Upgrade to Sonnet 4.6 + add Gemini 3.1 Pro support

- Refactored analyze_with_claude into shared _build_analysis_prompt + _parse_llm_response helpers
- New analyze_with_gemini() using Gemini 3.1 Pro ($2/M in, MMMLU 92.6% — best multilingual)
- Unified analyze_with_llm(provider) dispatcher with auto-fallback (Claude → Gemini)
- API endpoint accepts llm_provider in StartJobIn (claude/gemini/auto)
- Frontend dropdown to pick LLM
- Default model is now Sonnet 4.6 (was Haiku 4.5) — 3x quality at 3x price (~3 cents/video)
- Gemini support is opt-in: needs GEMINI_API_KEY env var to activate

app/main.py

@@ -237,6 +237,11 @@ def process_job(job_id):
             ]
             if job.get("include_prebuild"):
                 cmd += ["--include-prebuild"]
+            # LLM provider (claude/gemini/auto)
+            if job.get("llm_provider"):
+                cmd += ["--llm-provider", job["llm_provider"]]
+            if job.get("llm_model"):
+                cmd += ["--llm-model", job["llm_model"]]
             # lang: če None ali 'auto', pusti analyze.py auto-detect
             if job.get("lang") and job["lang"] not in ("auto", ""):
                 cmd += ["--lang", job["lang"]]
@@ -409,6 +414,9 @@ class StartJobIn(BaseModel):
     subtitle_style: str = "reels"
     whisper_model: str = "large-v3"
     quality: str = "medium"
+    # LLM za semantično analizo + popravke
+    llm_provider: str = "claude"  # claude / gemini / auto
+    llm_model: Optional[str] = None  # specifičen model (privzeto najboljši za provider)
 
 
 # ────────────────────────────────────────────────────────────────
@@ -512,6 +520,8 @@ async def start_processing(
         subtitle_style=payload.subtitle_style,
         whisper_model=payload.whisper_model,
         quality=payload.quality,
+        llm_provider=payload.llm_provider,
+        llm_model=payload.llm_model,
         current_step="V vrsti za obdelavo",
     )
     background.add_task(process_job, payload.job_id)