auräumaktion

math-tutor/README.md

@@ -15,7 +15,11 @@ cd math-tutor/frontend
 npm install
 npm run dev
 ```
-Um es im Netzwerk verfügbar zu machen mit: npm run dev -- --host 0.0.0.0
+To make it accessible over the network.
+Add the frontend- and backend-adress in the .env file in the frontend- and backend-folder.
+
+python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
+npm run dev -- --host 0.0.0.0
 ## Database (pgvector) setup
 
 Ensure `POSTGRES_URL` and embedding env vars are in `backend/.env`:
@@ -51,7 +55,7 @@ python -m scripts.retrieval_cli query --q "Was ist eine Teilmenge?" --k 8 --expa
 
 ## Configuration
 
-System prompt (LLM):
+Differend Orchestrators:
 - Edit `math-tutor/backend/app/api/chat.py` and update `SYSTEM_PROMPT`.
 
 Retrieval settings:

math-tutor/backend/app/api/chat.py

 from __future__ import annotations
 
+from logging import config
 from typing import List, Optional
 
 import logging
@@ -8,9 +9,12 @@ from fastapi import APIRouter, HTTPException, Path, Query
 from pydantic import BaseModel, Field
 
 from app.deterministic_services import session_store
-from app.deterministic_services.orchestrators import orchestrator_tutor as orchestrator
-from app.deterministic_services.orchestrators import orchestrator_QA 
+import app.config as config
 
+if config.get_orchestrator() == "tutor":
+    from app.deterministic_services.orchestrators import orchestrator_tutor as orchestrator
+else:
+    from app.deterministic_services.orchestrators import orchestrator_QA as orchestrator
 
 
 router = APIRouter()

math-tutor/backend/app/api/context.py

@@ -6,7 +6,6 @@ from fastapi import APIRouter, HTTPException
 from pydantic import BaseModel, Field
 
 from app.deterministic_services import context_store
-from app.deterministic_services import Source
 
 
 router = APIRouter()

math-tutor/backend/app/api/health.py

 from __future__ import annotations
 
 import logging
-import os
 from typing import Any, Dict
 
 import httpx
 import psycopg
 from fastapi import APIRouter
 
+import app.config as config
+
 router = APIRouter()
 logger = logging.getLogger(__name__)
 
 
 def _check_ollama() -> dict:
-    base_url = os.getenv("OLLAMA_URL", "http://localhost:11434").rstrip("/")
+    base_url = config.get_ollama_settings().base_url.rstrip("/")
     url = f"{base_url}/api/tags"
     try:
         with httpx.Client(timeout=5.0) as client:
@@ -34,14 +35,13 @@ def _normalize_openai_models_url(base_url: str) -> str:
 
 
 def _check_openai() -> dict:
-    base_url = os.getenv("OPENAI_BASE_URL")
-    api_key = os.getenv("OPENAI_API_KEY")
-    if not base_url or not api_key:
+    settings = config.get_openai_base_settings()
+    if not settings:
         return {"status": "missing_config"}
 
-    url = _normalize_openai_models_url(base_url)
+    url = _normalize_openai_models_url(settings.base_url)
     try:
-        headers = {"Authorization": f"Bearer {api_key}"}
+        headers = {"Authorization": f"Bearer {settings.api_key}"}
         with httpx.Client(timeout=5.0) as client:
             response = client.get(url, headers=headers)
         if response.status_code in (401, 403):
@@ -53,8 +53,9 @@ def _check_openai() -> dict:
 
 
 def _check_postgres() -> dict:
-    pg_url = os.getenv("POSTGRES_URL")
-    if not pg_url:
+    try:
+        pg_url = config.get_postgres_url()
+    except ValueError:
         return {"status": "missing_config"}
     try:
         with psycopg.connect(pg_url, connect_timeout=5) as conn:

math-tutor/backend/app/api/retrieval.py

-from __future__ import annotations
-
-from __future__ import annotations
-
-import os
-from pathlib import Path
-from typing import List, Optional
-
-from fastapi import APIRouter, HTTPException
-from pydantic import BaseModel, Field
-
-from app import config
-from app.deterministic_services.embeddings import OpenAILikeEmbeddings
-from app.deterministic_services import vector_store
-
-router = APIRouter()
-
-def _get_embedder() -> OpenAILikeEmbeddings:
-    settings = config.get_embedding_settings()
-    return OpenAILikeEmbeddings(
-        base_url=settings.base_url,
-        api_key=settings.api_key,
-        model=settings.model,
-        target_dim=settings.target_dim,
-    )
-
-
-class InitDbRequest(BaseModel):
-    pg_url: Optional[str] = None
-
-
-class IngestRequest(BaseModel):
-    pg_url: Optional[str] = None
-    base_dir: str = Field(default="markdown")
