agno 2.1.2__py3-none-any.whl → 2.3.13__py3-none-any.whl

agno/utils/models/llama.py

@@ -19,13 +19,17 @@ TOOL_CALL_ROLE_MAP = {
 }
 
 
-def format_message(message: Message, openai_like: bool = False, tool_calls: bool = False) -> Dict[str, Any]:
+def format_message(
+    message: Message, openai_like: bool = False, tool_calls: bool = False, compress_tool_results: bool = False
+) -> Dict[str, Any]:
     """
     Format a message into the format expected by Llama API.
 
     Args:
         message (Message): The message to format.
         openai_like (bool): Whether to format the message as an OpenAI-like message.
+        tool_calls (bool): Whether tool calls are present.
+        compress_tool_results: Whether to compress tool results.
 
     Returns:
         Dict[str, Any]: The formatted message.
@@ -52,10 +56,13 @@ def format_message(message: Message, openai_like: bool = False, tool_calls: bool
         log_warning("Audio input is currently unsupported.")
 
     if message.role == "tool":
+        # Use compressed content if compression is active
+        content = message.get_content(use_compressed_content=compress_tool_results)
+
         message_dict = {
             "role": "tool",
             "tool_call_id": message.tool_call_id,
-            "content": message.content,
+            "content": content,
         }
 
     if message.role == "assistant":

agno/utils/models/mistral.py

@@ -48,7 +48,7 @@ def _format_image_for_message(image: Image) -> Optional[ImageURLChunk]:
     return None
 
 
-def format_messages(messages: List[Message]) -> List[MistralMessage]:
+def format_messages(messages: List[Message], compress_tool_results: bool = False) -> List[MistralMessage]:
     mistral_messages: List[MistralMessage] = []
 
     for message in messages:
@@ -84,7 +84,9 @@ def format_messages(messages: List[Message]) -> List[MistralMessage]:
         elif message.role == "system":
             mistral_message = SystemMessage(role="system", content=message.content)
         elif message.role == "tool":
-            mistral_message = ToolMessage(name="tool", content=message.content, tool_call_id=message.tool_call_id)
+            # Get compressed content if compression is active
+            tool_content = message.get_content(use_compressed_content=compress_tool_results)
+            mistral_message = ToolMessage(name="tool", content=tool_content, tool_call_id=message.tool_call_id)
         else:
             raise ValueError(f"Unknown role: {message.role}")
 

agno/utils/print_response/agent.py

@@ -1,4 +1,5 @@
 import json
+import warnings
 from collections.abc import Set
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Union, cast, get_args
 
@@ -10,6 +11,7 @@ from rich.markdown import Markdown
 from rich.status import Status
 from rich.text import Text
 
+from agno.filters import FilterExpr
 from agno.media import Audio, File, Image, Video
 from agno.models.message import Message
 from agno.reasoning.step import ReasoningStep
@@ -29,12 +31,12 @@ def print_response_stream(
     session_id: Optional[str] = None,
     session_state: Optional[Dict[str, Any]] = None,
     user_id: Optional[str] = None,
+    run_id: Optional[str] = None,
     audio: Optional[Sequence[Audio]] = None,
     images: Optional[Sequence[Image]] = None,
     videos: Optional[Sequence[Video]] = None,
     files: Optional[Sequence[File]] = None,
-    stream_intermediate_steps: bool = False,
-    knowledge_filters: Optional[Dict[str, Any]] = None,
+    knowledge_filters: Optional[Union[Dict[str, Any], List[FilterExpr]]] = None,
     debug_mode: Optional[bool] = None,
     markdown: bool = False,
     show_message: bool = True,
@@ -85,12 +87,12 @@ def print_response_stream(
             session_id=session_id,
             session_state=session_state,
             user_id=user_id,
+            run_id=run_id,
             audio=audio,
             images=images,
             videos=videos,
             files=files,
             stream=True,
-            stream_intermediate_steps=stream_intermediate_steps,
             knowledge_filters=knowledge_filters,
             debug_mode=debug_mode,
             add_history_to_context=add_history_to_context,
@@ -179,6 +181,7 @@ def print_response_stream(
                 show_reasoning=show_reasoning,
                 show_full_reasoning=show_full_reasoning,
                 accumulated_tool_calls=accumulated_tool_calls,
+                compression_manager=agent.compression_manager,
             )
             panels.extend(additional_panels)
             if panels:
@@ -204,6 +207,10 @@ def print_response_stream(
             live_log.update(Group(*panels))
             agent.session_summary_manager.summaries_updated = False
 
