HE SPEAKS?!??!?!?!?!

V3.1/build/ai_bridge_server/server.py

@@ -33,6 +33,7 @@ import time
 import traceback
 import wave
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from urllib.error import URLError
 from urllib.request import Request, urlopen
 
 
@@ -81,8 +82,13 @@ def _ollama_chat(user_text: str) -> str:
         method="POST",
     )
 
-    with urlopen(req, timeout=60) as resp:
-        data = json.loads(resp.read().decode("utf-8"))
+    try:
+        with urlopen(req, timeout=60) as resp:
+            data = json.loads(resp.read().decode("utf-8"))
+    except Exception as e:
+        # Let caller decide how to respond; include a useful hint in logs.
+        _log(f"Ollama request failed url={ollama_url!r} err={e!r}")
+        raise
 
     msg = data.get("message") or {}
     content = msg.get("content")
@@ -91,6 +97,19 @@ def _ollama_chat(user_text: str) -> str:
     return content.strip()
 
 
+def _short_err(e: BaseException) -> str:
+    s = str(e) or e.__class__.__name__
+    s = " ".join(s.split())
+    if len(s) > 240:
+        s = s[:240] + "..."
+    return s
+
+
+def _ollama_down_reply() -> str:
+    # Keep it short and speakable.
+    return "My AI server isn't reachable right now. Please start Ollama on the computer, then try again."
+
+
 def _wav_diagnostics(wav_bytes: bytes) -> dict:
     """Best-effort WAV parsing + signal stats for debugging mic capture."""
     info: dict = {"bytes": len(wav_bytes)}
@@ -245,9 +264,16 @@ class Handler(BaseHTTPRequestHandler):
                     _json_response(self, 400, {"error": "Missing 'text'"})
                     return
                 _log(f"{client} /text prompt_chars={len(text)} prompt={text[:200]!r}")
-                reply = _ollama_chat(text)
-                _log(f"{client} /text ok reply_chars={len(reply)}")
-                _json_response(self, 200, {"reply": reply})
+                try:
+                    reply = _ollama_chat(text)
+                    _log(f"{client} /text ok reply_chars={len(reply)}")
+                    _json_response(self, 200, {"reply": reply})
+                except URLError as e:
+                    _log(f"{client} /text ollama_unreachable err={_short_err(e)!r}")
+                    _json_response(self, 200, {"reply": _ollama_down_reply(), "ollama_ok": False, "ollama_error": _short_err(e)})
+                except ConnectionRefusedError as e:
+                    _log(f"{client} /text ollama_refused err={_short_err(e)!r}")
+                    _json_response(self, 200, {"reply": _ollama_down_reply(), "ollama_ok": False, "ollama_error": _short_err(e)})
                 return
 
             if self.path == "/v1/chat/audio":
@@ -295,9 +321,24 @@ class Handler(BaseHTTPRequestHandler):
                     _json_response(self, 200, {"reply": "I didn't catch that. Could you say it again?", "text": ""})
                     return
                 _log(f"{client} /audio transcript_chars={len(transcript)} transcript={transcript[:200]!r}")
-                reply = _ollama_chat(transcript)
-                _log(f"{client} /audio ok text_chars={len(transcript)} reply_chars={len(reply)}")
-                _json_response(self, 200, {"reply": reply, "text": transcript})
+                try:
+                    reply = _ollama_chat(transcript)
+                    _log(f"{client} /audio ok text_chars={len(transcript)} reply_chars={len(reply)}")
+                    _json_response(self, 200, {"reply": reply, "text": transcript})
+                except URLError as e:
+                    _log(f"{client} /audio ollama_unreachable err={_short_err(e)!r}")
+                    _json_response(
+                        self,
+                        200,
+                        {"reply": _ollama_down_reply(), "text": transcript, "ollama_ok": False, "ollama_error": _short_err(e)},
+                    )
+                except ConnectionRefusedError as e:
+                    _log(f"{client} /audio ollama_refused err={_short_err(e)!r}")
+                    _json_response(
+                        self,
+                        200,
+                        {"reply": _ollama_down_reply(), "text": transcript, "ollama_ok": False, "ollama_error": _short_err(e)},
+                    )
                 return
 
             _json_response(self, 404, {"error": "Not found"})
@@ -332,6 +373,7 @@ def main():
         + " model="
         + os.environ.get("OLLAMA_MODEL", "phi3.5")
     )
+    _log("Ollama health check: curl -s http://127.0.0.1:11434/api/tags | head")
     if not _whisper.available():
         _log("STT: faster-whisper not installed; /v1/chat/audio will return 503")
     server.serve_forever()