Merge branch 'dev' into 'main'

Dev

See merge request kantz/tutor_react!23

math-tutor/.gitignore

@@ -4,6 +4,7 @@ venv/
 .env-openai
 .env-ollama
 .env-gwdg
+.env-stochastik
 __pycache__/
 drawings/
 sources/

math-tutor/README.md

@@ -10,7 +10,7 @@
 if you want to use the task mode you need :
 
 - task folder in sources with your tasks
-- `_subsection_map.json`.
+- `_subsection_map.yaml`.
 
 ## Setup
 
@@ -217,7 +217,7 @@ python -m test.hint_test --chat-id draft_session_mlgmxxzc_avmjfb
 
 python -m test.retrieval_store_test --query "Was ist eine Teilmenge?" --k 8 --expand
 
-python -m test.retrieval_store_test --query "Was ist eine Teilmenge?" --k 3 --subsections 1:1:1
+python -m test.retrieval_store_test --query "Was ist eine Teilmenge?" --k 3 --subsections 1-1-1
 
 python -m test.math_intent_test --input "Integrate x^2" --input "Was ist 2+2?"
 

math-tutor/backend/.env-example

@@ -9,6 +9,7 @@ FRONTEND_URL="http://frontend:3000"
 
 ORCHESTRATOR="task" # "tutor", "task" or "qa"
 RETRIEVAL_IMPL="child" # "child" or "subsection"
+TASK_FOLDER="tasks"
 LLM_PROVIDER="gwdg" # "openai", "gwdg", "mistral", or "ollama"
 
 EMBEDDING_PROVIDER="sentence-transformer" # "sentence-transformer", "openai", or "gwdg"

math-tutor/backend/app/config.py

 import os
 from dataclasses import dataclass
+from pathlib import Path
 from dotenv import load_dotenv
 
 from pydantic import BaseModel
@@ -7,6 +8,9 @@ from typing import Optional
 
 load_dotenv()
 
+BACKEND_ROOT = Path(__file__).resolve().parents[1]
+SOURCES_ROOT = BACKEND_ROOT / "sources"
+
 SUPPORTED_LLM_PROVIDERS = {"openai", "gwdg", "mistral", "ollama"}
 SUPPORTED_EMBEDDING_PROVIDERS = {"sentence-transformer", "openai", "gwdg"}
 
@@ -30,6 +34,17 @@ def get_retrieval_impl() -> str:
     return "child"
 
 
+def get_task_folder() -> Path:
+    value = os.getenv("TASK_FOLDER", "tasks").strip()
+    if not value:
+        value = "tasks"
+
+    folder = Path(value)
+    if folder.is_absolute():
+        return folder
+    return SOURCES_ROOT / folder
+
+
 def get_llm_provider() -> str:
     value = os.getenv("LLM_PROVIDER")
     if value is None or not value.strip():

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

@@ -41,7 +41,7 @@ def _ensure_context_task_fields(state: base.ChatState, query_text: str) -> tuple
 
     base.append_tool_log(
         state.tool_log,
-        "task_json_selected",
+        "task_yaml_selected",
         {"tasks_dir": str(task_catalog.TASKS_DIR)},
         {
             "file": task_file.get("_path", ""),

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

 from __future__ import annotations
 
-import json
 import re
 import unicodedata
 from functools import lru_cache
 from pathlib import Path
-from urllib.parse import quote
 from typing import Any
+from urllib.parse import quote
+
+import yaml
 
+from app import config
 from app.deterministic_services import context_store
 from app.deterministic_services.vector_store import parse_markdown_with_frontmatter
 
-TASKS_DIR = Path(__file__).resolve().parents[2] / "sources" / "tasks"
+TASKS_DIR = config.get_task_folder()
 TASK_IMAGES_DIR = TASKS_DIR / "images"
 SOURCES_DIR = Path(__file__).resolve().parents[2] / "sources"
-SUBSECTION_MAP_PATH = TASKS_DIR / "_subsection_map.json"
+SUBSECTION_MAP_PATH = TASKS_DIR / "_subsection_map.yaml"
 TASK_ASSET_URL_PREFIX = "/api/tasks/assets"
 
 def _normalize_text(value: str) -> str:
@@ -34,7 +36,7 @@ def _normalize_subsection_key(value: str) -> str:
 