+        # Clear compression stats after final display
+        if agent.compression_manager is not None:
+            agent.compression_manager.stats.clear()
+
         response_timer.stop()
 
         # Final update to remove the "Thinking..." status
@@ -217,12 +224,12 @@ async def aprint_response_stream(
     session_id: Optional[str] = None,
     session_state: Optional[Dict[str, Any]] = None,
     user_id: Optional[str] = None,
+    run_id: Optional[str] = None,
     audio: Optional[Sequence[Audio]] = None,
     images: Optional[Sequence[Image]] = None,
     videos: Optional[Sequence[Video]] = None,
     files: Optional[Sequence[File]] = None,
-    stream_intermediate_steps: bool = False,
-    knowledge_filters: Optional[Dict[str, Any]] = None,
+    knowledge_filters: Optional[Union[Dict[str, Any], List[FilterExpr]]] = None,
     debug_mode: Optional[bool] = None,
     markdown: bool = False,
     show_message: bool = True,
@@ -271,12 +278,12 @@ async def aprint_response_stream(
             session_id=session_id,
             session_state=session_state,
             user_id=user_id,
+            run_id=run_id,
             audio=audio,
             images=images,
             videos=videos,
             files=files,
             stream=True,
-            stream_intermediate_steps=stream_intermediate_steps,
             knowledge_filters=knowledge_filters,
             debug_mode=debug_mode,
             add_history_to_context=add_history_to_context,
@@ -368,6 +375,7 @@ async def aprint_response_stream(
                 show_reasoning=show_reasoning,
                 show_full_reasoning=show_full_reasoning,
                 accumulated_tool_calls=accumulated_tool_calls,
+                compression_manager=agent.compression_manager,
             )
             panels.extend(additional_panels)
             if panels:
@@ -393,6 +401,10 @@ async def aprint_response_stream(
             live_log.update(Group(*panels))
             agent.session_summary_manager.summaries_updated = False
 
+        # Clear compression stats after final display
+        if agent.compression_manager is not None:
+            agent.compression_manager.stats.clear()
+
         response_timer.stop()
 
         # Final update to remove the "Thinking..." status
@@ -409,6 +421,7 @@ def build_panels_stream(
     show_reasoning: bool = True,
     show_full_reasoning: bool = False,
     accumulated_tool_calls: Optional[List] = None,
+    compression_manager: Optional[Any] = None,
 ):
     panels = []
 
@@ -449,8 +462,18 @@ def build_panels_stream(
         for formatted_tool_call in formatted_tool_calls:
             tool_calls_content.append(f"• {formatted_tool_call}\n")
 
