harden reset system

.gitignore

@@ -1,6 +1,6 @@
 .ddev
 .idea
-/var/knowledge
+/var/knowledge/*
 /.venv
 /src/Vector/vector.index
 /src/Vector/vector_meta.json

config/services.yaml

@@ -7,7 +7,7 @@ parameters:
 
   mto.index.chunk_size: 800
   mto.index.chunk_overlap: 100
-  mto.index.embedding_model: 'nomic-embed-text'
+  mto.index.embedding_model: 'all-MiniLM-L6-v2' #nomic-embed-text
   mto.index.embedding_dimension: 768
   mto.index.scoring_version: 1
 

src/Ingest/IngestFlow.php

@@ -6,7 +6,6 @@ namespace App\Ingest;
 
 use App\Entity\DocumentVersion;
 use App\Index\IndexMetaManager;
-use App\Index\IndexStructureChangedException;
 use App\Knowledge\ChunkManager;
 use App\Knowledge\Ingest\KnowledgeIngestService;
 use App\Vector\VectorIndexBuilder;

src/Vector/VectorIndexBuilder.php

@@ -84,14 +84,12 @@ final class VectorIndexBuilder
         // Wir erwarten/standardisieren (ab jetzt) CLI-Args:
         //   --index <path-to-index.ndjson>
         //   --out   <path-to-vector.index.tmp>
-        //
-        // Falls dein Python-Script aktuell andere Args hat,
-        // passen wir es im nächsten Schritt konsistent an.
         $cmd = [
             $this->pythonBin,
             $this->scriptPath,
             '--index', $this->indexNdjsonPath,
             '--out', $tmpVectorIndexPath,
+            '--model', 'all-MiniLM-L6-v2',
         ];
 
         $process = new Process($cmd);