 
 def _parse_subsection_ref(value: str) -> tuple[int, int, int] | None:
-    match = re.match(r"^\s*(\d+)\s*[:.]\s*(\d+)\s*[:.]\s*(\d+)\s*$", str(value))
+    match = re.match(r"^\s*(\d+)\s*[-.]\s*(\d+)\s*[-.]\s*(\d+)\s*$", str(value))
     if not match:
         return None
     return int(match.group(1)), int(match.group(2)), int(match.group(3))
@@ -44,7 +46,7 @@ def load_subsection_map(path: Path = SUBSECTION_MAP_PATH) -> dict[str, tuple[int
     if not path.exists():
         return {}
     try:
-        content = json.loads(path.read_text(encoding="utf-8"))
+        content = yaml.safe_load(path.read_text(encoding="utf-8"))
     except Exception:
         return {}
     if not isinstance(content, dict):
@@ -186,23 +188,144 @@ def _match_score(query_text: str, candidate_text: str) -> int:
     return len(query_tokens.intersection(candidate_tokens))
 
 
+def _slugify(value: str) -> str:
+    normalized = unicodedata.normalize("NFKD", str(value))
+    ascii_value = normalized.encode("ascii", "ignore").decode("ascii")
+    lowered = ascii_value.strip().lower()
+    collapsed = re.sub(r"[^a-z0-9]+", "-", lowered)
+    collapsed = re.sub(r"-{2,}", "-", collapsed)
+    return collapsed.strip("-")
+
+
+def _normalize_block_images(images_raw: list[dict[str, str]]) -> list[dict[str, str]]:
+    normalized_images: list[dict[str, str]] = []
+    for item in images_raw:
+        if not isinstance(item, dict):
+            continue
+        src = str(item.get("src", "")).strip().replace("\\", "/")
+        description = str(item.get("description", "") or item.get("alt", "")).strip()
+        if not src or not description:
+            continue
+        if src.startswith("/") or Path(src).is_absolute():
+            continue
+        normalized_path = Path(src)
+        if any(part in {"", ".", ".."} for part in normalized_path.parts):
+            continue
+        encoded_src = "/".join(quote(part) for part in normalized_path.parts)
+        normalized_images.append(
+            {
+                "src": f"{TASK_ASSET_URL_PREFIX}/{encoded_src}",
+                "description": description,
+            }
+        )
+    return normalized_images
+
+
+def _extract_text_and_images(blocks: Any) -> tuple[list[str], list[dict[str, str]]]:
+    if not isinstance(blocks, list):
+        return [], []
+
+    text_parts: list[str] = []
+    images: list[dict[str, str]] = []
+    for block in blocks:
+        if not isinstance(block, dict):
+            continue
+        block_type = str(block.get("type", "")).strip().lower()
+        if block_type == "text":
+            text = str(block.get("text", "")).strip()
+            if text:
+                text_parts.append(text)
+            continue
+        if block_type == "image":
+            src = str(block.get("src", "")).strip()
+            alt = str(block.get("alt", "")).strip()
+            if src and alt:
+                images.append({"src": src, "description": alt})
+    return text_parts, images
+
+
+def _normalize_yaml_task_entry(task_entry: Any, position: int) -> dict[str, Any] | None:
+    if not isinstance(task_entry, dict):
+        return None
+
+    statement_parts, statement_images = _extract_text_and_images(task_entry.get("aufgabe", []))
+    hint_parts, hint_images = _extract_text_and_images(task_entry.get("hinweise", []))
+    solution_parts, _ = _extract_text_and_images(task_entry.get("loesung", []))
+
+    return {
+        "id": str(position).zfill(2),
+        "statement": "\n".join(statement_parts).strip(),
+        "hints": hint_parts,
+        "solution": "\n".join(solution_parts).strip(),
+        "images": statement_images + hint_images,
+    }
+
+
+def _normalize_exercise_topics(raw_topics: Any) -> list[str]:
+    if not isinstance(raw_topics, list):
+        return []
+
+    topics: list[str] = []
+    for item in raw_topics:
+        topic_id = ""
+        if isinstance(item, dict):
+            topic_id = str(item.get("topic-id", "")).strip()
+        elif isinstance(item, str):
+            topic_id = item.strip()
+        if topic_id:
+            topics.append(topic_id)
+    return topics
+
+
+def _normalize_yaml_task_file(content: Any, path: Path) -> dict[str, Any] | None:
+    if not isinstance(content, dict):
+        return None
+    if not isinstance(content.get("tasks"), list):
+        return None
+
+    title = str(content.get("titel", "")).strip()
+    intro = str(content.get("description", "")).strip()
+    file_id = str(content.get("id", "")).strip() or path.stem
+    slug = str(content.get("slug", "")).strip() or _slugify(title or file_id)
+    level = str(content.get("level", "")).strip()
+    subsections = _normalize_exercise_topics(content.get("exercise_topic", []))
+
+    if not title or not file_id or not slug or not level:
+        return None
+
+    tasks: list[dict[str, Any]] = []
+    for index, raw_task in enumerate(content.get("tasks", []), start=1):
+        normalized_task = _normalize_yaml_task_entry(raw_task, index)
+        if normalized_task is None:
+            return None
+        tasks.append(normalized_task)
+
+    return {
+        "_path": str(path),
+        "_file_id": file_id,
+        "title": title,
+        "intro": intro,
+        "slug": slug,
+        "level": level,
+        "subsections": subsections,
+        "tasks": tasks,
+    }
+
+
 def load_task_files(tasks_dir: Path = TASKS_DIR) -> list[dict[str, Any]]:
     if not tasks_dir.exists():
         return []
 
