alternatives retrival hinzugefügt

math-tutor/README.md

@@ -68,6 +68,6 @@ Retrieval settings:
 
 ## Testing
 python -m test.hint_test --chat-id draft_session_mlgmxxzc_avmjfb
-python -m test.vector_store_test --query "Was ist eine Teilmenge?" --k 8 --expand
+python -m test.retrieval_store_test --query "Was ist eine Teilmenge?" --k 8 --expand
 python -m test.math_intent_test --input "Integrate x^2" --input "Was ist 2+2?"
-python -m test.decision_test --chat-id draft_session_mlgmxxzc_avmjfb
+python -m test.decision_test --chat-id draft_session_mlgmxxzc_avmjfb
\ No newline at end of file

math-tutor/backend/.env-Example

 MATHPIX_APP_ID=""
 MATHPIX_APP_KEY=""
 
-OPENAI_BASE_URL=""
+OPENAI_BASE_URL="https://chat-ai.academiccloud.de/v1/"
 OPENAI_API_KEY=""
-OPENAI_CHAT_MODEL=""
-OPENAI_CHAT_TEMPERATURE=""
-OPENAI_EMBED_MODEL=""
+OPENAI_CHAT_MODEL="mistral-large-3-675b-instruct-2512"
+OPENAI_CHAT_TEMPERATURE="0.2"
+OPENAI_EMBED_MODEL="e5-mistral-7b-instruct"
 
-EMBEDDING_TYPE="" # "openai-like" or "sentence-transformers"
+EMBEDDING_TYPE="sentence-transformer" # "openai-like" or "sentence-transformer"
+EMBEDDING_DIM="512"
+
+SENTENCE_TRANSFORMER_MODEL="jinaai/jina-embeddings-v4"
+
+ORCHESTRATOR="tutor" # "tutor" or "qa"
+RETRIEVAL_IMPL="child" # "child" or "subsection"
 
 POSTGRES_URL=""
 
-OLLAMA_URL=""
-OLLAMA_MODEL= ""
-OLLAMA_TEMPERATURE=""
+OLLAMA_URL="http://localhost:11434"
+OLLAMA_MODEL= "ministral-3"
+OLLAMA_TEMPERATURE="0.2"
 
-FRONTEND_URL=""
+FRONTEND_URL="http://localhost:5173"
\ No newline at end of file

math-tutor/backend/app/config.py

@@ -18,6 +18,13 @@ class EmbeddingSettings(BaseModel):
 def get_orchestrator() -> str:
     return os.getenv("ORCHESTRATOR", "qa").lower()
 
+
+def get_retrieval_impl() -> str:
+    value = os.getenv("RETRIEVAL_IMPL", "child").strip().lower()
+    if value in {"child", "subsection"}:
+        return value
+    return "child"
+
 def get_embedding_settings() -> EmbeddingSettings:
     embedding_type = os.getenv("EMBEDDING_TYPE", "openai-like")
     if embedding_type == "sentence-transformer":

math-tutor/backend/app/deterministic_services/orchestrators/orchestrator_base.py

@@ -8,8 +8,8 @@ from app.deterministic_services import (
     Source,
     context_store,
     referenz_decoder,
+    retrieval_store,
     tool_logging,
-    vector_store,
 )
 from app.deterministic_services.embeddings import EmbeddingFactory
 
@@ -46,7 +46,7 @@ def extract_user_messages(messages: list[dict]) -> list[str]:
 def retrieve_context(query_text: str, pg_url: str | None = None) -> List[Source]:
     embedder = EmbeddingFactory.create(config.get_embedding_settings())
     url = pg_url or config.get_postgres_url()
-    return vector_store.retrieve(
+    return retrieval_store.retrieve(
         pg_url=url,
         embedder=embedder,
         query=query_text,

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

+from __future__ import annotations
+
+from typing import List
+
+import app.config as config
+from app.deterministic_services import vector_store, vector_store_subsection
+from app.deterministic_services.vector_store import EmbeddingLike, Source
+
+
+def _use_subsection_retrieval() -> bool:
+    return config.get_retrieval_impl() == "subsection"
+
+
+def retrieve(
+    pg_url: str,
+    embedder: EmbeddingLike,
+    query: str,
+    k: int = 4,
+    section_index: int | None = None,
+    subsection_index: int | None = None,
+    source_type_filter: list[str] | None = None,
+    expand_links: bool = True,
+    neighbor_expand: int = 0,
+) -> List[Source]:
+    if _use_subsection_retrieval():
+        return vector_store_subsection.retrieve(
+            pg_url=pg_url,
+            embedder=embedder,
+            query=query,
+            k=k,
+            section_index=section_index,
+            subsection_index=subsection_index,
+            source_type_filter=source_type_filter,
+            expand_links=expand_links,
+            neighbor_expand=neighbor_expand,
+        )
+
+    return vector_store.retrieve(
+        pg_url=pg_url,
+        embedder=embedder,
+        query=query,
+        k=k,
+        section_index=section_index,
+        subsection_index=subsection_index,
+        source_type_filter=source_type_filter,
+        expand_links=expand_links,
+        neighbor_expand=neighbor_expand,
+    )

