fast-agent-mcp 0.3.6__py3-none-any.whl → 0.3.8__py3-none-any.whl

fast_agent/llm/fastagent_llm.py

@@ -198,7 +198,7 @@ class FastAgentLLM(ContextDependent, FastAgentLLMProtocol, Generic[MessageParamT
         if messages[-1].first_text().startswith("***SAVE_HISTORY"):
             parts: list[str] = messages[-1].first_text().split(" ", 1)
             filename: str = (
-                parts[1].strip() if len(parts) > 1 else f"{self.name or 'assistant'}_prompts.txt"
+                parts[1].strip() if len(parts) > 1 else f"{self.name or 'assistant'}.json"
             )
             await self._save_history(filename)
             return Prompt.assistant(f"History saved to {filename}")
@@ -589,10 +589,10 @@ class FastAgentLLM(ContextDependent, FastAgentLLMProtocol, Generic[MessageParamT
         Uses JSON format for .json files (MCP SDK compatible format) and
         delimited text format for other extensions.
         """
-        from fast_agent.mcp.prompt_serialization import save_messages_to_file
+        from fast_agent.mcp.prompt_serialization import save_messages
 
         # Save messages using the unified save function that auto-detects format
-        save_messages_to_file(self._message_history, filename)
+        save_messages(self._message_history, filename)
 
     @property
     def message_history(self) -> List[PromptMessageExtended]:

fast_agent/llm/model_database.py

@@ -164,7 +164,11 @@ class ModelDatabase:
     )
 
     # FIXME: xAI has not documented the max output tokens for Grok 4. Using Grok 3 as a placeholder. Will need to update when available (if ever)
-    GROK_4 = ModelParameters(context_window=256000, max_output_tokens=16385, tokenizes=XAI_VISION)
+    GROK_4 = ModelParameters(context_window=256000, max_output_tokens=16385, tokenizes=TEXT_ONLY)
+
+    GROK_4_VLM = ModelParameters(
+        context_window=2000000, max_output_tokens=16385, tokenizes=XAI_VISION
+    )
 
     # Source for Grok 3 max output: https://www.reddit.com/r/grok/comments/1j7209p/exploring_grok_3_beta_output_capacity_a_simple/
     # xAI does not document Grok 3 max output tokens, using the above source as a reference.
@@ -240,6 +244,8 @@ class ModelDatabase:
         "gemini-2.5-flash-preview-05-20": GEMINI_FLASH,
         "gemini-2.5-pro-preview-05-06": GEMINI_PRO,
         # xAI Grok Models
+        "grok-4-fast-reasoning": GROK_4_VLM,
+        "grok-4-fast-non-reasoning": GROK_4_VLM,
         "grok-4": GROK_4,
         "grok-4-0709": GROK_4,
         "grok-3": GROK_3,

fast_agent/llm/model_factory.py

@@ -12,9 +12,6 @@ from fast_agent.llm.internal.slow import SlowLLM
 from fast_agent.llm.provider_types import Provider
 from fast_agent.types import RequestParams
 
-# from fast_agent.workflows.llm.augmented_llm_deepseek import DeekSeekAugmentedLLM
-
-
 # Type alias for LLM classes
 LLMClass = Union[Type[PassthroughLLM], Type[PlaybackLLM], Type[SilentLLM], Type[SlowLLM], type]
 
@@ -123,6 +120,8 @@ class ModelFactory:
         "kimi": "groq.moonshotai/kimi-k2-instruct-0905",
         "gpt-oss": "groq.openai/gpt-oss-120b",
         "gpt-oss-20b": "groq.openai/gpt-oss-20b",
+        "grok-4-fast": "xai.grok-4-fast-non-reasoning",
+        "grok-4-fast-reasoning": "xai.grok-4-fast-reasoning",
     }
 
