Add rapidfuzz

.changeset/embedding-clustering-rapidfuzz.md

@@ -0,0 +1,6 @@
+---
+'@platforma-open/milaboratories.runenv-python-3.12.10-embedding-clustering': minor
+'@platforma-open/milaboratories.runenv-python-3': minor
+---
+
+Add `rapidfuzz` 3.14.6 to the Embedding Clustering run environment, for string distances such as normalized Levenshtein. It has no required dependencies and is installed with `noDeps`, like the rest of the pinned set.

python-3.12.10-embedding-clustering/README.md

@@ -1,4 +1,4 @@
-Python 3.12.10 run environment for the Embedding Clustering block (`embedding-clustering`): `fast_hdbscan` and `pynndescent` on `numba`, plus `scikit-learn` (PCA, medoid distances), `polars` and `pyarrow` (Parquet loading and output) and `pandas`.
+Python 3.12.10 run environment for the Embedding Clustering block (`embedding-clustering`): `fast_hdbscan` and `pynndescent` on `numba`, plus `scikit-learn` (PCA, medoid distances), `polars` and `pyarrow` (Parquet loading and output), `pandas` and `rapidfuzz` (string distances such as normalized Levenshtein).
 
 The set is pinned flat and matches the block's `requirements.txt`. The block installs it with `pip --no-index --find-links <runenv packages>`, so each pin in the block must exist here as a wheel. Change the pins here and in the block together.
 

python-3.12.10-embedding-clustering/config.json

@@ -14,7 +14,8 @@
       "polars-lts-cpu==1.33.1",
       "pyarrow==24.0.0",
       "pynndescent==0.6.0",
-      "fast_hdbscan==0.3.2"
+      "fast_hdbscan==0.3.2",
+      "rapidfuzz==3.14.6"
     ],
     "noDeps": [
       "numpy",
@@ -32,7 +33,8 @@
       "numba",
       "llvmlite",
       "pynndescent",
-      "fast_hdbscan"
+      "fast_hdbscan",
+      "rapidfuzz"
     ],
     "skip": {},
     "overrides": {},