     loaded: list[dict[str, Any]] = []
-    for path in sorted(tasks_dir.glob("*.json")):
+    for path in sorted([*tasks_dir.glob("*.yaml"), *tasks_dir.glob("*.yml")]):
         try:
-            content = json.loads(path.read_text(encoding="utf-8"))
+            content = yaml.safe_load(path.read_text(encoding="utf-8"))
         except Exception:
             continue
-        if not isinstance(content, dict):
+        normalized = _normalize_yaml_task_file(content, path)
+        if normalized is None:
             continue
-        if not isinstance(content.get("tasks"), list):
-            continue
-        content["_path"] = str(path)
-        content["_file_id"] = path.stem
-        loaded.append(content)
+        loaded.append(normalized)
     return loaded
 
 
@@ -226,28 +349,7 @@ def _normalize_task_images(task_entry: dict[str, Any]) -> list[dict[str, str]]:
     images_raw = task_entry.get("images", [])
     if not isinstance(images_raw, list):
         return []
-
-    normalized_images: list[dict[str, str]] = []
-    for item in images_raw:
-        if not isinstance(item, dict):
-            continue
-        src = str(item.get("src", "")).strip().replace("\\", "/")
-        description = str(item.get("description", "")).strip()
-        if not src or not description:
-            continue
-        if src.startswith("/") or Path(src).is_absolute():
-            continue
-        normalized_path = Path(src)
-        if any(part in {"", ".", ".."} for part in normalized_path.parts):
-            continue
-        encoded_src = "/".join(quote(part) for part in normalized_path.parts)
-        normalized_images.append(
-            {
-                "src": f"{TASK_ASSET_URL_PREFIX}/{encoded_src}",
-                "description": description,
-            }
-        )
-    return normalized_images
+    return _normalize_block_images(images_raw)
 
