fastworkflow 2.15.3__tar.gz → 2.15.4__tar.gz

{fastworkflow-2.15.3 → fastworkflow-2.15.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: fastworkflow
-Version: 2.15.3
+Version: 2.15.4
 Summary: A framework for rapidly building large-scale, deterministic, interactive workflows with a fault-tolerant, conversational UX
 License: Apache-2.0
 Keywords: fastworkflow,ai,workflow,llm,openai

{fastworkflow-2.15.3 → fastworkflow-2.15.4}/fastworkflow/_workflows/command_metadata_extraction/_commands/wildcard.py

@@ -602,6 +602,7 @@ class ResponseGenerator:
         if cnp_output.error_msg:
             workflow_context = workflow.context
             workflow_context["NLU_Pipeline_Stage"] = NLUPipelineStage.INTENT_AMBIGUITY_CLARIFICATION
+            workflow_context["command"] = command
             workflow.context = workflow_context
             return CommandOutput(
                 command_responses=[

{fastworkflow-2.15.3 → fastworkflow-2.15.4}/fastworkflow/chat_session.py

@@ -424,14 +424,25 @@ class ChatSession:
         if os.path.exists("action.json"):
             os.remove("action.json")
 
-        refined_message = f'messsage\n{self._think_and_plan(message)}'
+        refined_message = f'messsage\n{self._think_and_plan(message, self.conversation_history)}'
 
         lm = dspy_utils.get_lm("LLM_AGENT", "LITELLM_API_KEY_AGENT")
-        with dspy.context(lm=lm):
-            agent_result = self._workflow_tool_agent(
-                user_query=refined_message,
-                conversation_history=self.conversation_history
-            )
+        from dspy.utils.exceptions import AdapterParseError
+        # Retry logic for AdapterParseError
+        max_retries = 2
+        for attempt in range(max_retries):
+            try:
+                with dspy.context(lm=lm, adapter=dspy.ChatAdapter()):
+                    agent_result = self._workflow_tool_agent(
+                        user_query=refined_message,
+                        conversation_history=self.conversation_history
+                    )
+                break  # Success, exit retry loop
+            except AdapterParseError as _:
+                if attempt == max_retries - 1:  # Last attempt
+                    raise  # Re-raise the exception if all retries failed
+                # Continue to next attempt
+
             # dspy.inspect_history(n=1)
 
         # Extract the final result from the agent
@@ -441,18 +452,21 @@ class ChatSession:
             else str(agent_result)
         )
 
-        self.conversation_history.messages.append(
-            {"user_query": message, 
-             "agent_response": result_text}
-        )
-
         # Create CommandOutput with the agent's response
         command_response = fastworkflow.CommandResponse(response=result_text)
+
+        user_instructions_summary = message
         # Attach actions captured during agent execution as artifacts if available
         if os.path.exists("action.json"):
             with open("action.json", "r", encoding="utf-8") as f:
                 actions = [json.loads(line) for line in f if line.strip()]
-            command_response.artifacts["extracted_actions"] = actions
+            user_instructions_summary = self._extract_user_instructions(message, actions)
+            command_response.artifacts["user_instructions_summary"] = user_instructions_summary
+
+        self.conversation_history.messages.append(
+            {"user_instructions": user_instructions_summary, 
+             "agent_response": result_text}
+        )
 
         command_output = fastworkflow.CommandOutput(
             command_responses=[command_response]
@@ -588,16 +602,17 @@ class ChatSession:
 
         return command_output
 
-    def _think_and_plan(self, user_query: str) -> str:
+    def _think_and_plan(self, user_query: str, conversation_history: dspy.History) -> str:
         """
-        Returns a refined plan by breaking down a user_query into simpler tasks based only on available commands and returns a todo list.
+        Returns a refined plan by breaking down a user_query into simpler tasks.
         """
         class TaskPlannerSignature(dspy.Signature):
             """
-            Break down a user_query into simpler tasks based only on available commands and return a todo list.
+            Break down a user_query into simpler tasks based only on available commands and conversation_history.
             If user_query is simple, return a single todo that is the user_query as-is
             """
             user_query: str = dspy.InputField()
+            conversation_history: dspy.History = dspy.InputField()
             available_commands: list[str] = dspy.InputField()
             todo_list: list[str] = dspy.OutputField(desc="task descriptions as short sentences")
 
