# # This workflow coordinates multiple agents in order to address complex user requests # according to the "Magentic" orchestration pattern introduced by AutoGen. # # For this workflow, several agents used, each with a prompt specific to their role: # # 1. Analyst Agent: Able to analyze the current task. # Enable "Bing Grounding Tool" in the agent settings. # See: ./setup/AnalystAgent.yaml # # 2. Manager Agent: Able to create plans and delegate tasks to other agents. # See: ./setup/ManagerAgent.yaml # # 3. Research Agent: # Enable "Bing Grounding" in the agent settings. # See: ./setup/WebAgent.yaml # # With instructions: # # Only provide requested information in a way that is throughfully organized and formatted. # Never include any analysis or code. # Never generate a file. # Avoid repeating yourself. # # 4. Coder Agent: # Enable "Code Interpreter" in the agent settings. # See: ./setup/CoderAgent.yaml # # 5. Weather Agent: Able to retrieve factual information from the web. # Enable "Open API" in the agent settings using the wttr.json schema. # See: ./setup/WeatherAgent.yaml # kind: Workflow trigger: kind: OnConversationStart id: workflow_demo actions: - kind: SetVariable id: setVariable_aASlmF displayName: List all available agents for this orchestrator variable: Local.AvailableAgents value: |- =[ { name: "WeatherAgent", description: "Able to retrieve weather information", agentid: Env.FOUNDRY_AGENT_RESEARCHWEATHER }, { name: "CoderAgent", description: "Able to write and execute Python code", agentid: Env.FOUNDRY_AGENT_RESEARCHCODER }, { name: "WebAgent", description: "Able to perform generic websearches", agentid: Env.FOUNDRY_AGENT_RESEARCHWEB } ] - kind: SetVariable id: setVariable_V6yEbo displayName: Get a summary of all the agents for use in prompts variable: Local.TeamDescription value: "=Concat(ForAll(Local.AvailableAgents, $\"- \" & name & $\": \" & description), Value, \"\n\")" - kind: SetVariable id: setVariable_NZ2u0l displayName: Set Task variable: Local.InputTask value: =System.LastMessage.Text - kind: SetVariable id: setVariable_10u2ZN displayName: Set Task variable: Local.SeedTask value: =UserMessage(Local.InputTask) - kind: SendActivity id: sendActivity_yFsbRy activity: Analyzing facts... - kind: CreateConversation id: conversation_1a2b3c conversationId: Local.InternalConversationId - kind: InvokeAzureAgent id: question_UDoMUw displayName: Get Facts conversationId: =Local.InternalConversationId agent: name: =Env.FOUNDRY_AGENT_RESEARCHANALYST output: autoSend: false messages: Local.TaskFacts input: messages: =UserMessage(Local.InputTask) additionalInstructions: |- In order to help begin addressing the user request, please answer the following pre-survey to the best of your ability. Keep in mind that you are Ken Jennings-level with trivia, and Mensa-level with puzzles, so there should be a deep well to draw from. Here is the pre-survey: 1. Please list any specific facts or figures that are GIVEN in the request itself. It is possible that there are none. 2. Please list any facts that may need to be looked up, and WHERE SPECIFICALLY they might be found. In some cases, authoritative sources are mentioned in the request itself. 3. Please list any facts that may need to be derived (e.g., via logical deduction, simulation, or computation) 4. Please list any facts that are recalled from memory, hunches, well-reasoned guesses, etc. When answering this survey, keep in mind that 'facts' will typically be specific names, dates, statistics, etc. Your answer must only use the headings: 1. GIVEN OR VERIFIED FACTS 2. FACTS TO LOOK UP 3. FACTS TO DERIVE 4. EDUCATED GUESSES DO NOT include any other headings or sections in your response. DO NOT list next steps or plans until asked to do so. - kind: SendActivity id: sendActivity_yFsbRz activity: Creating a plan... - kind: InvokeAzureAgent id: question_DsBaJU displayName: Create a Plan conversationId: =Local.InternalConversationId agent: name: =Env.FOUNDRY_AGENT_RESEARCHMANAGER output: autoSend: false messages: Local.Plan input: messages: =UserMessage(Local.InputTask) additionalInstructions: |- Your only job is to devise an efficient plan that identifies (by name) how a team member may contribute to addressing the user request. Only select the following team which is listed as "- [Name]: [Description]" {Local.TeamDescription} The plan must be a bullet point list must be in the form "- [AgentName]: [Specific action or task for that agent to perform]" Remember, there is no requirement to involve the entire team -- only select team member's whose particular expertise is required for this task. - kind: SetVariable id: setVariable_Kk2LDL displayName: Define instructions variable: Local.TaskInstructions