fixing

src/crewai/llm.py

@@ -67,33 +67,72 @@ class FilteredStream(io.TextIOBase):
             self._lock = threading.Lock()
 
         with self._lock:
-            # Filter out extraneous messages from LiteLLM
+            lower_s = s.lower()
+
+            # Skip common noisy LiteLLM banners and any other lines that contain "litellm"
             if (
-                "Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new"
-                in s
-                or "LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()`"
-                in s
+                "give feedback / get help" in lower_s
+                or "litellm.info:" in lower_s
+                or "litellm" in lower_s
+                or "Consider using a smaller input or implementing a text splitting strategy" in lower_s
             ):
                 return 0
+
             return self._original_stream.write(s)
 
     def flush(self):
         with self._lock:
             return self._original_stream.flush()
 
+    def __getattr__(self, name):
+        """Delegate attribute access to the wrapped original stream.
+
+        This ensures compatibility with libraries (e.g., Rich) that rely on
+        attributes such as `encoding`, `isatty`, `buffer`, etc., which may not
+        be explicitly defined on this proxy class.
+        """
+        return getattr(self._original_stream, name)
+
+    # Delegate common properties/methods explicitly so they aren't shadowed by
+    # the TextIOBase defaults (e.g., .encoding returns None by default, which
+    # confuses Rich). These explicit pass-throughs ensure the wrapped Console
+    # still sees a fully-featured stream.
+    @property
+    def encoding(self):
+        return getattr(self._original_stream, "encoding", "utf-8")
+
+    def isatty(self):
+        return self._original_stream.isatty()
+
+    def fileno(self):
+        return self._original_stream.fileno()
+
+    def writable(self):
+        return True
+
+
+# Apply the filtered stream globally so that any subsequent writes containing the filtered
+# keywords (e.g., "litellm") are hidden from terminal output. We guard against double
+# wrapping to ensure idempotency in environments where this module might be reloaded.
+if not isinstance(sys.stdout, FilteredStream):
+    sys.stdout = FilteredStream(sys.stdout)
+if not isinstance(sys.stderr, FilteredStream):
+    sys.stderr = FilteredStream(sys.stderr)
+
 
 LLM_CONTEXT_WINDOW_SIZES = {
     # openai
     "gpt-4": 8192,
     "gpt-4o": 128000,
-    "gpt-4o-mini": 128000,
+    "gpt-4o-mini": 200000,
     "gpt-4-turbo": 128000,
     "gpt-4.1": 1047576,  # Based on official docs
     "gpt-4.1-mini-2025-04-14": 1047576,
     "gpt-4.1-nano-2025-04-14": 1047576,
     "o1-preview": 128000,
     "o1-mini": 128000,
-    "o3-mini": 200000,  # Based on official o3-mini specifications
+    "o3-mini": 200000,
+    "o4-mini": 200000,
     # gemini
     "gemini-2.0-flash": 1048576,
     "gemini-2.0-flash-thinking-exp-01-21": 32768,
@@ -208,7 +247,7 @@ LLM_CONTEXT_WINDOW_SIZES = {
 }
 
 DEFAULT_CONTEXT_WINDOW_SIZE = 8192
-CONTEXT_WINDOW_USAGE_RATIO = 0.75
+CONTEXT_WINDOW_USAGE_RATIO = 0.85
 
 
 @contextmanager
@@ -219,12 +258,7 @@ def suppress_warnings():
             "ignore", message="open_text is deprecated*", category=DeprecationWarning
         )
 
-        # Redirect stdout and stderr
-        with (
-            redirect_stdout(FilteredStream(sys.stdout)),
-            redirect_stderr(FilteredStream(sys.stderr)),
-        ):
-            yield
+        yield
 
 
 class Delta(TypedDict):

src/crewai/utilities/agent_utils.py

@@ -44,7 +44,7 @@ def render_text_description_and_args(
     tools: Sequence[Union[CrewStructuredTool, BaseTool]],
 ) -> str:
     """Render the tool name, description, and args in plain text.
-    
+
         search: This tool is used for search, args: {"query": {"type": "string"}}
         calculator: This tool is used for math, \
         args: {"expression": {"type": "string"}}
@@ -309,7 +309,7 @@ def handle_context_length(
     """
     if respect_context_window:
         printer.print(
-            content="Context length exceeded. Summarizing content to fit the model context window.",
+            content="Context length exceeded. Summarizing content to fit the model context window. Might take a while...",
             color="yellow",
         )
         summarize_messages(messages, llm, callbacks, i18n)
@@ -337,15 +337,22 @@ def summarize_messages(
         callbacks: List of callbacks for LLM
         i18n: I18N instance for messages
     """
+    messages_string = " ".join([message["content"] for message in messages])
     messages_groups = []
-    for message in messages:
-        content = message["content"]
-        cut_size = llm.get_context_window_size()
-        for i in range(0, len(content), cut_size):
-            messages_groups.append({"content": content[i : i + cut_size]})
+
+    cut_size = llm.get_context_window_size()
+
+    for i in range(0, len(messages_string), cut_size):
+        messages_groups.append({"content": messages_string[i : i + cut_size]})
 
     summarized_contents = []
-    for group in messages_groups:
+
+    total_groups = len(messages_groups)
+    for idx, group in enumerate(messages_groups, 1):
+        Printer().print(
+            content=f"Summarizing {idx}/{total_groups}...",
+            color="yellow",
+        )
         summary = llm.call(
             [
                 format_message_for_llm(