Release v0.1.4

Author: adaamko

CHANGELOG.md

@@ -2,6 +2,15 @@
 
 All notable changes to Squeez are documented here.
 
+## [0.1.4] - 2026-04-27
+
+### Added
+- **Extractive backend (`highlighter`)**: auto-detected support for `KRLabsOrg/verbatim-rag-modern-bert-v2` (and any compatible Verbatim-RAG ModernBERT span model). Same CLI, same Python API — just point `SQUEEZ_LOCAL_MODEL` at the HF id.
+- **Session stats**: `--stats` prints token/line savings per call; `--summary` prints accumulated savings across the session. Logged to `~/.cache/squeez/session_stats.jsonl`.
+
+### Changed
+- Docs reframed around two model types (generative vs extractive) instead of internal backend names.
+
 ## [0.1.3] - 2026-03-18
 
 ### Fixed

README.md

@@ -254,9 +254,9 @@ pytest -q 2>&1 | squeez "find the failing test"
 git log --oneline -50 | squeez "find the auth commit"
 ```
 
-`KRLabsOrg/verbatim-rag-modern-bert-v2` is a 150M ModernBERT span model
-trained on a multi-domain mix that includes Squeez tool-output. See
-[RESULTS.md](RESULTS.md) for the head-to-head with Squeez-2B.
+[`KRLabsOrg/verbatim-rag-modern-bert-v2`](https://huggingface.co/KRLabsOrg/verbatim-rag-modern-bert-v2)
+is a 150M ModernBERT span model trained on a multi-domain mix that includes
+Squeez tool-output — its model card has the head-to-head numbers.
 
 To train your own extractive model, see [TRAINING.md](TRAINING.md).
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "squeez"
-version = "0.1.3"
+version = "0.1.4"
 description = "Squeeze verbose LLM agent tool output down to only the relevant lines"
 readme = {file = "README.md", content-type = "text/markdown"}
 requires-python = ">=3.10"

squeez/__init__.py

@@ -1,6 +1,6 @@
 """Squeez — Squeeze verbose LLM agent tool output down to only the relevant evidence blocks."""
 
-__version__ = "0.1.3"
+__version__ = "0.1.4"
 
 
 def __getattr__(name: str):