value: |- ="# TASK Address the following user request: " & Local.InputTask & " # TEAM Use the following team to answer this request: " & Local.TeamDescription & " # FACTS Consider this initial fact sheet: " & Trim(Last(Local.TaskFacts).Text) & " # PLAN Here is the plan to follow as best as possible: " & Last(Local.Plan).Text - kind: SendActivity id: sendActivity_bwNZiM activity: {Local.TaskInstructions} - kind: InvokeAzureAgent id: question_o3BQkf displayName: Progress Ledger Prompt conversationId: =Local.InternalConversationId agent: name: =Env.FOUNDRY_AGENT_RESEARCHMANAGER output: autoSend: false messages: Local.ProgressLedgerUpdate input: messages: =UserMessage(Local.AgentResponseText) additionalInstructions: |- Recall we are working on the following request: {Local.InputTask} And we have assembled the following team: {Local.TeamDescription} To make progress on the request, please answer the following questions, including necessary reasoning: - Is the request fully satisfied? (True if complete, or False if the original request has yet to be SUCCESSFULLY and FULLY addressed) - Are we in a loop where we are repeating the same requests and / or getting the same responses from an agent multiple times? Loops can span multiple turns, and can include repeated actions like scrolling up or down more than a handful of times. - Are we making forward progress? (True if just starting, or recent messages are adding value. False if recent messages show evidence of being stuck in a loop or if there is evidence of significant barriers to success such as the inability to read from a required file) - Who should speak next? (select from: {Concat(Local.AvailableAgents, name, ",")}) - What instruction or question would you give this team member? (Phrase as if speaking directly to them, and include any specific information they may need) Please output an answer in pure JSON format according to the following schema. The JSON object must be parsable as-is. DO NOT OUTPUT ANYTHING OTHER THAN JSON, AND DO NOT DEVIATE FROM THIS SCHEMA: {{ "is_request_satisfied": {{ "reason": string, "answer": boolean }}, "is_in_loop": {{ "reason": string, "answer": boolean }}, "is_progress_being_made": {{ "reason": string, "answer": boolean }}, "next_speaker": {{ "reason": string, "answer": string (select from: {Concat(Local.AvailableAgents, name, ",")}) }}, "instruction_or_question": {{ "reason": string, "answer": string }} }} - kind: ParseValue id: parse_rNZtlV displayName: Parse ledger response variable: Local.TypedProgressLedger value: =Last(Local.ProgressLedgerUpdate).Text valueType: kind: Record properties: instruction_or_question: type: kind: Record properties: answer: String reason: String is_in_loop: type: kind: Record properties: answer: Boolean reason: String is_progress_being_made: type: kind: Record properties: answer: Boolean reason: String is_request_satisfied: type: kind: Record properties: answer: Boolean reason: String next_speaker: type: kind: Record properties: answer: String reason: String - kind: ConditionGroup id: conditionGroup_mVIecC conditions: - id: conditionItem_fj432c condition: =Local.TypedProgressLedger.is_request_satisfied.answer displayName: If Done actions: - kind: SendActivity id: sendActivity_kdl3mC activity: Completed! {Local.TypedProgressLedger.is_request_satisfied.reason} - kind: InvokeAzureAgent id: question_Ke3l1d displayName: Generate Response conversationId: =System.ConversationId agent: name: =Env.FOUNDRY_AGENT_RESEARCHMANAGER output: messages: Local.FinalResponse input: messages: =Local.SeedTask additionalInstructions: |- We have completed the task. Based only on the conversation and without adding any new information, synthesize the result of the conversation as a complete response to the user task. The user will only every see this last response and not the entire conversation, so please ensure it is complete and self-contained. - kind: EndConversation id: end_SVoNSV - id: conditionItem_yiqund condition: =Local.TypedProgressLedger.is_in_loop.answer || Not(Local.TypedProgressLedger.is_progress_being_made.answer) displayName: If Stalling actions: - kind: SetVariable id: setVariable_H5lXdD displayName: Increase stall count variable: Local.StallCount value: =Local.StallCount + 1 - kind: ConditionGroup id: conditionGroup_vBTQd3 conditions: - id: conditionItem_fpaNL9 condition: =Local.TypedProgressLedger.is_in_loop.answer displayName: Is Loop actions: - kind: SendActivity id: sendActivity_fpaNL9 activity: {Local.TypedProgressLedger.is_in_loop.reason} - id: conditionItem_NnqvXh condition: =Not(Local.TypedProgressLedger.is_progress_being_made.answer) displayName: Is No Progress actions: - kind: SendActivity id: sendActivity_NnqvXh