     @staticmethod

fast_agent/llm/provider/bedrock/llm_bedrock.py

@@ -126,7 +126,7 @@ class ModelCapabilities:
 
 class BedrockLLM(FastAgentLLM[BedrockMessageParam, BedrockMessage]):
     """
-    AWS Bedrock implementation of AugmentedLLM using the Converse API.
+    AWS Bedrock implementation of FastAgentLLM using the Converse API.
     Supports all Bedrock models including Nova, Claude, Meta, etc.
     """
 

fast_agent/llm/provider/google/llm_google_native.py

@@ -36,10 +36,8 @@ GOOGLE_EXCLUDE_FIELDS = {
     FastAgentLLM.PARAM_MESSAGES,  # Handled by contents
     FastAgentLLM.PARAM_MODEL,  # Handled during client/call setup
     FastAgentLLM.PARAM_SYSTEM_PROMPT,  # Handled by system_instruction in config
-    # AugmentedLLM.PARAM_PARALLEL_TOOL_CALLS, # Handled by tool_config in config
-    FastAgentLLM.PARAM_USE_HISTORY,  # Handled by AugmentedLLM base / this class's logic
+    FastAgentLLM.PARAM_USE_HISTORY,  # Handled by FastAgentLLM base / this class's logic
     FastAgentLLM.PARAM_MAX_ITERATIONS,  # Handled by this class's loop
-    # Add any other OpenAI-specific params not applicable to google.genai
     FastAgentLLM.PARAM_MCP_METADATA,
 }.union(FastAgentLLM.BASE_EXCLUDE_FIELDS)
 

fast_agent/llm/provider/openai/llm_azure.py

@@ -23,7 +23,7 @@ DEFAULT_AZURE_API_VERSION = "2024-10-21"
 
 class AzureOpenAILLM(OpenAILLM):
     """
-    Azure OpenAI implementation extending OpenAIAugmentedLLM.
+    Azure OpenAI implementation extending OpenAILLM.
     Handles both API Key and DefaultAzureCredential authentication.
     """
 

fast_agent/llm/provider/openai/llm_openai.py

@@ -7,7 +7,7 @@ from mcp.types import (
     ContentBlock,
     TextContent,
 )
-from openai import AsyncOpenAI, AuthenticationError
+from openai import APIError, AsyncOpenAI, AuthenticationError
 from openai.lib.streaming.chat import ChatCompletionStreamState
 
 # from openai.types.beta.chat import
@@ -19,19 +19,17 @@ from openai.types.chat import (
 )
 from pydantic_core import from_json
 
+from fast_agent.constants import FAST_AGENT_ERROR_CHANNEL
 from fast_agent.core.exceptions import ProviderKeyError
 from fast_agent.core.logging.logger import get_logger
 from fast_agent.core.prompt import Prompt
 from fast_agent.event_progress import ProgressAction
-from fast_agent.llm.fastagent_llm import (
-    FastAgentLLM,
-    RequestParams,
-)
+from fast_agent.llm.fastagent_llm import FastAgentLLM, RequestParams
 from fast_agent.llm.provider.openai.multipart_converter_openai import OpenAIConverter, OpenAIMessage
 from fast_agent.llm.provider_types import Provider
 from fast_agent.llm.usage_tracking import TurnUsage
-from fast_agent.types import PromptMessageExtended
-from fast_agent.types.llm_stop_reason import LlmStopReason
+from fast_agent.mcp.helpers.content_helpers import text_content
+from fast_agent.types import LlmStopReason, PromptMessageExtended
 
 _logger = get_logger(__name__)
 
@@ -348,7 +346,11 @@ class OpenAILLM(FastAgentLLM[ChatCompletionMessageParam, ChatCompletionMessage])
         # Use basic streaming API
         stream = await self._openai_client().chat.completions.create(**arguments)
         # Process the stream
-        response = await self._process_stream(stream, model_name)
+        try:
+            response = await self._process_stream(stream, model_name)
+        except APIError as error:
+            self.logger.error("Streaming APIError during OpenAI completion", exc_info=error)
+            return self._stream_failure_response(error, model_name)
         # Track usage if response is valid and has usage data
         if (
             hasattr(response, "usage")
@@ -438,6 +440,53 @@ class OpenAILLM(FastAgentLLM[ChatCompletionMessageParam, ChatCompletionMessage])
             *response_content_blocks, stop_reason=stop_reason, tool_calls=requested_tool_calls
         )
 