 
 def _build_task_payload(

math-tutor/backend/scripts/convert_tasks_json_to_yaml.py

+from __future__ import annotations
+
+import argparse
+import json
+import shutil
+import unicodedata
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+BACKEND_ROOT = Path(__file__).resolve().parents[1]
+TASKS_DIR = BACKEND_ROOT / "sources" / "tasks"
+TASK_IMAGES_DIR = TASKS_DIR / "images"
+DEFAULT_LEVEL = "Uebungen"
+
+
+def slugify(value: str) -> str:
+    normalized = unicodedata.normalize("NFKD", str(value))
+    ascii_value = normalized.encode("ascii", "ignore").decode("ascii")
+    lowered = ascii_value.strip().lower()
+    collapsed = []
+    last_was_dash = False
+    for char in lowered:
+        if char.isalnum():
+            collapsed.append(char)
+            last_was_dash = False
+        elif not last_was_dash:
+            collapsed.append("-")
+            last_was_dash = True
+    return "".join(collapsed).strip("-")
+
+
+def load_json_task(path: Path) -> dict[str, Any]:
+    payload = json.loads(path.read_text(encoding="utf-8"))
+    if not isinstance(payload, dict):
+        raise ValueError(f"{path.name}: expected object")
+    if not isinstance(payload.get("tasks"), list):
+        raise ValueError(f"{path.name}: expected tasks list")
+    return payload
+
+
+def _build_topic_entries(subsections: Any) -> list[dict[str, str]]:
+    if not isinstance(subsections, list):
+        return []
+    topics: list[dict[str, str]] = []
+    for item in subsections:
+        topic_id = str(item).strip()
+        if topic_id:
+            topics.append({"topic-id": topic_id})
+    return topics
+
+
+def _build_text_block(text: str) -> dict[str, str]:
+    return {"type": "text", "text": text}
+
+
+def _build_image_block(src: str, description: str) -> dict[str, str]:
+    return {"type": "image", "src": src, "alt": description}
+
+
+def convert_task_payload(payload: dict[str, Any], source_file_id: str, image_path_map: dict[str, str]) -> dict[str, Any]:
+    title = str(payload.get("title", "")).strip()
+    intro = str(payload.get("intro", "")).strip()
+    slug = slugify(title or source_file_id)
+
+    converted_tasks: list[dict[str, Any]] = []
+    for raw_task in payload.get("tasks", []):
+        if not isinstance(raw_task, dict):
+            continue
+
+        aufgabe: list[dict[str, str]] = []
+        statement = str(raw_task.get("statement", "")).strip()
+        if statement:
+            aufgabe.append(_build_text_block(statement))
+
+        for raw_image in raw_task.get("images", []) if isinstance(raw_task.get("images"), list) else []:
+            if not isinstance(raw_image, dict):
+                continue
+            old_src = str(raw_image.get("src", "")).strip().replace("\\", "/")
+            description = str(raw_image.get("description", "")).strip()
+            new_src = image_path_map.get(old_src, old_src)
+            if old_src and description:
+                aufgabe.append(_build_image_block(new_src, description))
+
+        hinweise: list[dict[str, str]] = []
+        for hint in raw_task.get("hints", []) if isinstance(raw_task.get("hints"), list) else []:
+            hint_text = str(hint).strip()
+            if hint_text:
+                hinweise.append(_build_text_block(hint_text))
+
+        loesung_text = str(raw_task.get("solution", "")).strip()
+        loesung = [_build_text_block(loesung_text)] if loesung_text else []
+
+        converted_tasks.append(
+            {
+                "aufgabe": aufgabe,
+                "hinweise": hinweise,
+                "loesung": loesung,
+            }
+        )
+
+    return {
+        "id": source_file_id,
+        "titel": title,
+        "level": DEFAULT_LEVEL,
+        "slug": slug,
+        "description": intro,
+        "exercise_topic": _build_topic_entries(payload.get("subsections", [])),
+        "tasks": converted_tasks,
+    }
+
+
+def build_image_path_map(payload: dict[str, Any], slug: str) -> dict[str, str]:
+    mapping: dict[str, str] = {}
+    used_targets: set[str] = set()
+
+    for raw_task in payload.get("tasks", []):
+        if not isinstance(raw_task, dict):
+            continue
+        for raw_image in raw_task.get("images", []) if isinstance(raw_task.get("images"), list) else []:
+            if not isinstance(raw_image, dict):
+                continue
+            old_src = str(raw_image.get("src", "")).strip().replace("\\", "/")
+            if not old_src:
+                continue
+            basename = Path(old_src).name
+            stem = Path(basename).stem
+            suffix = Path(basename).suffix
+            target = f"{slug}/{basename}"
+            counter = 2
+            while target in used_targets:
+                target = f"{slug}/{stem}-{counter}{suffix}"
+                counter += 1
