fix(diarization-ui): validate non-empty content before LLM call, add OLLAMA_THINK flag

Empty documents caused the model to spin in its thinking loop and waste
all tokens. Now raises a clear job error before the Ollama call.
Also adds OLLAMA_THINK env var (default false) to control whether
the model uses extended thinking - disabling it avoids runaway thinking
loops on ambiguous inputs.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app.py

@@ -16,6 +16,7 @@ API_BASE = os.getenv("API_BASE", "http://gx10.aquantico.lan:8093").rstrip("/")
 OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://gx10.aquantico.lan:11434").rstrip("/")
 OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "qwen3.5:9b")
 OLLAMA_NUM_PREDICT = int(os.getenv("OLLAMA_NUM_PREDICT", "4096"))
+OLLAMA_THINK = os.getenv("OLLAMA_THINK", "false").lower() in ("1", "true", "yes")
 DB_PATH = os.getenv("DB_PATH", "/data/ui.db")
 
 app = FastAPI(title="Diarization UI")
@@ -371,6 +372,8 @@ def _process_analysis_job(job_id: int):
             prm = c.execute("SELECT * FROM prompts WHERE id=?", (j["prompt_id"],)).fetchone()
         if not doc or not prm:
             raise RuntimeError("Dokument oder Prompt nicht gefunden")
+        if not (doc["content_md"] or "").strip():
+            raise RuntimeError("Dokument hat keinen Inhalt – bitte zuerst das Transkript prüfen")
 
         user_extra = (j.get("user_prompt") or "").strip()
         llm_prompt = (
@@ -387,7 +390,7 @@ def _process_analysis_job(job_id: int):
 
         r = requests.post(
             f"{OLLAMA_BASE_URL}/api/generate",
-            json={"model": OLLAMA_MODEL, "prompt": llm_prompt, "stream": True, "options": {
+            json={"model": OLLAMA_MODEL, "prompt": llm_prompt, "stream": True, "think": OLLAMA_THINK, "options": {
                 "num_ctx": num_ctx,
                 "num_predict": OLLAMA_NUM_PREDICT,
                 "repeat_penalty": 1.15,