+    def _stream_failure_response(self, error: APIError, model_name: str) -> PromptMessageExtended:
+        """Convert streaming API errors into a graceful assistant reply."""
+
+        provider_label = (
+            self.provider.value if isinstance(self.provider, Provider) else str(self.provider)
+        )
+        detail = getattr(error, "message", None) or str(error)
+        detail = detail.strip() if isinstance(detail, str) else ""
+
+        parts: list[str] = [f"{provider_label} request failed"]
+        if model_name:
+            parts.append(f"for model '{model_name}'")
+        code = getattr(error, "code", None)
+        if code:
+            parts.append(f"(code: {code})")
+        status = getattr(error, "status_code", None)
+        if status:
+            parts.append(f"(status={status})")
+
+        message = " ".join(parts)
+        if detail:
+            message = f"{message}: {detail}"
+
+        user_summary = " ".join(message.split()) if message else ""
+        if user_summary and len(user_summary) > 280:
+            user_summary = user_summary[:277].rstrip() + "..."
+
+        if user_summary:
+            assistant_text = f"I hit an internal error while calling the model: {user_summary}"
+            if not assistant_text.endswith((".", "!", "?")):
+                assistant_text += "."
+            assistant_text += " See fast-agent-error for additional details."
+        else:
+            assistant_text = (
+                "I hit an internal error while calling the model; see fast-agent-error for details."
+            )
+
+        assistant_block = text_content(assistant_text)
+        error_block = text_content(message)
+
+        return PromptMessageExtended(
+            role="assistant",
+            content=[assistant_block],
+            channels={FAST_AGENT_ERROR_CHANNEL: [error_block]},
+            stop_reason=LlmStopReason.ERROR,
+        )
+
     async def _is_tool_stop_reason(self, finish_reason: str) -> bool:
         return True
 

fast_agent/llm/provider/openai/llm_tensorzero_openai.py

@@ -26,7 +26,7 @@ class TensorZeroOpenAILLM(OpenAILLM):
         self._t0_function_name = kwargs.get("model", "")
 
         super().__init__(*args, provider=Provider.TENSORZERO, **kwargs)
-        self.logger.info("TensorZeroOpenAIAugmentedLLM initialized.")
+        self.logger.info("TensorZeroOpenAILLM initialized.")
 
     def _initialize_default_params(self, kwargs: dict) -> RequestParams:
         """

fast_agent/llm/request_params.py

@@ -11,7 +11,7 @@ from pydantic import Field
 
 class RequestParams(CreateMessageRequestParams):
     """
-    Parameters to configure the AugmentedLLM 'generate' requests.
+    Parameters to configure the FastAgentLLM 'generate' requests.
     """
 
     messages: List[SamplingMessage] = Field(exclude=True, default=[])

fast_agent/mcp/__init__.py

@@ -24,11 +24,9 @@ from .helpers import (
     split_thinking_content,
     text_content,
 )
-from .prompt_message_extended import PromptMessageExtended
 
 __all__ = [
     "Prompt",
-    "PromptMessageExtended",
     # Helpers
     "get_text",
     "get_image_data",
@@ -51,4 +49,5 @@ def __getattr__(name: str):
         from .prompt import Prompt  # local import
 
         return Prompt
+
     raise AttributeError(f"module '{__name__}' has no attribute '{name}'")

fast_agent/mcp/mcp_aggregator.py

@@ -470,12 +470,13 @@ class MCPAggregator(ContextDependent):
         """
         instructions = {}
 
-        if self.connection_persistence and hasattr(self, '_persistent_connection_manager'):
+        if self.connection_persistence and hasattr(self, "_persistent_connection_manager"):
             # Get instructions from persistent connections
             for server_name in self.server_names:
                 try:
                     server_conn = await self._persistent_connection_manager.get_server(
-                        server_name, client_session_factory=self._create_session_factory(server_name)
+                        server_name,
+                        client_session_factory=self._create_session_factory(server_name),
                     )
                     # Always include server, even if no instructions
                     # Get tool names for this server
@@ -1074,7 +1075,9 @@ class MCPAggregator(ContextDependent):
             logger.debug(f"Server '{server_name}' does not support tools")
             return
 