activity: {Local.TypedProgressLedger.is_progress_being_made.reason} - kind: ConditionGroup id: conditionGroup_xzNrdM conditions: - id: conditionItem_NlQTBv condition: =Local.StallCount > 2 displayName: Stall Count Exceeded actions: - kind: SendActivity id: sendActivity_H5lXdD activity: Unable to make sufficient progress... - kind: ConditionGroup id: conditionGroup_4s1Z27 conditions: - id: conditionItem_EXAlhZ condition: =Local.RestartCount > 2 actions: - kind: SendActivity id: sendActivity_xKxFUU activity: Stopping after attempting {Local.RestartCount} restarts... - kind: EndConversation id: end_GHVrFh - kind: SendActivity id: sendActivity_cwNZiM activity: Re-analyzing facts... - kind: InvokeAzureAgent id: question_wFJ123 displayName: Get New Facts Prompt conversationId: =Local.InternalConversationId agent: name: =Env.FOUNDRY_AGENT_RESEARCHANALYST output: autoSend: false messages: Local.TaskFacts input: messages: |- =UserMessage( "As a reminder, we are working to solve the following task: " & Local.InputTask) additionalInstructions: |- It's clear we aren't making as much progress as we would like, but we may have learned something new. Please rewrite the following fact sheet, updating it to include anything new we have learned that may be helpful. Example edits can include (but are not limited to) adding new guesses, moving educated guesses to verified facts if appropriate, etc. Updates may be made to any section of the fact sheet, and more than one section of the fact sheet can be edited. This is an especially good time to update educated guesses, so please at least add or update one educated guess or hunch, and explain your reasoning. Here is the old fact sheet: {Local.TaskFacts} - kind: SendActivity id: sendActivity_dsBaJU activity: Re-analyzing plan... - kind: InvokeAzureAgent id: question_uEJ456 displayName: Create new Plan Prompt conversationId: =Local.InternalConversationId agent: name: =Env.FOUNDRY_AGENT_RESEARCHMANAGER output: autoSend: false messages: Local.Plan input: additionalInstructions: |- Please briefly explain what went wrong on this last run (the root cause of the failure), and then come up with a new plan that takes steps and/or includes hints to overcome prior challenges and especially avoids repeating the same mistakes. As before, the new plan should be concise, be expressed in bullet-point form, and consider the following team composition (do not involve any other outside people since we cannot contact anyone else): {Local.TeamDescription} - kind: SetVariable id: setVariable_jW7tmM displayName: Set Plan as Context variable: Local.TaskInstructions value: |- ="# TASK Address the following user request: " & Local.InputTask & " # TEAM Use the following team to answer this request: " & Local.TeamDescription & " # FACTS Consider this initial fact sheet: " & Local.TaskFacts.Text & " # PLAN Here is the plan to follow as best as possible: " & Local.Plan.Text - kind: SetVariable id: setVariable_6J2snP displayName: Reset Stall count variable: Local.StallCount value: 0 - kind: SetVariable id: setVariable_S6HCgh displayName: Increase Restart count variable: Local.RestartCount value: =Local.RestartCount + 1 - kind: GotoAction id: goto_LzfJ8u actionId: question_o3BQkf elseActions: - kind: SendActivity id: sendActivity_L7ooQO activity: |- ({Local.TypedProgressLedger.next_speaker.reason}) {Local.TypedProgressLedger.next_speaker.answer} - {Local.TypedProgressLedger.instruction_or_question.answer} - kind: SetVariable id: setVariable_L7ooQO variable: Local.StallCount value: 0 - kind: SetVariable id: setVariable_nxN1mE variable: Local.NextSpeaker value: =Search(Local.AvailableAgents, Local.TypedProgressLedger.next_speaker.answer, name) - kind: ConditionGroup id: conditionGroup_QFPiF5 conditions: - id: conditionItem_GmigcU condition: =CountRows(Local.NextSpeaker) = 1 displayName: If next Agent tool Exists actions: - kind: InvokeAzureAgent id: question_orsBf06 displayName: Progress Ledger Prompt conversationId: =System.ConversationId agent: name: =First(Local.NextSpeaker).agentid output: messages: Local.AgentResponse input: messages: =Local.SeedTask additionalInstructions: |- {Local.TypedProgressLedger.instruction_or_question.answer} - kind: SetVariable id: setVariable_XzNrdM variable: Local.AgentResponseText value: =Last(Local.AgentResponse).Text - kind: ResetVariable id: setVariable_8eIx2A displayName: Clear seed task variable: Local.SeedTask elseActions: - kind: SendActivity id: sendActivity_BhcsI7 activity: Unable to choose next agent... - kind: SetVariable id: setVariable_BhcsI7 displayName: Increase stall count variable: Local.StallCount value: =Local.StallCount + 1 - kind: GotoAction id: goto_76Hne8 actionId: question_o3BQkf