+            used_targets.add(target)
+            mapping[old_src] = target
+    return mapping
+
+
+def filter_existing_image_path_map(image_path_map: dict[str, str], images_root: Path) -> tuple[dict[str, str], list[str]]:
+    existing: dict[str, str] = {}
+    missing: list[str] = []
+    for old_src, new_src in image_path_map.items():
+        source_path = images_root / old_src
+        if source_path.exists():
+            existing[old_src] = new_src
+        else:
+            missing.append(old_src)
+    return existing, missing
+
+
+def move_task_images(image_path_map: dict[str, str], images_root: Path, dry_run: bool = False) -> list[str]:
+    moved: list[str] = []
+    for old_src, new_src in image_path_map.items():
+        source_path = images_root / old_src
+        target_path = images_root / new_src
+        if source_path.resolve() == target_path.resolve():
+            continue
+        if dry_run:
+            moved.append(f"{source_path} -> {target_path}")
+            continue
+        target_path.parent.mkdir(parents=True, exist_ok=True)
+        shutil.move(str(source_path), str(target_path))
+        moved.append(f"{source_path} -> {target_path}")
+    return moved
+
+
+def remove_empty_dirs(root: Path) -> None:
+    for path in sorted(root.rglob("*"), reverse=True):
+        if path.is_dir():
+            try:
+                next(path.iterdir())
+            except StopIteration:
+                path.rmdir()
+
+
+def build_output_yaml_path(tasks_dir: Path, slug: str, source_file_id: str, used_names: set[str]) -> Path:
+    candidates = [slug, slugify(source_file_id)]
+    for candidate in candidates:
+        if candidate and candidate not in used_names:
+            used_names.add(candidate)
+            return tasks_dir / f"{candidate}.yaml"
+
+    base_name = slug or slugify(source_file_id) or "task"
+    counter = 2
+    candidate = f"{base_name}-{counter}"
+    while candidate in used_names:
+        counter += 1
+        candidate = f"{base_name}-{counter}"
+    used_names.add(candidate)
+    return tasks_dir / f"{candidate}.yaml"
+
+
+def convert_file(
+    json_path: Path,
+    output_path: Path | None = None,
+    images_root: Path | None = None,
+    dry_run: bool = False,
+) -> tuple[Path, list[str], list[str]]:
+    payload = load_json_task(json_path)
+    source_file_id = json_path.stem
+    slug = slugify(str(payload.get("title", "")).strip() or source_file_id)
+    yaml_path = output_path or json_path.with_name(f"{slug}.yaml")
+    resolved_images_root = images_root or TASK_IMAGES_DIR
+    image_path_map = build_image_path_map(payload, slug)
+    existing_image_map, missing_images = filter_existing_image_path_map(image_path_map, resolved_images_root)
+    converted = convert_task_payload(payload, source_file_id, existing_image_map)
+
+    moves = move_task_images(existing_image_map, resolved_images_root, dry_run=dry_run)
+
+    if not dry_run:
+        yaml_path.write_text(
+            yaml.safe_dump(converted, sort_keys=False, allow_unicode=True),
+            encoding="utf-8",
+        )
+        json_path.unlink()
+    return yaml_path, moves, missing_images
+
+
+def convert_all(tasks_dir: Path = TASKS_DIR, dry_run: bool = False) -> tuple[list[Path], list[str]]:
+    converted_paths: list[Path] = []
+    missing_images: list[str] = []
+    images_root = tasks_dir / "images"
+    used_names: set[str] = set()
+    for json_path in sorted(tasks_dir.glob("*.json")):
+        if json_path.name == "_subsection_map.json":
+            continue
+        payload = load_json_task(json_path)
+        source_file_id = json_path.stem
+        slug = slugify(str(payload.get("title", "")).strip() or source_file_id)
+        yaml_path = build_output_yaml_path(tasks_dir, slug, source_file_id, used_names)
+        yaml_path, _, missing = convert_file(
+            json_path,
+            output_path=yaml_path,
+            images_root=images_root,
+            dry_run=dry_run,
+        )
+        converted_paths.append(yaml_path)
+        missing_images.extend(missing)
+    if not dry_run and images_root.exists():
+        remove_empty_dirs(images_root)
+    return converted_paths, missing_images
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Convert backend task sources from JSON to YAML.")
+    parser.add_argument("--dry-run", action="store_true", help="Preview conversion without writing files.")
+    args = parser.parse_args()
+
+    converted_paths, missing_images = convert_all(dry_run=args.dry_run)
+    print(f"Converted {len(converted_paths)} task files.")
+    if missing_images:
+        print(f"Missing images preserved with original src ({len(missing_images)}):")
+        for item in missing_images:
+            print(f"- {item}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