+        tool_calls_text = tool_calls_content.plain.rstrip()
+
+        # Add compression stats if available (don't clear - caller will clear after final display)
+        if compression_manager is not None and compression_manager.stats:
+            stats = compression_manager.stats
+            saved = stats.get("original_size", 0) - stats.get("compressed_size", 0)
+            orig = stats.get("original_size", 1)
+            if stats.get("tool_results_compressed", 0) > 0:
+                tool_calls_text += f"\n\ncompressed: {stats.get('tool_results_compressed', 0)} | Saved: {saved:,} chars ({saved / orig * 100:.0f}%)"
+
         tool_calls_panel = create_panel(
-            content=tool_calls_content.plain.rstrip(),
+            content=tool_calls_text,
             title="Tool Calls",
             border_style="yellow",
         )
@@ -471,11 +494,23 @@ def build_panels_stream(
         and response_event.citations is not None
         and response_event.citations.urls is not None
     ):
-        md_content = "\n".join(
+        md_lines = []
+
+        # Add search queries if present
+        if response_event.citations.search_queries:
+            md_lines.append("**Search Queries:**")
+            for query in response_event.citations.search_queries:
+                md_lines.append(f"- {query}")
+            md_lines.append("")  # Empty line before URLs
+
+        # Add URL citations
+        md_lines.extend(
             f"{i + 1}. [{citation.title or citation.url}]({citation.url})"
             for i, citation in enumerate(response_event.citations.urls)
             if citation.url  # Only include citations with valid URLs
         )
+
+        md_content = "\n".join(md_lines)
         if md_content:  # Only create panel if there are citations
             citations_panel = create_panel(
                 content=Markdown(md_content),
@@ -493,12 +528,14 @@ def print_response(
     session_id: Optional[str] = None,
     session_state: Optional[Dict[str, Any]] = None,
     user_id: Optional[str] = None,
+    run_id: Optional[str] = None,
     audio: Optional[Sequence[Audio]] = None,
     images: Optional[Sequence[Image]] = None,
     videos: Optional[Sequence[Video]] = None,
     files: Optional[Sequence[File]] = None,
-    stream_intermediate_steps: bool = False,
-    knowledge_filters: Optional[Dict[str, Any]] = None,
+    knowledge_filters: Optional[Union[Dict[str, Any], List[FilterExpr]]] = None,
+    stream_events: Optional[bool] = None,
+    stream_intermediate_steps: Optional[bool] = None,
     debug_mode: Optional[bool] = None,
     markdown: bool = False,
     show_message: bool = True,
@@ -513,6 +550,19 @@ def print_response(
     metadata: Optional[Dict[str, Any]] = None,
     **kwargs: Any,
 ):
+    if stream_events is not None:
+        warnings.warn(
+            "The 'stream_events' parameter is deprecated and will be removed in future versions. Event streaming is always enabled using the print_response function.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+    if stream_intermediate_steps is not None:
+        warnings.warn(
+            "The 'stream_intermediate_steps' parameter is deprecated and will be removed in future versions. Event streaming is always enabled using the print_response function.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+
     with Live(console=console) as live_log:
         status = Status("Thinking...", spinner="aesthetic", speed=0.4, refresh_per_second=10)
         live_log.update(status)
@@ -538,12 +588,13 @@ def print_response(
             session_id=session_id,
             session_state=session_state,
             user_id=user_id,
+            run_id=run_id,
             audio=audio,
             images=images,
             videos=videos,
             files=files,
             stream=False,
-            stream_intermediate_steps=stream_intermediate_steps,
+            stream_events=True,
             knowledge_filters=knowledge_filters,
             debug_mode=debug_mode,
             add_history_to_context=add_history_to_context,
@@ -577,6 +628,7 @@ def print_response(
             show_full_reasoning=show_full_reasoning,
             tags_to_include_in_markdown=tags_to_include_in_markdown,
             markdown=markdown,
+            compression_manager=agent.compression_manager,
         )
         panels.extend(additional_panels)
 
