fix: change token counting fallback log from warning to debug (#2220)

Author: opieter-aws

src/strands/models/anthropic.py

@@ -412,7 +412,7 @@ async def count_tokens(
             )
             return total_tokens
         except Exception as e:
-            logger.warning(
+            logger.debug(
                 "model_id=<%s>, error=<%s> | native token counting failed, falling back to estimation",
                 self.config["model_id"],
                 e,

src/strands/models/bedrock.py

@@ -798,7 +798,7 @@ async def count_tokens(
             logger.debug("model_id=<%s>, total_tokens=<%d> | native token count", self.config["model_id"], total_tokens)
             return total_tokens
         except Exception as e:
-            logger.warning(
+            logger.debug(
                 "model_id=<%s>, error=<%s> | native token counting failed, falling back to estimation",
                 self.config["model_id"],
                 e,

src/strands/models/gemini.py

@@ -486,7 +486,7 @@ async def count_tokens(
             )
             return total_tokens
         except Exception as e:
-            logger.warning(
+            logger.debug(
                 "model_id=<%s>, error=<%s> | native token counting failed, falling back to estimation",
                 self.config["model_id"],
                 e,

src/strands/models/llamacpp.py

@@ -555,7 +555,7 @@ async def count_tokens(
             )
             return total_tokens
         except Exception as e:
-            logger.warning(
+            logger.debug(
                 "model_id=<%s>, error=<%s> | native token counting failed, falling back to estimation",
                 self.config.get("model_id", "default"),
                 e,

src/strands/models/openai_responses.py

@@ -226,7 +226,7 @@ async def count_tokens(
             )
             return total_tokens
         except Exception as e:
-            logger.warning(
+            logger.debug(
                 "model_id=<%s>, error=<%s> | native token counting failed, falling back to estimation",
                 self.config["model_id"],
                 e,

tests/strands/models/test_anthropic.py

@@ -1131,10 +1131,10 @@ async def test_fallback_on_generic_exception(self, model_with_client, anthropic_
         assert result >= 0
 
     @pytest.mark.asyncio
-    async def test_fallback_logs_warning(self, model_with_client, anthropic_client, messages, caplog):
+    async def test_fallback_logs_debug(self, model_with_client, anthropic_client, messages, caplog):
         anthropic_client.messages.count_tokens = unittest.mock.AsyncMock(side_effect=RuntimeError("API down"))
 
-        with caplog.at_level(logging.WARNING):
+        with caplog.at_level(logging.DEBUG, logger="strands.models.anthropic"):
             await model_with_client.count_tokens(messages=messages)
 
         assert any("native token counting failed" in record.message for record in caplog.records)

tests/strands/models/test_bedrock.py

@@ -3220,10 +3220,10 @@ async def test_fallback_on_none_input_tokens(self, model_with_client, bedrock_cl
         assert result >= 0
 
     @pytest.mark.asyncio
-    async def test_fallback_logs_warning(self, model_with_client, bedrock_client, messages, caplog):
+    async def test_fallback_logs_debug(self, model_with_client, bedrock_client, messages, caplog):
         bedrock_client.count_tokens.side_effect = RuntimeError("API down")
 
-        with caplog.at_level(logging.WARNING):
+        with caplog.at_level(logging.DEBUG, logger="strands.models.bedrock"):
             await model_with_client.count_tokens(messages=messages)
 
         assert any("native token counting failed" in record.message for record in caplog.records)

tests/strands/models/test_gemini.py

@@ -1197,10 +1197,10 @@ async def test_fallback_on_generic_exception(self, model, gemini_client, message
         assert result >= 0
 
     @pytest.mark.asyncio
-    async def test_fallback_logs_warning(self, model, gemini_client, messages, caplog):
+    async def test_fallback_logs_debug(self, model, gemini_client, messages, caplog):
         gemini_client.aio.models.count_tokens.side_effect = RuntimeError("API down")
 
-        with caplog.at_level(logging.WARNING):
+        with caplog.at_level(logging.DEBUG, logger="strands.models.gemini"):
             await model.count_tokens(messages=messages)
 
         assert any("native token counting failed" in record.message for record in caplog.records)

tests/strands/models/test_llamacpp.py

@@ -796,10 +796,10 @@ async def test_fallback_on_connection_error(self, model, messages):
         assert result >= 0
 
     @pytest.mark.asyncio
-    async def test_fallback_logs_warning(self, model, messages, caplog):
+    async def test_fallback_logs_debug(self, model, messages, caplog):
         model.client.post = AsyncMock(side_effect=RuntimeError("Server down"))
 
-        with caplog.at_level(logging.WARNING):
+        with caplog.at_level(logging.DEBUG, logger="strands.models.llamacpp"):
             await model.count_tokens(messages=messages)
 
         assert any("native token counting failed" in record.message for record in caplog.records)

tests/strands/models/test_openai_responses.py

@@ -1289,12 +1289,12 @@ async def test_fallback_on_generic_exception(self, model, openai_client, message
         assert result >= 0
 
     @pytest.mark.asyncio
-    async def test_fallback_logs_warning(self, model, openai_client, messages, caplog):
+    async def test_fallback_logs_debug(self, model, openai_client, messages, caplog):
         import logging
 
         openai_client.responses.input_tokens.count.side_effect = RuntimeError("API down")
 
-        with caplog.at_level(logging.WARNING):
+        with caplog.at_level(logging.DEBUG, logger="strands.models.openai_responses"):
             await model.count_tokens(messages=messages)
 
         assert any("native token counting failed" in record.message for record in caplog.records)