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

+from __future__ import annotations
+
+from typing import Any, Dict, List, Optional
+
+import psycopg
+from pgvector import Vector
+from pgvector.psycopg import register_vector
+from psycopg.rows import dict_row
+
+from app.deterministic_services.vector_store import (
+    EmbeddingLike,
+    Source,
+    _retrivla_to_sources,
+    _row_to_retrieved,
+    embed_query,
+)
+
+
+def retrieve(
+    pg_url: str,
+    embedder: EmbeddingLike,
+    query: str,
+    k: int = 4,
+    section_index: Optional[int] = None,
+    subsection_index: Optional[int] = None,
+    source_type_filter: Optional[List[str]] = None,
+    expand_links: bool = False,
+    neighbor_expand: int = 0,
+) -> List[Source]:
+    # Parameters kept for drop-in compatibility with child-level retrieve.
+    _ = expand_links
+    _ = neighbor_expand
+
+    qvec = Vector(embed_query(embedder, query))
+
+    where = ["doc_type = ANY(%(sub_doc_types)s)"]
+    params: Dict[str, Any] = {
+        "qvec": qvec,
+        "k": k,
+        "sub_doc_types": ["subsection", "chapter"],
+    }
+
+    if section_index is not None:
+        where.append("section_index = %(section_index)s")
+        params["section_index"] = section_index
+
+    if subsection_index is not None:
+        where.append("subsection_index = %(subsection_index)s")
+        params["subsection_index"] = subsection_index
+
+    if source_type_filter:
+        where.append("source_type = ANY(%(source_type_filter)s)")
+        params["source_type_filter"] = source_type_filter
+
+    where_sql = " AND ".join(where)
+    sql = f"""
+    SELECT
+      uid, doc_type,
+      section_index, subsection_index, child_index,
+      section_title, subsection_title, title, source_type,
+      path, markdown,
+      1 - (embedding <=> %(qvec)s) AS score
+    FROM docs
+    WHERE {where_sql}
+    ORDER BY embedding <=> %(qvec)s
+    LIMIT %(k)s;
+    """
+
+    with psycopg.connect(pg_url, row_factory=dict_row) as conn:
+        register_vector(conn)
+        with conn.cursor() as cur:
+            cur.execute(sql, params)
+            rows = cur.fetchall()
+
+    subsections = [_row_to_retrieved(row, source_type="subsection") for row in rows]
+    return _retrivla_to_sources({"subsections_direct": subsections})

math-tutor/backend/scripts/retrieval_cli.py

@@ -18,7 +18,7 @@ if str(ROOT_DIR) not in sys.path:
     sys.path.insert(0, str(ROOT_DIR))
 
 from app.deterministic_services.embeddings import BaseEmbeddings, EmbeddingFactory
-from app.deterministic_services import vector_store
+from app.deterministic_services import retrieval_store, vector_store
 
 
 def build_embedder() -> BaseEmbeddings:
@@ -51,7 +51,7 @@ def cli_query(args: argparse.Namespace) -> None:
     pg_url = args.pg or os.getenv("POSTGRES_URL")
     if not pg_url:
         raise ValueError("Missing POSTGRES_URL")
-    result = vector_store.retrieve(
+    result = retrieval_store.retrieve(
         pg_url=pg_url,
         embedder=embedder,
         query=args.q,

math-tutor/backend/test/vector_store_test.py → math-tutor/backend/test/retrieval_store_test.py

 import argparse
 
 from app import config
-from app.deterministic_services.embeddings import EmbeddingFactory, OpenAILikeEmbeddings
-from app.deterministic_services import vector_store
-
-
-def _normalize_sources(result: object) -> list[vector_store.Source]:
-    if isinstance(result, list):
-        return result
-    if isinstance(result, dict):
-        groups = {k: v for k, v in result.items() if isinstance(v, list)}
-        return vector_store._retrivla_to_sources(groups)
-    raise TypeError("Unexpected retrieval result type")
+from app.deterministic_services.embeddings import EmbeddingFactory
+from app.deterministic_services import retrieval_store
 
 
 def main() -> None:
@@ -30,7 +21,7 @@ def main() -> None:
     pg_url = args.pg or config.get_postgres_url()
     embedder = EmbeddingFactory.create(config.get_embedding_settings())
 
-    result = vector_store.retrieve(
+    sources = retrieval_store.retrieve(
         pg_url=pg_url,
         embedder=embedder,
         query=args.query,
@@ -42,7 +33,6 @@ def main() -> None:
         neighbor_expand=args.neighbor_expand,
     )
 
-    sources = _normalize_sources(result)
     if not sources:
         print("Keine Quellen gefunden.")
         return