-        await self.display.show_tool_update(aggregator=self, updated_server=server_name)
+        await self.display.show_tool_update(
+            updated_server=server_name, agent_name="Tool List Change Notification"
+        )
 
         async with self._refresh_lock:
             try:

fast_agent/mcp/prompt_message_extended.py

@@ -12,6 +12,8 @@ from mcp.types import (
 from pydantic import BaseModel
 
 from fast_agent.mcp.helpers.content_helpers import get_text
+
+# Import directly to avoid circular dependency with types/__init__.py
 from fast_agent.types.llm_stop_reason import LlmStopReason
 
 

fast_agent/mcp/prompt_serialization.py

@@ -23,6 +23,7 @@ from mcp.types import (
     EmbeddedResource,
     GetPromptResult,
     ImageContent,
+    PromptMessage,
     TextContent,
     TextResourceContents,
 )
@@ -34,12 +35,30 @@ from fast_agent.mcp.prompts.prompt_constants import (
 )
 from fast_agent.types import PromptMessageExtended
 
+# -------------------------------------------------------------------------
+# Serialization Helpers
+# -------------------------------------------------------------------------
+
+
+def serialize_to_dict(obj, exclude_none: bool = True):
+    """Standardized Pydantic serialization to dictionary.
+
+    Args:
+        obj: Pydantic model object to serialize
+        exclude_none: Whether to exclude None values (default: True)
+
+    Returns:
+        Dictionary representation suitable for JSON serialization
+    """
+    return obj.model_dump(by_alias=True, mode="json", exclude_none=exclude_none)
+
+
 # -------------------------------------------------------------------------
 # JSON Serialization Functions
 # -------------------------------------------------------------------------
 
 
-def multipart_messages_to_get_prompt_result(
+def to_get_prompt_result(
     messages: List[PromptMessageExtended],
 ) -> GetPromptResult:
     """
@@ -60,35 +79,58 @@ def multipart_messages_to_get_prompt_result(
     return GetPromptResult(messages=flat_messages)
 
 
-def multipart_messages_to_json(messages: List[PromptMessageExtended]) -> str:
+
+def to_get_prompt_result_json(messages: List[PromptMessageExtended]) -> str:
+    """
+    Convert PromptMessageExtended objects to MCP-compatible GetPromptResult JSON.
+
+    This is a lossy conversion that flattens multipart messages and loses extended fields
+    like tool_calls, channels, and stop_reason. Use for MCP server compatibility.
+
+    Args:
+        messages: List of PromptMessageExtended objects
+
+    Returns:
+        JSON string in GetPromptResult format
+    """
+    result = to_get_prompt_result(messages)
+    result_dict = serialize_to_dict(result)
+    return json.dumps(result_dict, indent=2)
+
+
+def to_json(messages: List[PromptMessageExtended]) -> str:
     """
-    Convert PromptMessageExtended objects to a pure JSON string in GetPromptResult format.
+    Convert PromptMessageExtended objects directly to JSON, preserving all extended fields.
 
-    This approach preserves all data and structure exactly as is, compatible with
-    the MCP GetPromptResult type.
+    This preserves tool_calls, tool_results, channels, and stop_reason that would be lost
+    in the standard GetPromptResult conversion.
 
