction |
|
|
42
|
+
|
|
5
43
|
## [2.0.1-alpha.18] - 2025-12-31
|
|
6
44
|
|
|
7
45
|
### Added
|