superlocalmemory 3.4.55 → 3.4.56

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "superlocalmemory",
-  "version": "3.4.55",
+  "version": "3.4.56",
   "description": "Information-geometric agent memory with mathematical guarantees. 4-channel retrieval, Fisher-Rao similarity, zero-LLM mode, EU AI Act compliant. Works with Claude, Cursor, Windsurf, and 17+ AI tools.",
   "keywords": [
     "ai-memory",

package/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "superlocalmemory"
-version = "3.4.55"
+version = "3.4.56"
 description = "Information-geometric agent memory with mathematical guarantees"
 readme = "README.md"
 license = {text = "AGPL-3.0-or-later"}

package/src/superlocalmemory/__init__.py

@@ -3,7 +3,7 @@
 import os
 os.environ.setdefault("KMP_DUPLICATE_LIB_OK", "TRUE")
 
-__version__ = "3.4.55"
+__version__ = "3.4.56"
 
 _REQUIRED_VERSIONS = {
     "sentence_transformers": "5.3.0",

package/src/superlocalmemory/core/engine_wiring.py

@@ -78,20 +78,21 @@ def init_embedder(config: SLMConfig) -> Any | None:
     emb_cfg = config.embedding
     provider = emb_cfg.provider
 
-    # All modes use sentence-transformers subprocess as primary so the
-    # embedding space matches stored vectors. Ollama is fallback only —
-    # Ollama's nomic-embed-text and sentence-transformers nomic-embed-text-v1.5
-    # produce different vectors, so mixing them against an ST-indexed
-    # corpus degrades semantic recall quality.
+    # v3.4.55: When provider is ollama, use Ollama's embedding API as
+    # PRIMARY. The stored vectors were created by Ollama's nomic-embed-text;
+    # sentence-transformers' nomic-embed-text-v1.5 produces different vectors.
+    # Mixing them degrades semantic recall. ST subprocess is the fallback,
+    # not the primary — this also eliminates the 2-3 minute cold-start
+    # where the ST model loads from disk (~2.1 GB) on every daemon restart.
     if provider == "ollama":
-        st_emb = _try_service_embedder(EmbeddingService, emb_cfg)
-        if st_emb is not None:
-            logger.info("Using sentence-transformers subprocess (matches stored embedding space)")
-            return st_emb
         result = _try_ollama_embedder(emb_cfg)
         if result is not None:
-            logger.warning("sentence-transformers unavailable; falling back to Ollama (semantic quality may degrade)")
+            logger.info("Using Ollama embeddings (nomic-embed-text, local)")
             return result
+        st_emb = _try_service_embedder(EmbeddingService, emb_cfg)
+        if st_emb is not None:
+            logger.warning("Ollama unavailable; falling back to sentence-transformers subprocess")
+            return st_emb
         return None
 
     # --- V3.4.24: Explicit OpenAI-compatible provider ---

package/src/superlocalmemory.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: superlocalmemory
-Version: 3.4.55
+Version: 3.4.56
 Summary: Information-geometric agent memory with mathematical guarantees
 Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
 License: AGPL-3.0-or-later