Add semantic tagging, search, favourite/NSFW filtering, and LLM job queue

Replaces old list-format tags (which duplicated prompt content) with structured
dict tags per category (origin_series, outfit_type, participants, style_type,
scene_type, etc.). Tags are now purely organizational metadata — removed from
the prompt pipeline entirely.

Adds is_favourite and is_nsfw columns to all 8 resource models. Favourite is
DB-only (user preference); NSFW is mirrored in JSON tags for rescan persistence.
All library pages get filter controls and favourites-first sorting.

Introduces a parallel LLM job queue (_enqueue_task + _llm_queue_worker) for
background tag regeneration, with the same status polling UI as ComfyUI jobs.
Fixes call_llm() to use has_request_context() fallback for background threads.

Adds global search (/search) across resources and gallery images, with navbar
search bar. Adds gallery image sidecar JSON for per-image favourite/NSFW metadata.

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

services/llm.py

@@ -2,7 +2,7 @@ import os
 import json
 import asyncio
 import requests
-from flask import request as flask_request
+from flask import has_request_context, request as flask_request
 from mcp import ClientSession, StdioServerParameters
 from mcp.client.stdio import stdio_client
 from models import Settings
@@ -77,6 +77,28 @@ def call_mcp_tool(name, arguments):
         return json.dumps({"error": str(e)})
 
 
+async def _run_character_mcp_tool(name, arguments):
+    server_params = StdioServerParameters(
+        command="docker",
+        args=["run", "--rm", "-i",
+               "-v", "character-cache:/root/.local/share/character_details",
+               "character-mcp:latest"],
+    )
+    async with stdio_client(server_params) as (read, write):
+        async with ClientSession(read, write) as session:
+            await session.initialize()
+            result = await session.call_tool(name, arguments)
+            return result.content[0].text
+
+
+def call_character_mcp_tool(name, arguments):
+    try:
+        return asyncio.run(_run_character_mcp_tool(name, arguments))
+    except Exception as e:
+        print(f"Character MCP Tool Error: {e}")
+        return None
+
+
 def load_prompt(filename):
     path = os.path.join('data/prompts', filename)
     if os.path.exists(path):
@@ -100,7 +122,7 @@ def call_llm(prompt, system_prompt="You are a creative assistant."):
         headers = {
             "Authorization": f"Bearer {settings.openrouter_api_key}",
             "Content-Type": "application/json",
-            "HTTP-Referer": flask_request.url_root,
+            "HTTP-Referer": flask_request.url_root if has_request_context() else "http://localhost:5000/",
             "X-Title": "Character Browser"
         }
         model = settings.openrouter_model or 'google/gemini-2.0-flash-001'
@@ -120,7 +142,8 @@ def call_llm(prompt, system_prompt="You are a creative assistant."):
         {"role": "user", "content": prompt}
     ]
 
-    max_turns = 10
+    max_turns = 15
+    tool_turns_remaining = 8  # stop offering tools after this many tool-calling turns
     use_tools = True
     format_retries = 3  # retries allowed for unexpected response format
 
@@ -131,13 +154,13 @@ def call_llm(prompt, system_prompt="You are a creative assistant."):
             "messages": messages,
         }
 
-        # Only add tools if supported/requested
-        if use_tools:
+        # Only add tools if supported/requested and we haven't exhausted tool turns
+        if use_tools and tool_turns_remaining > 0:
             data["tools"] = DANBOORU_TOOLS
             data["tool_choice"] = "auto"
 
         try:
-            response = requests.post(url, headers=headers, json=data)
+            response = requests.post(url, headers=headers, json=data, timeout=120)
 
             # If 400 Bad Request and we were using tools, try once without tools
             if response.status_code == 400 and use_tools:
@@ -158,6 +181,7 @@ def call_llm(prompt, system_prompt="You are a creative assistant."):
                 raise KeyError('message')
 
             if message.get('tool_calls'):
+                tool_turns_remaining -= 1
                 messages.append(message)
                 for tool_call in message['tool_calls']:
                     name = tool_call['function']['name']
@@ -170,6 +194,8 @@ def call_llm(prompt, system_prompt="You are a creative assistant."):
                         "name": name,
                         "content": tool_result
                     })
+                if tool_turns_remaining <= 0:
+                    print("Tool turn limit reached — next request will not offer tools")
                 continue
 
             return message['content']