@@ -611,10 +626,36 @@ class ChatSession:
         planner_lm = dspy_utils.get_lm("LLM_PLANNER", "LITELLM_API_KEY_PLANNER")
         with dspy.context(lm=planner_lm):
             task_planner_func = dspy.ChainOfThought(TaskPlannerSignature)
-            prediction = task_planner_func(user_query=user_query, available_commands=available_commands)
+            prediction = task_planner_func(
+                user_query=user_query,
+                conversation_history=conversation_history, 
+                available_commands=available_commands)
 
             if not prediction.todo_list or (len(prediction.todo_list) == 1 and prediction.todo_list[0] == user_query):
                 return user_query
 
             steps_list = '\n'.join([f'{i + 1}. {task}' for i, task in enumerate(prediction.todo_list)])
             return f"{user_query}\nNext steps:\n{steps_list}"
+
+
+    def _extract_user_instructions(self, 
+        user_query: str, workflow_actions: list[dict[str, str]]) -> str:
+        """
+        Summarizes user instructions based on original user query and subsequent user feedback in workflow actions.
+        """
+        class UserInstructionCompilerSignature(dspy.Signature):
+            """
+            Concise summary of user instructions based on their commands to the workflow. 
+            Include parameter values passed in commands in the summary.
+            """
+            commands_list: list[str] = dspy.InputField()
+            user_instructions_summary: str = dspy.OutputField(desc="A single paragraph summary")
+
+        commands_list: list[str] = [user_query]
+        commands_list.extend([wf_action['command'] for wf_action in workflow_actions if 'command' in wf_action])
+
+        planner_lm = dspy_utils.get_lm("LLM_PLANNER", "LITELLM_API_KEY_PLANNER")
+        with dspy.context(lm=planner_lm):
+            uic_func = dspy.ChainOfThought(UserInstructionCompilerSignature)
+            prediction = uic_func(commands_list=commands_list)
+            return prediction.user_instructions_summary

{fastworkflow-2.15.3 → fastworkflow-2.15.4}/fastworkflow/examples/retail_workflow/_commands/get_user_details.py

@@ -50,6 +50,7 @@ class Signature:
         "What details do you have on my user account?",
         "Can you show me everything tied to my profile?",
         "I'd like to review my account and recent activity.",
+        "retrieve user details"
     ]
     template_utterances: List[str] = []
 

{fastworkflow-2.15.3 → fastworkflow-2.15.4}/fastworkflow/workflow_agent.py

@@ -14,7 +14,7 @@ from fastworkflow.mcp_server import FastWorkflowMCPServer
 
 class WorkflowAgentSignature(dspy.Signature):
     """
-    Carefully review the user request and execute the todo list using available tools for building the final answer.
+    Carefully review the user request and conversation_history, then execute the todo list using available tools for building the final answer.
     All the tasks in the todo list must be completed before returning the final answer.
     """
     user_query = dspy.InputField(desc="The natural language user query.")
@@ -185,12 +185,18 @@ def initialize_workflow_tool_agent(mcp_server: FastWorkflowMCPServer, max_iters:
         Commands must be formatted using plain text for command name followed by XML tags enclosing parameter values (if any) as follows: command_name <param1_name>param1_value</param1_name> <param2_name>param2_value</param2_name> ...
         Don't use this tool to respond to a clarification requests in PARAMETER EXTRACTION ERROR state
         """
-        try:
-            return _execute_workflow_query(command, chat_session_obj=chat_session_obj)
-        except Exception as e:
-            message = f"Terminate immediately! Exception processing {command}: {str(e)}"
-            logger.critical(message)                    
-            return message
+        # Retry logic for workflow execution
+        max_retries = 2
+        for attempt in range(max_retries):
+            try:
+                return _execute_workflow_query(command, chat_session_obj=chat_session_obj)
+            except Exception as e:
+                if attempt == max_retries - 1:  # Last attempt
+                    message = f"Terminate immediately! Exception processing {command}: {str(e)}"
+                    logger.critical(message)                    
+                    return message
+                # Continue to next attempt
+                logger.warning(f"Attempt {attempt + 1} failed for command '{command}': {str(e)}")
 
     def missing_information_guidance(how_to_find_request: str) -> str:
         """

{fastworkflow-2.15.3 → fastworkflow-2.15.4}/pyproject.toml

@@ -9,7 +9,7 @@ repository = "https://github.com/radiantlogicinc/fastworkflow"
 
 [tool.poetry]
 name = "fastworkflow"
-version = "2.15.3"
+version = "2.15.4"
 description = "A framework for rapidly building large-scale, deterministic, interactive workflows with a fault-tolerant, conversational UX"
 authors = ["Dhar Rawal <drawal@radiantlogic.com>"]
 license = "Apache-2.0"