math-tutor/backend/scripts/generate_socratic_chats.py

@@ -144,7 +144,7 @@ def _subsection_key(subsection: MarkdownDoc) -> str:
 
 
 def _format_ref(ref: SubsectionRef) -> str:
-    return f"{ref[0]}:{ref[1]}:{ref[2]}"
+    return f"{ref[0]}-{ref[1]}-{ref[2]}"
 
 
 def build_llm_sources_text(subsection: MarkdownDoc, children: list[MarkdownDoc]) -> str:

math-tutor/backend/test/task_catalog_socratic_test.py

@@ -99,11 +99,11 @@ class TaskCatalogSocraticTest(unittest.TestCase):
             return_value={
                 "quadratische gleichungen": {
                     "subsection": "Quadratische Gleichungen",
-                    "index": "1:3:3",
+                    "index": "1-3-3",
                 },
                 "mengen": {
                     "subsection": "Mengen",
-                    "index": "1:1:1",
+                    "index": "1-1-1",
                 },
             },
         ), patch(

math-tutor/backend/test/task_yaml_migration_test.py

+from __future__ import annotations
+
+import json
+import os
+import shutil
+import tempfile
+import unittest
+from pathlib import Path
+
+import yaml
+
+os.environ.setdefault("OPENAI_BASE_URL", "http://localhost:9999")
+os.environ.setdefault("OPENAI_API_KEY", "test-key")
+os.environ.setdefault("POSTGRES_URL", "postgresql://localhost/test")
+
+from app.deterministic_services import task_catalog
+from scripts import convert_tasks_json_to_yaml
+
+
+class TaskYamlMigrationTest(unittest.TestCase):
+    def test_load_task_files_normalizes_yaml_schema(self) -> None:
+        temp_dir = Path(tempfile.mkdtemp(prefix="task-yaml-"))
+        yaml_path = temp_dir / "analysis.yaml"
+        yaml_path.write_text(
+            yaml.safe_dump(
+                {
+                    "id": "analysis_1",
+                    "titel": "Analysis",
+                    "level": "Uebungen",
+                    "slug": "analysis",
+                    "description": "Intro",
+                    "exercise_topic": [{"topic-id": "quadratische-gleichungen"}],
+                    "tasks": [
+                        {
+                            "aufgabe": [
+                                {"type": "text", "text": "Bestimme f(x)."},
+                                {"type": "image", "src": "analysis/01.png", "alt": "Graph einer Parabel."},
+                            ],
+                            "hinweise": [
+                                {"type": "text", "text": "Nutze die Ableitung."},
+                                {"type": "image", "src": "analysis/h1.png", "alt": "Hinweisgrafik."},
+                            ],
+                            "loesung": [{"type": "text", "text": "f'(x)=2x"}],
+                        }
+                    ],
+                },
+                sort_keys=False,
+                allow_unicode=True,
+            ),
+            encoding="utf-8",
+        )
+
+        try:
+            loaded = task_catalog.load_task_files(temp_dir)
+        finally:
+            shutil.rmtree(temp_dir, ignore_errors=True)
+
+        self.assertEqual(len(loaded), 1)
+        self.assertEqual(loaded[0]["_file_id"], "analysis_1")
+        self.assertEqual(loaded[0]["title"], "Analysis")
+        self.assertEqual(loaded[0]["intro"], "Intro")
+        self.assertEqual(loaded[0]["subsections"], ["quadratische-gleichungen"])
+        self.assertEqual(loaded[0]["tasks"][0]["id"], "01")
+        self.assertEqual(loaded[0]["tasks"][0]["statement"], "Bestimme f(x).")
+        self.assertEqual(loaded[0]["tasks"][0]["hints"], ["Nutze die Ableitung."])
+        self.assertEqual(loaded[0]["tasks"][0]["solution"], "f'(x)=2x")
+        self.assertEqual(
+            loaded[0]["tasks"][0]["images"],
+            [
+                {"src": "analysis/01.png", "description": "Graph einer Parabel."},
+                {"src": "analysis/h1.png", "description": "Hinweisgrafik."},
+            ],
+        )
+
+    def test_load_task_files_skips_invalid_yaml(self) -> None:
+        temp_dir = Path(tempfile.mkdtemp(prefix="task-yaml-invalid-"))
+        valid_path = temp_dir / "valid.yaml"
+        invalid_path = temp_dir / "invalid.yaml"
+        malformed_path = temp_dir / "malformed.yaml"
+
