sampler-cli 0.4.2__tar.gz → 0.4.3__tar.gz

{sampler_cli-0.4.2 → sampler_cli-0.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sampler-cli
-Version: 0.4.2
+Version: 0.4.3
 Summary: Token-efficient CLI for indexing and searching code symbols (Python-first, designed for minimal LLM/agent context size)
 Author: Samuel Ignacio Carmona Rodriguez
 License: MIT
@@ -52,7 +52,7 @@ Dynamic: license-file
 
 Token-efficient CLI for indexing and searching code symbols across multiple projects.
 
-Current version: 0.4.2
+Current version: 0.4.3
 
 Designed for humans and agents: compact default output, short paths, and low-noise symbol views.
 

{sampler_cli-0.4.2 → sampler_cli-0.4.3}/README.md

@@ -2,7 +2,7 @@
 
 Token-efficient CLI for indexing and searching code symbols across multiple projects.
 
-Current version: 0.4.2
+Current version: 0.4.3
 
 Designed for humans and agents: compact default output, short paths, and low-noise symbol views.
 

{sampler_cli-0.4.2 → sampler_cli-0.4.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "sampler-cli"
-version = "0.4.2"
+version = "0.4.3"
 description = "Token-efficient CLI for indexing and searching code symbols (Python-first, designed for minimal LLM/agent context size)"
 readme = "README.md"
 license = { text = "MIT" }

{sampler_cli-0.4.2 → sampler_cli-0.4.3}/src/sampler/__init__.py

@@ -1,3 +1,3 @@
 __all__ = ["__version__"]
 
-__version__ = "0.4.2"
+__version__ = "0.4.3"

{sampler_cli-0.4.2 → sampler_cli-0.4.3}/src/sampler/query/semantic.py

@@ -92,7 +92,7 @@ class SemanticEngine:
         scored.sort(key=lambda pair: pair[0], reverse=True)
         return scored[:pool]
 
-    def _hash_scored_candidates(self, query: str, rows: list[dict], project_name: str | None, pool: int):
+    def _hash_scored_candidates(self, query: str, rows: list[dict], pool: int):
         """Last-resort backend: deterministic hash fingerprints (no ML, always available)."""
         try:
             import numpy as np
@@ -101,18 +101,9 @@ class SemanticEngine:
                 "Semantic search requires numpy. Install with: pip install 'sampler-cli[semantic]'"
             ) from exc
 
-        embeddings = self.db.get_embeddings_for_project(project_name) if project_name else []
+        # ponytail: always derive on-the-fly for hash (cheap+deterministic); precomp only for real providers (avoids all cross-dim crashes when provider switched)
         query_vec = self.embedder.hash_fingerprint_vector(query)
 
-        if embeddings:
-            scored = []
-            for row in embeddings:
-                vec = np.frombuffer(row["vector"], dtype="float32")
-                sim = float(np.dot(query_vec, vec))
-                scored.append((sim, row))
-            scored.sort(key=lambda pair: pair[0], reverse=True)
-            return scored[:pool]
-
         if not rows:
             return []
 
@@ -139,7 +130,7 @@ class SemanticEngine:
             return tfidf
 
         # 3. Hash fingerprint (always works, deterministic local fallback)
-        return self._hash_scored_candidates(query, rows, project_name, pool)
+        return self._hash_scored_candidates(query, rows, pool)
 
     def semantic_search(self, query: str, project_name: str | None = None, limit: int = 10) -> list[dict]:
         """Pure cosine-similarity search over stored symbol embeddings, no graph/text signals."""

{sampler_cli-0.4.2 → sampler_cli-0.4.3}/src/sampler_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sampler-cli
-Version: 0.4.2
+Version: 0.4.3
 Summary: Token-efficient CLI for indexing and searching code symbols (Python-first, designed for minimal LLM/agent context size)
 Author: Samuel Ignacio Carmona Rodriguez
 License: MIT
@@ -52,7 +52,7 @@ Dynamic: license-file
 
 Token-efficient CLI for indexing and searching code symbols across multiple projects.
 
-Current version: 0.4.2
+Current version: 0.4.3
 
 Designed for humans and agents: compact default output, short paths, and low-noise symbol views.