     Args:
         messages: List of PromptMessageExtended objects
 
     Returns:
-        JSON string representation with GetPromptResult container
+        JSON string representation preserving all PromptMessageExtended data
     """
-    # First convert to GetPromptResult
-    result = multipart_messages_to_get_prompt_result(messages)
+    # Convert each message to dict using standardized serialization
+    messages_dicts = [serialize_to_dict(msg) for msg in messages]
 
-    # Convert to dictionary using model_dump with proper JSON mode
-    result_dict = result.model_dump(by_alias=True, mode="json", exclude_none=True)
+    # Wrap in a container similar to GetPromptResult for consistency
+    result_dict = {"messages": messages_dicts}
 
     # Convert to JSON string
     return json.dumps(result_dict, indent=2)
 
 
-def json_to_extended_messages(json_str: str) -> List[PromptMessageExtended]:
+def from_json(json_str: str) -> List[PromptMessageExtended]:
     """
-    Parse a JSON string in GetPromptResult format into PromptMessageExtended objects.
+    Parse a JSON string into PromptMessageExtended objects.
+
+    Handles both:
+    - Enhanced format with full PromptMessageExtended data
+    - Legacy GetPromptResult format (missing extended fields default to None)
 
     Args:
-        json_str: JSON string representation of GetPromptResult
+        json_str: JSON string representation
 
     Returns:
         List of PromptMessageExtended objects
@@ -96,31 +138,66 @@ def json_to_extended_messages(json_str: str) -> List[PromptMessageExtended]:
     # Parse JSON to dictionary
     result_dict = json.loads(json_str)
 
-    # Parse as GetPromptResult
-    result = GetPromptResult.model_validate(result_dict)
+    # Extract messages array
+    messages_data = result_dict.get("messages", [])
+
+    extended_messages: List[PromptMessageExtended] = []
+    basic_buffer: List[PromptMessage] = []
+
+    def flush_basic_buffer() -> None:
+        nonlocal basic_buffer
+        if not basic_buffer:
+            return
+        extended_messages.extend(PromptMessageExtended.to_extended(basic_buffer))
+        basic_buffer = []
+
+    for msg_data in messages_data:
+        content = msg_data.get("content")
+        is_enhanced = isinstance(content, list)
+        if is_enhanced:
+            try:
+                msg = PromptMessageExtended.model_validate(msg_data)
+            except Exception:
+                is_enhanced = False
+            else:
+                flush_basic_buffer()
+                extended_messages.append(msg)
+                continue
+
+        try:
+            basic_msg = PromptMessage.model_validate(msg_data)
+        except Exception:
+            continue
+        basic_buffer.append(basic_msg)
 
-    # Convert to multipart messages
-    return PromptMessageExtended.to_extended(result.messages)
+    flush_basic_buffer()
 
+    return extended_messages
 
-def save_messages_to_json_file(messages: List[PromptMessageExtended], file_path: str) -> None:
+
+def save_json(messages: List[PromptMessageExtended], file_path: str) -> None:
     """
-    Save PromptMessageExtended objects to a JSON file.
+    Save PromptMessageExtended objects to a JSON file using enhanced format.
+
+    Uses the enhanced format that preserves tool_calls, tool_results, channels,
+    and stop_reason data.
 
     Args:
         messages: List of PromptMessageExtended objects
         file_path: Path to save the JSON file
     """
-    json_str = multipart_messages_to_json(messages)
+    json_str = to_json(messages)
 
     with open(file_path, "w", encoding="utf-8") as f:
         f.write(json_str)
 
 
-def load_messages_from_json_file(file_path: str) -> List[PromptMessageExtended]:
+def load_json(file_path: str) -> List[PromptMessageExtended]:
     """
     Load PromptMessageExtended objects from a JSON file.
 
+    Handles both enhanced format and legacy GetPromptResult format.
+
     Args:
         file_path: Path to the JSON file
 
@@ -130,14 +207,14 @@ def load_messages_from_json_file(file_path: str) -> List[PromptMessageExtended]:
     with open(file_path, "r", encoding="utf-8") as f:
         json_str = f.read()
 
-    return json_to_extended_messages(json_str)
+    return from_json(json_str)
 
 
-def save_messages_to_file(messages: List[PromptMessageExtended], file_path: str) -> None:
+def save_messages(messages: List[PromptMessageExtended], file_path: str) -> None:
     """
     Save PromptMessageExtended objects to a file, with format determined by file extension.
 
-    Uses GetPromptResult JSON format for .json files (fully MCP compatible) and
+    Uses enhanced JSON format for .json files (preserves all fields) and
     delimited text format for other extensions.
 
