Limit HF providers to Groq/Cerebras/OpenRouter and enable traceback logging

gateway.py

@@ -295,12 +295,24 @@ def llm_route(messages, model=None, tools=None, tool_choice=None):
         ]
     else:
         providers = [
-            ("ollama",     lambda: {"role":"assistant","content": llm_ollama(messages, model=model or "dolor3v_coder:turbo")}),
-            ("groq",       lambda: llm_groq(messages, model=model or "llama-3.3-70b-versatile")),
-            ("cerebras",   lambda: {"role":"assistant","content": llm_cerebras(messages, model=model or "gpt-oss-120b")}),
-            ("glm",        lambda: {"role":"assistant","content": llm_glm(messages, model=model or "glm-4.7-flash")}),
-            ("openrouter", lambda: llm_openrouter(messages, model=model or "nvidia/nemotron-super-49b-v1:free")),
-            ("openai",     lambda: {"role":"assistant","content": llm_openai(messages, model=model or "gpt-4o-mini")}),
+            ("groq", lambda: llm_groq(
+                messages,
+                model=model or "llama-3.3-70b-versatile"
+            )),
+            ("cerebras", lambda: {
+                "role":"assistant",
+                "content": llm_cerebras(
+                    messages,
+                    model=model or "gpt-oss-120b"
+                )
+            }),
+            ("openrouter", lambda: llm_openrouter(
+                messages,
+                model=model or os.getenv(
+                    "OPENROUTER_MODEL",
+                    "qwen/qwen3-coder:free"
+                )
+            )),
         ]
 
     for name, fn in providers:
@@ -315,7 +327,10 @@ def llm_route(messages, model=None, tools=None, tool_choice=None):
             emit_event("custom", {"provider":name,"tokens":len(content_preview)}, "mcp-llm")
             return message, name
         except Exception as e:
-            log(f"LLM {name} failed: {e}")
+            import traceback
+            traceback.print_exc()
+            print(f"[LLM ERROR] {name}: {repr(e)}")
+            log(f"LLM {name} failed: {repr(e)}")
             continue
     return {"role":"assistant","content":"All LLM providers failed"}, "none"
 

patch_llm_route.py

@@ -0,0 +1,61 @@
+from pathlib import Path
+
+p = Path("gateway.py")
+text = p.read_text()
+
+old = '''    else:
+        providers = [
+            ("ollama",     lambda: {"role":"assistant","content": llm_ollama(messages, model=model or "dolor3v_coder:turbo")}),
+            ("groq",       lambda: llm_groq(messages, model=model or "llama-3.3-70b-versatile")),
+            ("cerebras",   lambda: {"role":"assistant","content": llm_cerebras(messages, model=model or "gpt-oss-120b")}),
+            ("glm",        lambda: {"role":"assistant","content": llm_glm(messages, model=model or "glm-4.7-flash")}),
+            ("openrouter", lambda: llm_openrouter(messages, model=model or "nvidia/nemotron-super-49b-v1:free")),
+            ("openai",     lambda: {"role":"assistant","content": llm_openai(messages, model=model or "gpt-4o-mini")}),
+        ]'''
+
+new = '''    else:
+        providers = [
+            ("groq", lambda: llm_groq(
+                messages,
+                model=model or "llama-3.3-70b-versatile"
+            )),
+            ("cerebras", lambda: {
+                "role":"assistant",
+                "content": llm_cerebras(
+                    messages,
+                    model=model or "gpt-oss-120b"
+                )
+            }),
+            ("openrouter", lambda: llm_openrouter(
+                messages,
+                model=model or os.getenv(
+                    "OPENROUTER_MODEL",
+                    "qwen/qwen3-coder:free"
+                )
+            )),
+        ]'''
+
+if old not in text:
+    raise SystemExit("Provider block not found.")
+
+text = text.replace(old, new)
+
+old2 = '''        except Exception as e:
+            log(f"LLM {name} failed: {e}")
+            continue'''
+
+new2 = '''        except Exception as e:
+            import traceback
+            traceback.print_exc()
+            print(f"[LLM ERROR] {name}: {repr(e)}")
+            log(f"LLM {name} failed: {repr(e)}")
+            continue'''
+
+if old2 not in text:
+    raise SystemExit("Exception block not found.")
+
+text = text.replace(old2, new2)
+
+p.write_text(text)
+
+print("gateway.py patched successfully.")