Skip to content

Commit b2bac1f

Browse files
authored
Merge pull request #34 from VecGrep/chore/release-1.5.0
chore: release 1.5.0 changelog
2 parents 5544aea + a6e4fad commit b2bac1f

File tree

1 file changed

+53
-0
lines changed

1 file changed

+53
-0
lines changed

CHANGELOG.md

Lines changed: 53 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -4,6 +4,59 @@ All notable changes to VecGrep are documented here.
44

55
---
66

7+
## [1.5.0] — 2026-02-28
8+
9+
### Breaking Changes
10+
11+
- **Store backend replaced**: SQLite + numpy vector store replaced with LanceDB.
12+
Existing indexes at `~/.vecgrep/` must be re-indexed after upgrading —
13+
simply run `index_codebase` again on your project.
14+
15+
### Added
16+
17+
**Performance**
18+
- **fastembed ONNX backend** — default embedding backend switched from
19+
sentence-transformers (PyTorch) to fastembed (ONNX Runtime).
20+
MCP server startup: ~6.6s → ~1.25s. First embed call: ~2–3s → ~100ms.
21+
- **IVF-PQ ANN index**`build_index()` creates an approximate nearest
22+
neighbour index after indexing for sub-linear search on large codebases.
23+
- **`file_stats` table** — O(files) change detection table. Incremental
24+
re-indexing now reads one row per file instead of scanning all chunks.
25+
- **Auto device detection** — selects Metal (Apple Silicon), CUDA (NVIDIA),
26+
or CPU automatically for the torch backend.
27+
28+
**Configuration**
29+
- `VECGREP_BACKEND` env var — `onnx` (default, fast) or `torch` (any HF model).
30+
- `VECGREP_MODEL` env var — override the default HuggingFace model ID.
31+
- Default model switched to [`isuruwijesiri/all-MiniLM-L6-v2-code-search-512`](https://huggingface.co/isuruwijesiri/all-MiniLM-L6-v2-code-search-512),
32+
fine-tuned for semantic code search (384-dim, ~80MB one-time download).
33+
34+
**MCP tools**
35+
- `stop_watching` — new tool to explicitly stop watching a codebase path.
36+
- Watch state persisted to `~/.vecgrep/watched.json` and restored on restart.
37+
- Background startup sync via daemon thread — server starts immediately.
38+
39+
**Developer experience**
40+
- `.githooks/pre-commit` lint check — run `git config core.hooksPath .githooks`
41+
once per clone to enable.
42+
43+
### Fixed
44+
45+
- `build_index()` passed `"vector"` as positional arg to `create_index()`,
46+
mapping to `metric` instead of `vector_column_name`.
47+
- `list_tables()` returned `ListTablesResponse` instead of a plain list,
48+
causing `set()` to fail with an unhashable type error.
49+
50+
### Performance summary
51+
52+
| Metric | 1.0.0 | 1.5.0 |
53+
|---|---|---|
54+
| MCP server startup | ~6.6s | ~1.25s |
55+
| Model load (first embed) | ~2–3s | ~100ms |
56+
| Change detection | O(chunks) SHA-256 | O(files) mtime+size |
57+
58+
---
59+
760
## [1.0.0] — 2026-02-22
861

962
First stable release.

0 commit comments

Comments
 (0)