feat: debugging asr and minicpm

app.py

@@ -1275,7 +1275,8 @@ with gr.Blocks(title="CityQuest · AI") as demo:
             with gr.Row():
                 with gr.Column():
                     gr.HTML("<div class='section-label'>Voice journal</div>")
-                    d_audio = gr.Audio(sources=["microphone"], type="filepath", label="Record")
+                    d_audio = gr.Audio(sources=["microphone", "upload"], type="filepath",
+                                       format="wav", label="Record or upload audio")
                     d_lang = gr.Dropdown(
                         ["en", "fr", "de", "it", "es", "pt", "el", "nl", "pl",
                          "zh", "ja", "ko", "vi", "ar"],

app/services/minicpm.py

@@ -123,7 +123,22 @@ def warm() -> bool:
     return _get_llm() is not None
 
 
-def ask(question: str, context: str = "", max_tokens: int = 320) -> Optional[str]:
+def _strip_think(text: str) -> str:
+    """Remove MiniCPM5 reasoning blocks.
+
+    MiniCPM5 is a reasoning model that emits ``<think>…</think>`` before its
+    answer. Those tags render invisibly in Markdown (they look like HTML), so
+    we drop the whole block and keep only the final answer. Also handles an
+    unclosed ``<think>`` (answer was truncated by max_tokens).
+    """
+    import re
+    text = re.sub(r"<think>.*?</think>", "", text, flags=re.DOTALL | re.IGNORECASE)
+    # If a think block was opened but never closed (truncation), drop the rest.
+    text = re.sub(r"<think>.*$", "", text, flags=re.DOTALL | re.IGNORECASE)
+    return text.strip()
+
+
+def ask(question: str, context: str = "", max_tokens: int = 640) -> Optional[str]:
     """In-game MiniCPM helper: answer a player's question (clue / explain a
     task / tell me about this place).  Returns plain text, or ``None`` if the
     model is unavailable so the caller can show a graceful fallback."""
@@ -134,7 +149,7 @@ def ask(question: str, context: str = "", max_tokens: int = 320) -> Optional[str
         "You are CityQuest's friendly in-game guide for a real-world city "
         "adventure game. Give a short, warm, practical answer (2-4 sentences). "
         "Help with clues, understanding tasks, and local places. Never reveal "
-        "winners or scores. Plain text only."
+        "winners or scores. Plain text only. Do not show your reasoning."
     )
     user = question if not context else f"Context:\n{context}\n\nQuestion: {question}"
 
@@ -145,8 +160,9 @@ def ask(question: str, context: str = "", max_tokens: int = 320) -> Optional[str
                       {"role": "user", "content": user}],
             max_tokens=max_tokens, temperature=0.6, top_p=0.9,
         )
-        text = (result["choices"][0]["message"]["content"] or "").strip()
-        print(f"[minicpm] ask(chat) len={len(text)} preview={text[:80]!r}")
+        raw = (result["choices"][0]["message"]["content"] or "").strip()
+        text = _strip_think(raw)
+        print(f"[minicpm] ask(chat) raw_len={len(raw)} answer_len={len(text)} preview={text[:80]!r}")
         if text:
             return text
     except Exception as e:
@@ -164,8 +180,8 @@ def ask(question: str, context: str = "", max_tokens: int = 320) -> Optional[str
             prompt, max_tokens=max_tokens, temperature=0.6, top_p=0.9,
             stop=["<|im_end|>", "<|im_start|>", "</s>"],
         )
-        text = (out["choices"][0]["text"] or "").strip()
-        print(f"[minicpm] ask(raw) len={len(text)} preview={text[:80]!r}")
+        text = _strip_think((out["choices"][0]["text"] or "").strip())
+        print(f"[minicpm] ask(raw) answer_len={len(text)} preview={text[:80]!r}")
         return text or None
     except Exception as e:
         print(f"[minicpm] ask(raw) failed: {type(e).__name__}: {e}")