box_hint entfernt

math-tutor/README.md

@@ -64,4 +64,4 @@ Retrieval settings:
 
 ## Testing
 python -m test.hint_test ...
-pyhton -m test.vector_store_test ...
\ No newline at end of file
+python -m test.vector_store_test --query "Was ist eine Teilmenge?" --k 8 --expand

math-tutor/backend/app/deterministic_services/vector_store.py

@@ -131,7 +131,6 @@ CREATE TABLE IF NOT EXISTS docs (
   subsection_title TEXT NULL,
   title TEXT NULL,
   type TEXT NULL,
-  box_hint TEXT NULL,
 
   path TEXT NOT NULL,
   markdown TEXT NOT NULL,
@@ -152,12 +151,12 @@ UPSERT_SQL = """
 INSERT INTO docs (
   uid, doc_type,
   section_index, subsection_index, child_index,
-  section_title, subsection_title, title, type, box_hint,
+  section_title, subsection_title, title, type,
   path, markdown, embedding
 ) VALUES (
   %(uid)s, %(doc_type)s,
   %(section_index)s, %(subsection_index)s, %(child_index)s,
-  %(section_title)s, %(subsection_title)s, %(title)s, %(type)s, %(box_hint)s,
+  %(section_title)s, %(subsection_title)s, %(title)s, %(type)s,
   %(path)s, %(markdown)s, %(embedding)s
 )
 ON CONFLICT (uid) DO UPDATE SET
@@ -169,7 +168,6 @@ ON CONFLICT (uid) DO UPDATE SET
   subsection_title = EXCLUDED.subsection_title,
   title = EXCLUDED.title,
   type = EXCLUDED.type,
-  box_hint = EXCLUDED.box_hint,
   path = EXCLUDED.path,
   markdown = EXCLUDED.markdown,
   embedding = EXCLUDED.embedding
@@ -213,7 +211,6 @@ def upsert_docs(pg_url: str, docs: List[DocRecord], embeddings: List[List[float]
                 "subsection_title": m.get("subsection_title"),
                 "title": m.get("title"),
                 "type": m.get("type"),
-                "box_hint": m.get("box_hint"),
                 "path": doc.path,
                 "markdown": doc.markdown,
                 "embedding": Vector(emb),
@@ -294,7 +291,6 @@ def _row_to_retrieved(row: Dict[str, Any]) -> Retrieved:
         "subsection_title": row["subsection_title"],
         "title": row["title"],
         "type": row["type"],
-        "box_hint": row["box_hint"],
         "path": row["path"],
     }
     return Retrieved(
@@ -360,7 +356,7 @@ def retrieve(
     SELECT
       uid, doc_type,
       section_index, subsection_index, child_index,
-      section_title, subsection_title, title, type, box_hint,
+      section_title, subsection_title, title, type,
       path, markdown,
       1 - (embedding <=> %(qvec)s) AS score
     FROM docs
@@ -402,7 +398,7 @@ def retrieve(
                         SELECT
                           d.uid, d.doc_type,
                           d.section_index, d.subsection_index, d.child_index,
-                          d.section_title, d.subsection_title, d.title, d.type, d.box_hint,
+                          d.section_title, d.subsection_title, d.title, d.type,
                           d.path, d.markdown,
                           1.0 AS score
                         FROM docs d
@@ -419,7 +415,7 @@ def retrieve(
                         SELECT
                           uid, doc_type,
                           section_index, subsection_index, child_index,
-                          section_title, subsection_title, title, type, box_hint,
+                          section_title, subsection_title, title, type,
                           path, markdown,
                           1.0 AS score
                         FROM docs
@@ -435,7 +431,7 @@ def retrieve(
                         SELECT
                           d.uid, d.doc_type,
                           d.section_index, d.subsection_index, d.child_index,
-                          d.section_title, d.subsection_title, d.title, d.type, d.box_hint,
+                          d.section_title, d.subsection_title, d.title, d.type,
                           d.path, d.markdown,
                           1 - (d.embedding <=> %(qvec)s) AS score
                         FROM docs d
@@ -478,7 +474,7 @@ def retrieve(
                                 SELECT
                                   d.uid, d.doc_type,
                                   d.section_index, d.subsection_index, d.child_index,
-                                  d.section_title, d.subsection_title, d.title, d.type, d.box_hint,
+                                  d.section_title, d.subsection_title, d.title, d.type,
                                   d.path, d.markdown,
                                   0.9 AS score
                                 FROM docs d