+        valid_path.write_text(
+            yaml.safe_dump(
+                {
+                    "id": "valid",
+                    "titel": "Valid",
+                    "level": "Uebungen",
+                    "slug": "valid",
+                    "description": "Intro",
+                    "exercise_topic": [],
+                    "tasks": [{"aufgabe": [], "hinweise": [], "loesung": []}],
+                },
+                sort_keys=False,
+                allow_unicode=True,
+            ),
+            encoding="utf-8",
+        )
+        invalid_path.write_text(
+            yaml.safe_dump(
+                {
+                    "id": "invalid",
+                    "titel": "Invalid",
+                    "slug": "invalid",
+                    "description": "Intro",
+                    "exercise_topic": [],
+                    "tasks": [],
+                },
+                sort_keys=False,
+                allow_unicode=True,
+            ),
+            encoding="utf-8",
+        )
+        malformed_path.write_text("tasks: [\n", encoding="utf-8")
+
+        try:
+            loaded = task_catalog.load_task_files(temp_dir)
+        finally:
+            shutil.rmtree(temp_dir, ignore_errors=True)
+
+        self.assertEqual([item["_file_id"] for item in loaded], ["valid"])
+
+    def test_build_task_catalog_keeps_api_shape_for_yaml_sources(self) -> None:
+        task_files = [
+            {
+                "_file_id": "analysis_1",
+                "title": "Analysis",
+                "intro": "Intro",
+                "subsections": ["quadratische-gleichungen"],
+                "tasks": [
+                    {
+                        "id": "01",
+                        "statement": "Bestimme f(x).",
+                        "hints": [],
+                        "solution": "f'(x)=2x",
+                        "images": [{"src": "analysis/01.png", "description": "Graph einer Parabel."}],
+                    }
+                ],
+            }
+        ]
+
+        catalog = task_catalog.build_task_catalog(task_files)
+
+        self.assertEqual(catalog[0]["file_id"], "analysis_1")
+        self.assertEqual(catalog[0]["title"], "Analysis")
+        self.assertEqual(catalog[0]["intro"], "Intro")
+        self.assertEqual(catalog[0]["tasks"][0]["task_id"], "01")
+        self.assertEqual(catalog[0]["tasks"][0]["statement"], "Bestimme f(x).")
+        self.assertEqual(
+            catalog[0]["tasks"][0]["images"],
+            [{"src": "/api/tasks/assets/analysis/01.png", "description": "Graph einer Parabel."}],
+        )
+
+    def test_convert_task_payload_maps_json_shape_to_yaml(self) -> None:
+        payload = {
+            "title": "Reelle Funktionen bestimmen",
+            "subsections": ["eigenschaften-reeller-funktionen"],
+            "intro": "Geben Sie Definitions- und Wertebereich der Funktion an:",
+            "tasks": [
+                {
+                    "id": "01",
+                    "statement": "$f(x)=x^2-3$",
+                    "images": [{"src": "reelle_funktionen_bestimmen/Bild1.png", "description": "Graph"}],
+                    "hints": ["Hinweis 1", ""],
+                    "solution": "Loesung",
+                }
+            ],
+        }
+
+        converted = convert_tasks_json_to_yaml.convert_task_payload(
+            payload,
+            source_file_id="Reelle Funktionen bestimmen",
+            image_path_map={"reelle_funktionen_bestimmen/Bild1.png": "reelle-funktionen-bestimmen/Bild1.png"},
+        )
+
+        self.assertEqual(converted["id"], "Reelle Funktionen bestimmen")
+        self.assertEqual(converted["slug"], "reelle-funktionen-bestimmen")
+        self.assertEqual(converted["level"], "Uebungen")
+        self.assertEqual(converted["exercise_topic"], [{"topic-id": "eigenschaften-reeller-funktionen"}])
+        self.assertEqual(converted["tasks"][0]["aufgabe"][0], {"type": "text", "text": "$f(x)=x^2-3$"})
+        self.assertEqual(
+            converted["tasks"][0]["aufgabe"][1],
+            {"type": "image", "src": "reelle-funktionen-bestimmen/Bild1.png", "alt": "Graph"},
+        )
+        self.assertEqual(converted["tasks"][0]["hinweise"], [{"type": "text", "text": "Hinweis 1"}])