-
-
-class QueryRequest(BaseModel):
-    pg_url: Optional[str] = None
-    query: str = Field(..., min_length=1)
-    k: int = 4
-    expand_links: bool = True
-    section_index: Optional[int] = None
-    subsection_index: Optional[int] = None
-    source_type_filter: Optional[List[str]] = None
-    neighbor_expand: int = 0
-
-
-@router.post("/api/retrieval/init-db")
-def init_db(request: InitDbRequest) -> dict:
-    pg_url = request.pg_url or os.getenv("POSTGRES_URL")
-    if not pg_url:
-        raise HTTPException(status_code=500, detail="Missing POSTGRES_URL")
-    vector_store.init_db(pg_url)
-    return {"status": "ok"}
-
-
-@router.post("/api/retrieval/ingest")
-def ingest(request: IngestRequest) -> dict:
-    pg_url = request.pg_url or os.getenv("POSTGRES_URL")
-    if not pg_url:
-        raise HTTPException(status_code=500, detail="Missing POSTGRES_URL")
-    base_dir = Path(request.base_dir)
-    if not base_dir.exists():
-        raise HTTPException(status_code=400, detail="base_dir does not exist")
-
-    embedder = _get_embedder()
-    docs = vector_store.load_docs(base_dir)
-    embeddings = vector_store.embed_documents(embedder, [doc.markdown for doc in docs])
-    upserted = vector_store.upsert_docs(pg_url, docs, embeddings)
-    return {"status": "ok", "upserted": upserted}
-
-
-@router.post("/api/retrieval/query")
-def query(request: QueryRequest) -> dict:
-    pg_url = request.pg_url or os.getenv("POSTGRES_URL")
-    if not pg_url:
-        raise HTTPException(status_code=500, detail="Missing POSTGRES_URL")
-    embedder = _get_embedder()
-    result = vector_store.retrieve(
-        pg_url=pg_url,
-        embedder=embedder,
-        query=request.query,
-        k=request.k,
-        section_index=request.section_index,
-        subsection_index=request.subsection_index,
-        source_type_filter=request.source_type_filter,
-        expand_links=request.expand_links,
-        neighbor_expand=request.neighbor_expand,
-    )
-    return result
-
-
-@router.get("/api/retrieval/sections")
-def sections(pg_url: Optional[str] = None) -> list[dict]:
-    url = pg_url or os.getenv("POSTGRES_URL")
-    if not url:
-        raise HTTPException(status_code=500, detail="Missing POSTGRES_URL")
-    return vector_store.list_sections(url)
-
-
-@router.get("/api/retrieval/subsections")
-def subsections(pg_url: Optional[str] = None, section_index: Optional[int] = None) -> list[dict]:
-    url = pg_url or os.getenv("POSTGRES_URL")
-    if not url:
-        raise HTTPException(status_code=500, detail="Missing POSTGRES_URL")
-    return vector_store.list_subsections(url, section_index)

math-tutor/backend/app/config.py

@@ -14,6 +14,10 @@ class EmbeddingSettings(BaseModel):
     model: str
     target_dim: int = 1024
 
+
+def get_orchestrator() -> str:
+    return os.getenv("ORCHESTRATOR", "qa").lower()
+
 def get_embedding_settings() -> EmbeddingSettings:
     embedding_type = os.getenv("EMBEDDING_TYPE", "openai-like")
     if embedding_type == "sentence-transformer":
@@ -56,6 +60,11 @@ class OpenAIChatSettings:
     timeout: float | None
     temperature: float | None
 
+@dataclass(frozen=True)
+class OpenAIBaseSettings:
+    base_url: str
+    api_key: str
+
 @dataclass(frozen=True)
 class MathpixSettings:
     app_id: str
@@ -91,17 +100,26 @@ def get_ollama_settings() -> OllamaSettings:
         temperature=_read_float(os.getenv("OLLAMA_TEMPERATURE")),
     )
 
+def get_openai_base_settings() -> OpenAIBaseSettings | None:
+    base_url = os.getenv("OPENAI_BASE_URL")
+    api_key = os.getenv("OPENAI_API_KEY")
+    if not base_url or not api_key:
+        return None
+    return OpenAIBaseSettings(
+        base_url=base_url,
+        api_key=api_key,
+    )
+
 def get_openai_chat_settings() -> OpenAIChatSettings | None:
     model = os.getenv("OPENAI_CHAT_MODEL")
     if not model:
         return None
-    base_url = os.getenv("OPENAI_BASE_URL")
-    api_key = os.getenv("OPENAI_API_KEY")
-    if not base_url or not api_key:
+    base_settings = get_openai_base_settings()
+    if not base_settings:
         raise ValueError("Missing OPENAI_BASE_URL or OPENAI_API_KEY for chat")
     return OpenAIChatSettings(
-        base_url=base_url,
-        api_key=api_key,
+        base_url=base_settings.base_url,
+        api_key=base_settings.api_key,
         model=model,
         timeout=_read_float(os.getenv("OPENAI_CHAT_TIMEOUT")),