@@ -611,17 +663,19 @@ async def aprint_response(
     session_id: Optional[str] = None,
     session_state: Optional[Dict[str, Any]] = None,
     user_id: Optional[str] = None,
+    run_id: Optional[str] = None,
     audio: Optional[Sequence[Audio]] = None,
     images: Optional[Sequence[Image]] = None,
     videos: Optional[Sequence[Video]] = None,
     files: Optional[Sequence[File]] = None,
-    stream_intermediate_steps: bool = False,
-    knowledge_filters: Optional[Dict[str, Any]] = None,
+    knowledge_filters: Optional[Union[Dict[str, Any], List[FilterExpr]]] = None,
     debug_mode: Optional[bool] = None,
     markdown: bool = False,
     show_message: bool = True,
     show_reasoning: bool = True,
     show_full_reasoning: bool = False,
+    stream_events: Optional[bool] = None,
+    stream_intermediate_steps: Optional[bool] = None,
     tags_to_include_in_markdown: Set[str] = {"think", "thinking"},
     console: Optional[Any] = None,
     add_history_to_context: Optional[bool] = None,
@@ -631,6 +685,19 @@ async def aprint_response(
     metadata: Optional[Dict[str, Any]] = None,
     **kwargs: Any,
 ):
+    if stream_events is not None:
+        warnings.warn(
+            "The 'stream_events' parameter is deprecated and will be removed in future versions. Event streaming is always enabled using the aprint_response function.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+    if stream_intermediate_steps is not None:
+        warnings.warn(
+            "The 'stream_intermediate_steps' parameter is deprecated and will be removed in future versions. Event streaming is always enabled using the aprint_response function.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+
     with Live(console=console) as live_log:
         status = Status("Thinking...", spinner="aesthetic", speed=0.4, refresh_per_second=10)
         live_log.update(status)
@@ -656,12 +723,13 @@ async def aprint_response(
             session_id=session_id,
             session_state=session_state,
             user_id=user_id,
+            run_id=run_id,
             audio=audio,
             images=images,
             videos=videos,
             files=files,
             stream=False,
-            stream_intermediate_steps=stream_intermediate_steps,
+            stream_events=True,
             knowledge_filters=knowledge_filters,
             debug_mode=debug_mode,
             add_history_to_context=add_history_to_context,
@@ -695,6 +763,7 @@ async def aprint_response(
             show_full_reasoning=show_full_reasoning,
             tags_to_include_in_markdown=tags_to_include_in_markdown,
             markdown=markdown,
+            compression_manager=agent.compression_manager,
         )
         panels.extend(additional_panels)
 
@@ -731,6 +800,7 @@ def build_panels(
     show_full_reasoning: bool = False,
     tags_to_include_in_markdown: Optional[Set[str]] = None,
     markdown: bool = False,
+    compression_manager: Optional[Any] = None,
 ):
     panels = []
 
@@ -782,8 +852,19 @@ def build_panels(
         for formatted_tool_call in formatted_tool_calls:
             tool_calls_content.append(f"• {formatted_tool_call}\n")
 
+        tool_calls_text = tool_calls_content.plain.rstrip()
+
+        # Add compression stats if available
+        if compression_manager is not None and compression_manager.stats:
+            stats = compression_manager.stats
+            saved = stats.get("original_size", 0) - stats.get("compressed_size", 0)
+            orig = stats.get("original_size", 1)
+            if stats.get("tool_results_compressed", 0) > 0:
+                tool_calls_text += f"\n\ncompressed: {stats.get('tool_results_compressed', 0)} | Saved: {saved:,} chars ({saved / orig * 100:.0f}%)"
+            compression_manager.stats.clear()
+
         tool_calls_panel = create_panel(
-            content=tool_calls_content.plain.rstrip(),
+            content=tool_calls_text,
             title="Tool Calls",
             border_style="yellow",
         )
@@ -821,11 +902,23 @@ def build_panels(
         and run_response.citations is not None
         and run_response.citations.urls is not None
     ):
-        md_content = "\n".join(
+        md_lines = []
+
+        # Add search queries if present
+        if run_response.citations.search_queries:
+            md_lines.append("**Search Queries:**")
+            for query in run_response.citations.search_queries:
+                md_lines.append(f"- {query}")
+            md_lines.append("")  # Empty line before URLs
+
+        # Add URL citations
+        md_lines.extend(
             f"{i + 1}. [{citation.title or citation.url}]({citation.url})"
             for i, citation in enumerate(run_response.citations.urls)
             if citation.url  # Only include citations with valid URLs
         )
+
+        md_content = "\n".join(md_lines)
         if md_content:  # Only create panel if there are citations
             citations_panel = create_panel(
                 content=Markdown(md_content),