     Args:
@@ -147,19 +224,16 @@ def save_messages_to_file(messages: List[PromptMessageExtended], file_path: str)
     path_str = str(file_path).lower()
 
     if path_str.endswith(".json"):
-        # Use GetPromptResult JSON format for .json files (fully MCP compatible)
-        save_messages_to_json_file(messages, file_path)
+        save_json(messages, file_path)
     else:
-        # Use delimited text format for other extensions
-        save_messages_to_delimited_file(messages, file_path)
+        save_delimited(messages, file_path)
 
 
-def load_messages_from_file(file_path: str) -> List[PromptMessageExtended]:
+def load_messages(file_path: str) -> List[PromptMessageExtended]:
     """
     Load PromptMessageExtended objects from a file, with format determined by file extension.
 
-    Uses GetPromptResult JSON format for .json files (fully MCP compatible) and
-    delimited text format for other extensions.
+    Uses JSON format for .json files and delimited text format for other extensions.
 
     Args:
         file_path: Path to the file
@@ -170,11 +244,9 @@ def load_messages_from_file(file_path: str) -> List[PromptMessageExtended]:
     path_str = str(file_path).lower()
 
     if path_str.endswith(".json"):
-        # Use GetPromptResult JSON format for .json files (fully MCP compatible)
-        return load_messages_from_json_file(file_path)
+        return load_json(file_path)
     else:
-        # Use delimited text format for other extensions
-        return load_messages_from_delimited_file(file_path)
+        return load_delimited(file_path)
 
 
 # -------------------------------------------------------------------------
@@ -238,7 +310,7 @@ def multipart_messages_to_delimited_format(
                     delimited_content.append(resource_delimiter)
 
                     # Convert to dictionary using proper JSON mode
-                    content_dict = content.model_dump(by_alias=True, mode="json", exclude_none=True)
+                    content_dict = serialize_to_dict(content)
 
                     # Add to delimited content as JSON
                     delimited_content.append(json.dumps(content_dict, indent=2))
@@ -253,7 +325,7 @@ def multipart_messages_to_delimited_format(
                     delimited_content.append(resource_delimiter)
 
                     # Convert to dictionary using proper JSON mode
-                    content_dict = content.model_dump(by_alias=True, mode="json", exclude_none=True)
+                    content_dict = serialize_to_dict(content)
 
                     # Add to delimited content as JSON
                     delimited_content.append(json.dumps(content_dict, indent=2))
@@ -281,6 +353,17 @@ def delimited_format_to_extended_messages(
     Returns:
         List of PromptMessageExtended objects
     """
+    if user_delimiter not in content and assistant_delimiter not in content:
+        stripped = content.strip()
+        if not stripped:
+            return []
+        return [
+            PromptMessageExtended(
+                role="user",
+                content=[TextContent(type="text", text=stripped)],
+            )
+        ]
+
     lines = content.split("\n")
     messages = []
 
@@ -365,11 +448,13 @@ def delimited_format_to_extended_messages(
                         resource_uri = f"resource://fast-agent/{resource_uri}"
 
                     # Create a simple resource with just the URI
+                    # For legacy format, we don't have the actual content, just the reference
                     resource = EmbeddedResource(
                         type="resource",
                         resource=TextResourceContents(
                             uri=resource_uri,
                             mimeType="text/plain",
+                            text="",  # Legacy format doesn't include content
                         ),
                     )
                     resource_contents.append(resource)
@@ -436,7 +521,7 @@ def delimited_format_to_extended_messages(
     return messages
 
 
-def save_messages_to_delimited_file(
+def save_delimited(
     messages: List[PromptMessageExtended],
     file_path: str,
     user_delimiter: str = USER_DELIMITER,
@@ -467,7 +552,7 @@ def save_messages_to_delimited_file(
         f.write("\n".join(delimited_content))
 
 
-def load_messages_from_delimited_file(
+def load_delimited(
     file_path: str,
     user_delimiter: str = USER_DELIMITER,
     assistant_delimiter: str = ASSISTANT_DELIMITER,