+        self.assertEqual(converted["tasks"][0]["loesung"], [{"type": "text", "text": "Loesung"}])
+
+    def test_convert_all_rewrites_json_and_moves_images(self) -> None:
+        temp_dir = Path(tempfile.mkdtemp(prefix="task-convert-"))
+        images_dir = temp_dir / "images" / "old-folder"
+        images_dir.mkdir(parents=True, exist_ok=True)
+        (images_dir / "Bild1.png").write_text("image", encoding="utf-8")
+        json_path = temp_dir / "Reelle Funktionen bestimmen.json"
+        json_path.write_text(
+            json.dumps(
+                {
+                    "title": "Reelle Funktionen bestimmen",
+                    "subsections": ["eigenschaften-reeller-funktionen"],
+                    "intro": "Intro",
+                    "tasks": [
+                        {
+                            "id": "01",
+                            "statement": "$f(x)=x^2-3$",
+                            "images": [{"src": "old-folder/Bild1.png", "description": "Graph"}],
+                            "hints": ["Hinweis 1"],
+                            "solution": "Loesung",
+                        }
+                    ],
+                },
+                ensure_ascii=False,
+            ),
+            encoding="utf-8",
+        )
+
+        try:
+            converted, missing_images = convert_tasks_json_to_yaml.convert_all(temp_dir)
+            yaml_path = temp_dir / "reelle-funktionen-bestimmen.yaml"
+            self.assertEqual(converted, [yaml_path])
+            self.assertEqual(missing_images, [])
+            self.assertFalse(json_path.exists())
+            self.assertTrue(yaml_path.exists())
+            self.assertTrue((temp_dir / "images" / "reelle-funktionen-bestimmen" / "Bild1.png").exists())
+
+            payload = yaml.safe_load(yaml_path.read_text(encoding="utf-8"))
+            self.assertEqual(payload["id"], "Reelle Funktionen bestimmen")
+            self.assertEqual(
+                payload["tasks"][0]["aufgabe"][1]["src"],
+                "reelle-funktionen-bestimmen/Bild1.png",
+            )
+        finally:
+            shutil.rmtree(temp_dir, ignore_errors=True)
+
+    def test_convert_all_uses_distinct_yaml_filenames_for_duplicate_titles(self) -> None:
+        temp_dir = Path(tempfile.mkdtemp(prefix="task-convert-duplicate-"))
+        first_json = temp_dir / "Auf Beschränktheit untersuchen.json"
+        second_json = temp_dir / "Auf Beschränktheit untersuchen_images.json"
+
+        payload = {
+            "title": "Auf Beschränktheit untersuchen",
+            "subsections": [],
+            "intro": "Intro",
+            "tasks": [{"id": "01", "statement": "s", "hints": [], "solution": "l"}],
+        }
+        first_json.write_text(json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+        second_json.write_text(json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+
+        try:
+            converted, missing_images = convert_tasks_json_to_yaml.convert_all(temp_dir)
+            self.assertEqual(missing_images, [])
+            self.assertEqual(
+                sorted(path.name for path in converted),
+                [
+                    "auf-beschranktheit-untersuchen-images.yaml",
+                    "auf-beschranktheit-untersuchen.yaml",
+                ],
+            )
+        finally:
+            shutil.rmtree(temp_dir, ignore_errors=True)
+
+
+if __name__ == "__main__":
+    unittest.main()

math-tutor/docker/docker-compose.yml

@@ -25,12 +25,12 @@ services:
     networks:
       - web
 
-  backend-ollama:
+  backend-stochastik:
     build:
       context: ../backend
       dockerfile: Dockerfile
     env_file:
-      - ../backend/.env-ollama
+      - ../backend/.env-stochastik
     expose:
       - "8000"
     healthcheck:
@@ -63,14 +63,14 @@ services:
     networks:
       - web
 
-  frontend-ollama:
+  frontend-stochastik:
     build:
       context: ../frontend
       dockerfile: Dockerfile
       args:
-        VITE_API_BASE_URL: http://backend-ollama:8000
+        VITE_API_BASE_URL: http://backend-stochastik:8000
     depends_on:
-      - backend-ollama
+      - backend-stochastik
     restart: unless-stopped
     networks:
       - web