Improve agent reliability and token efficiency

assistant_bot/ollama.py

@@ -1,5 +1,6 @@
 import asyncio
 import json
+import logging
 import urllib.error
 import urllib.request
 from typing import Any
@@ -7,6 +8,9 @@ from typing import Any
 from .ai import AIClientError
 
 
+logger = logging.getLogger(__name__)
+
+
 class OllamaError(AIClientError):
     pass
 
@@ -52,6 +56,26 @@ class OllamaClient:
             payload["format"] = "json"
 
         data = self._request_json("POST", "/api/chat", payload, timeout=180)
+        prompt_tokens = data.get("prompt_eval_count")
+        completion_tokens = data.get("eval_count")
+        logger.info(
+            "AI usage provider=ollama model=%s input_tokens=%s "
+            "output_tokens=%s total_tokens=%s prompt_eval_ms=%.1f "
+            "generation_ms=%.1f total_ms=%.1f load_ms=%.1f",
+            model,
+            prompt_tokens,
+            completion_tokens,
+            (
+                prompt_tokens + completion_tokens
+                if isinstance(prompt_tokens, int)
+                and isinstance(completion_tokens, int)
+                else None
+            ),
+            self._duration_ms(data.get("prompt_eval_duration")),
+            self._duration_ms(data.get("eval_duration")),
+            self._duration_ms(data.get("total_duration")),
+            self._duration_ms(data.get("load_duration")),
+        )
         content = data.get("message", {}).get("content")
         if isinstance(content, str) and content.strip():
             return content.strip()
@@ -62,6 +86,10 @@ class OllamaClient:
 
         raise OllamaError("Ollama returned an empty response.")
 
+    @staticmethod
+    def _duration_ms(value: Any) -> float:
+        return value / 1_000_000 if isinstance(value, int | float) else 0.0
+
     def _request_json(
         self,
         method: str,