Fix models sort (by output cost), filter non-chat models, add compression variants

assistant/assistant.py

@@ -228,7 +228,7 @@ class Assistant(commands.Cog):
 
     @assistant.command(name="models")
     async def assistant_models(self, ctx: commands.Context):
-        """List available models sorted by cost (cheapest first)."""
+        """List available chat models sorted by cost (cheapest first)."""
         # Pricing per 1M tokens (EUR) from GreenPT docs
         MODEL_PRICING = {
             "deepseek-v4-flash-0731": {"input": 0.14, "output": 0.35},
@@ -253,6 +253,23 @@ class Assistant(commands.Cog):
             "llama-3.3-70b-instruct": {"input": 1.10, "output": 1.10},
             "mistral-medium-3.5-128b": {"input": 1.80, "output": 9.00},
             "kimi-k3": {"input": 3.30, "output": 16.50},
+            # Compression variants (same price as glm-5.2)
+            "glm-5.2-caveman": {"input": 1.10, "output": 4.40},
+            "glm-5.2-caveman-lite": {"input": 1.10, "output": 4.40},
+            "glm-5.2-caveman-ultra": {"input": 1.10, "output": 4.40},
+            "glm-5.2-honey": {"input": 1.10, "output": 4.40},
+            "glm-5.2-honey-lite": {"input": 1.10, "output": 4.40},
+            "glm-5.2-honey-ultra": {"input": 1.10, "output": 4.40},
+            "glm-5.2-ponytail": {"input": 1.10, "output": 4.40},
+            "glm-5.2-ponytail-lite": {"input": 1.10, "output": 4.40},
+            "glm-5.2-ponytail-ultra": {"input": 1.10, "output": 4.40},
+        }
+
+        # Models that aren't usable for chat completions
+        NON_CHAT_MODELS = {
+            "green-embedding", "green-embeddings", "green-rerank",
+            "green-s", "green-s-pro", "bge-multilingual-gemma2",
+            "qwen3-embedding-8b",
         }
 
         api_key = await self.config.api_key()
@@ -280,23 +297,24 @@ class Assistant(commands.Cog):
             await ctx.send("No models returned by the API.")
             return
 
-        # Build list with pricing, sorted by combined cost (input + output)
+        # Build list with pricing, sorted by output cost (most relevant for chat)
         current_model = await self.config.guild(ctx.guild).model()
-        lines = []
         model_entries = []
         for m in models:
             model_id = m["id"]
+            if model_id in NON_CHAT_MODELS:
+                continue
             pricing = MODEL_PRICING.get(model_id)
             if pricing:
-                total = pricing["input"] + pricing["output"]
+                sort_key = pricing["output"]
-                model_entries.append((total, model_id, pricing))
+                model_entries.append((sort_key, model_id, pricing))
             else:
-                # Unknown pricing, put at the end
                 model_entries.append((9999, model_id, None))
 
-        model_entries.sort(key=lambda x: x[0])
+        model_entries.sort(key=lambda x: (x[0], x[1]))
 
-        for total, model_id, pricing in model_entries:
+        lines = []
+        for sort_key, model_id, pricing in model_entries:
             marker = " ◀" if model_id == current_model else ""
             if pricing:
                 lines.append(
@@ -310,7 +328,7 @@ class Assistant(commands.Cog):
             description="\n".join(lines),
             color=await ctx.embed_color(),
         )
-        embed.set_footer(text="Prices in EUR per 1M tokens. ◀ = current model.")
+        embed.set_footer(text="Prices in EUR per 1M tokens, sorted by output cost. ◀ = current model.")
         await ctx.send(embed=embed)
 
     @assistant.command(name="temp")