entari-plugin-hyw 4.0.0rc5__py3-none-any.whl → 4.0.0rc7__py3-none-any.whl

entari_plugin_hyw/__init__.py

@@ -22,7 +22,7 @@ from arclet.entari.event.command import CommandReceive
 from .modular_pipeline import ModularPipeline
 from .history import HistoryManager
 from .render_vue import ContentRenderer, get_content_renderer
-from .misc import process_onebot_json, process_images, resolve_model_name, render_refuse_answer, REFUSE_ANSWER_MARKDOWN
+from .misc import process_onebot_json, process_images, resolve_model_name, render_refuse_answer, render_image_unsupported, REFUSE_ANSWER_MARKDOWN
 from arclet.entari.event.lifespan import Cleanup
 
 import os
@@ -93,6 +93,7 @@ class ModelConfig:
     model_provider: Optional[str] = None
     input_price: Optional[float] = None
     output_price: Optional[float] = None
+    image_input: bool = True
 
 
 @dataclass
@@ -117,11 +118,10 @@ class HywConfig(BasicConfModel):
     instruct: Optional[ModelConfig] = None
     qa: Optional[ModelConfig] = None
     main: Optional[ModelConfig] = None  # Summary stage
+    vision: Optional[ModelConfig] = None  # Vision description stage
     
     # Search/Fetch Settings
-    search_engine: str = "bing"
-    enable_domain_blocking: bool = True
-    page_content_mode: str = "text"
+    search_engine: str = "google"
     
     # Rendering Settings
     headless: bool = False
@@ -131,10 +131,10 @@ class HywConfig(BasicConfModel):
     # Bot Behavior
     save_conversation: bool = False
     reaction: bool = False
-    quote: bool = True
+    quote: bool = False
     
     # UI Theme
-    theme_color: str = "#ef4444"
+    theme_color: str = "#ff0000"
     
     def __post_init__(self):
         """Parse and normalize theme color after initialization."""
@@ -146,6 +146,8 @@ class HywConfig(BasicConfModel):
             self.qa = ModelConfig(**self.qa)
         if isinstance(self.main, dict):
             self.main = ModelConfig(**self.main)
+        if isinstance(self.vision, dict):
+            self.vision = ModelConfig(**self.vision)
     
     def get_model_config(self, stage: str) -> Dict[str, Any]:
         """
@@ -170,6 +172,9 @@ class HywConfig(BasicConfModel):
             secondary = self.main
         elif stage == "main":
             primary = self.main
+        elif stage == "vision":
+            primary = self.vision
+            secondary = self.main  # Fallback to main if vision not fully configured
             
         # Build result with fallback logic
         def resolve(field_name: str, is_essential: bool = True):
@@ -315,14 +320,66 @@ async def process_request(
 
         images, err = await process_images(mc, vision_model)
         
-        # Start preparing render tab (async)
+        # Check image input support
+        model_cfg_dict = next((m for m in conf.models if m.get("name") == model), None)
+        image_input_supported = True
+        if model_cfg_dict:
+            image_input_supported = model_cfg_dict.get("image_input", True)
+        
+        # Log inferenced content mode
+        inferred_content_mode = "image" if image_input_supported else "text"
+        logger.info(f"Process Request: Model '{model}' Image Input: {image_input_supported} -> Mode: {inferred_content_mode}")
+        
+        if images and not image_input_supported:
+            logger.warning(f"Model '{model}' does not support images, but user sent {len(images)} images.")
+            
+            # Start renderer for the unsupported card
+            renderer = await get_content_renderer()
+            render_tab_task = asyncio.create_task(renderer.prepare_tab())
+            
+            # Wait for tab and render unsupported
+            try:
+                tab_id = await render_tab_task
+            except Exception as e:
+                tab_id = None
+                
+            import tempfile
+            with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as tf:
+                output_path = tf.name
+                
+            render_ok = await render_image_unsupported(
+                renderer=renderer,
+                output_path=output_path,
+                theme_color=conf.theme_color,
+                tab_id=tab_id
+            )
+            
+            if render_ok:
+                with open(output_path, "rb") as f:
+                    img_data = base64.b64encode(f.read()).decode()
+                await session.send(MessageChain(Image(src=f'data:image/png;base64,{img_data}')))
+                if os.path.exists(output_path):
+                    os.remove(output_path)
+            return
+
         renderer = await get_content_renderer()
         render_tab_task = asyncio.create_task(renderer.prepare_tab())
         tab_id = None
 
         # Call Pipeline directly
         safe_input = msg_text
-        pipeline = ModularPipeline(conf)
+        
+        async def send_noti(msg: str):
+            try:
+                # Send simple text notification
+                if conf.quote:
+                    await session.send([Quote(session.event.message.id), msg])
+                else:
+                    await session.send(msg)
+            except Exception as e:
+                logger.warning(f"Failed to send notification: {e}")
+
+        pipeline = ModularPipeline(conf, send_func=send_noti)
         try:
             resp = await pipeline.execute(
                 safe_input,
@@ -385,6 +442,7 @@ async def process_request(
                 tab_id=tab_id,
             )
         else:
+            logger.info(f"Rendering card with {len(structured.get('references', []))} references...")
             render_ok = await renderer.render(
                 markdown_content=content,
                 output_path=output_path,
@@ -396,6 +454,7 @@ async def process_request(
                 stages_used=final_resp.get("stages_used", []),
                 theme_color=conf.theme_color,
             )
+            logger.info(f"Render completed: {render_ok}")
         
         # Send & Save
         if not render_ok:
@@ -442,10 +501,13 @@ async def process_request(
         if conf.save_conversation and sent_id:
             try:
                 # Pass web_results to save fetched pages as markdown, and output image
+                # Also pass vision_trace and instruct_traces for dedicated logs
                 history_manager.save_to_disk(
                     sent_id, 
                     web_results=final_resp.get("web_results"),
-                    image_path=output_path if 'output_path' in locals() else None
+                    image_path=output_path if 'output_path' in locals() else None,
+                    vision_trace=final_resp.get("vision_trace"),
+                    instruct_traces=final_resp.get("instruct_traces"),
                 )
             except Exception as e:
                 logger.warning(f"Failed to save conversation: {e}")