         temperature=_read_float(os.getenv("OPENAI_CHAT_TEMPERATURE")),

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

-from app.deterministic_services import context_store, retrieval_service, tool_logging
-from app.deterministic_services import referenz_decoder
-from app.LLM_services import qa_LLM
 from typing import List
 
+from app.deterministic_services import context_store, tool_logging, vector_store, Source, referenz_decoder
+from app.LLM_services import qa_LLM
+from app.deterministic_services.embeddings import EmbeddingFactory
+
+import app.config as config
 
 def _is_new_chat(messages: list[dict]) -> bool:
     return not any(m.get("role") == "assistant" for m in messages)
@@ -13,7 +15,7 @@ def _append_tool_log(tool_log: list[dict], name: str, args: dict, response: obje
 
 def _bootstrap_context(sheet: dict, query_text: str, tool_log: list[dict]) -> None:
     # Erstelle einen neuen Retrieval-Block oder aktualisiere den bestehenden
-    sources = retrieval_service.retrieve_context(query_text=query_text)
+    sources = _retrieve_context(query_text=query_text)
     
     retrievals = sheet.get("retrieval_contexts", [])
     if retrievals:
@@ -36,6 +38,17 @@ def _extract_user_messages(messages: list[dict]) -> list[str]:
                 user_contents.append(content)
     return user_contents
 
+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()
+    sources = vector_store.retrieve(
+        pg_url=url,
+        embedder=embedder,
+        query=query_text,
+        k=8,
+        expand_links=True,
+    )
+    return sources
 
 def run_chat(messages: list[dict], draft: str | None = None) -> dict:
     # Input-Fehlerbehandlung

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

-from app.deterministic_services import context_store, retrieval_service, tool_logging
-from app.deterministic_services import referenz_decoder
+from typing import List
+
+from app.deterministic_services import context_store, tool_logging, vector_store, Source, referenz_decoder
 from app.LLM_services import hint_LLM, decision_LLM, math_intent_LLM, solver_LLM
+from app.deterministic_services.embeddings import EmbeddingFactory
+
+import app.config as config
 
 
 def _is_new_chat(messages: list[dict]) -> bool:
@@ -12,7 +16,7 @@ def _append_tool_log(tool_log: list[dict], name: str, args: dict, response: obje
 
 def _bootstrap_context(sheet: dict, query_text: str, tool_log: list[dict]) -> None:
     # Erstelle einen neuen Retrieval-Block oder aktualisiere den bestehenden
-    sources = retrieval_service.retrieve_context(query_text=query_text)
+    sources = _retrieve_context(query_text=query_text)
     
     retrievals = sheet.get("retrieval_contexts", [])
     if retrievals:
@@ -40,6 +44,18 @@ def _extract_user_messages(messages: list[dict]) -> list[str]:
                 user_contents.append(content)
     return user_contents
 
+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()
+    sources = vector_store.retrieve(
+        pg_url=url,
+        embedder=embedder,
+        query=query_text,
+        k=8,
+        expand_links=True,
+    )
+    return sources
+
 
 def run_chat(messages: list[dict], draft: str | None = None) -> dict:
     # Input-Fehlerbehandlung

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

-# app/deterministic_services/retrieval_service.py
-
-from __future__ import annotations
-
-from typing import List
-
-from app.deterministic_services import Source
-from app import config
-from app.deterministic_services.embeddings import EmbeddingFactory
-from app.deterministic_services import vector_store
-
-
-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()
-    sources = vector_store.retrieve(
-        pg_url=url,
-        embedder=embedder,
-        query=query_text,
-        k=8,
-        expand_links=True,
-    )
-    return sources
\ No newline at end of file

math-tutor/backend/app/main.py

@@ -2,7 +2,7 @@ from contextlib import asynccontextmanager
 
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
-from app.api import canvas, chat, health, retrieval, context
+from app.api import canvas, chat, health, context
 from app.config import get_frontend_url
 
 
@@ -24,6 +24,5 @@ app.add_middleware(
 
 app.include_router(chat.router)
 app.include_router(canvas.router)
-app.include_router(retrieval.router)
 app.include_router(context.router)
 app.include_router(health.router)