class openssa.deprecated.ooda_rag.prompts.OODAPrompts

Bases: object

PROVIDE_TOOLS = 'You have the following functions available to call, provided here in a dict of function names and descriptions. {tool_descriptions}'

FORMULATE_TASK = 'Reformulate the user\'s input as a task (a sentence, not a function call) that could be theoretically completed to satisfy the user. Return a JSON dictionary {"content": "your thinking here", "task": "your task here"}'

DECOMPOSE_INTO_SUBTASKS = 'Given the tools available, if the task cannot be completed directly with the current tools and resources, break it down into maximum 2 smaller subtasks that can be directly addressed in order. If it does not need to be broken down, return an empty list of subtasks. Return a JSON dictionary {"subtasks": ["subtask 1", "subtask 2", ...]} each subtask should be a sentence or command or question not a function call. Return json only, nothing else. Think step by step.'

DECOMPOSE_INTO_OODA = 'Given the task at hand and the tools available, decompose the task into an OODA procedure. Return a JSON dictionary {"observe": "instructions for observe step", "orient": "instructions for orient step", "decide": "instructions for decide step", "act": "instructions for act step"}'

OBSERVE = 'We are in the OBSERVE step. What progress has been made so far? What information do you have and what can you obtain via the function tools available? Return a JSON dictionary including the function calls you want to make in the form of function: params. {"content": "your thinking here", "calls": [{"function1": "params1"}, {"function2": "params2"}]}'

ORIENT = 'We are in the ORIENT step. Analyse and reflect on the information obtained. What are the possible courses of action you can take to complete the task? Return a JSON dictionary {"content": "your thinking here"}'

DECIDE = 'We are in the DECIDE step. Choose a course of action to take based on your analysis. What functions do we need to execute? Return a JSON dictionary including the function calls you want to make in the form of function: params. {"content": "your thinking here", "calls": [{"function1": "params1"}, {"function2": "params2"}]}'

ACT = 'We are in the ACT step. Given everything so far, finalise your written response to complete the task at hand. Return a JSON dictionary {"content": "your response here"}'

SYNTHESIZE_RESULT = "As an expert in reasoning, you are examining a dialogue involving a user, an assistant, and a system. Your task is to synthesize the final answer to the user's initial question based on this conversation. This is the concluding instruction and must be followed with precision. You will derive the final response by critically analyzing all the messages in the conversation and performing any necessary calculations. Be aware that some contributions from the assistant may not be relevant or could be misleading due to being based on incomplete information. {heuristic} If the conversation does not provide sufficient information to synthesize the answer then admit you cannot produce accurate answer. Do not use any information outside of the conversation context. Exercise discernment in selecting the appropriate messages to construct a logical and step-by-step reasoning process."

class openssa.deprecated.ooda_rag.prompts.BuiltInAgentPrompt

Bases: object

COMMUNICATION = "You are an expert in communication. Your will help to format following message with this instruction:\n###{instruction}###\n\nHere is the message:\n###{message}###\n\nOutput the response in JSON format with the keyword 'message'."

PROBLEM_STATEMENT = 'Formulate a concise problem statement from the most recent question asked by a user in a conversation with an AI expert. Focus on the question that remains unanswered (Question B), ignoring any questions that have already been addressed (Question A). The problem statement should be direct and to the point, avoiding any additional explanatory text or context not explicitly mentioned by the user. Ensure to correct any spelling or grammatical errors, and replace pronouns or ambiguous terms with precise names or technical terms. The outcome should be a succinct problem statement, accurately reflecting the user\'s query for use in a retrieval system. Format the response in JSON, clearly labeled as \'problem statement\'.\n\nExample:\nUser: What is the flow rate for Argon in machine XYZ? (Question A)\nAssistant: The flow rate for Argon is 500 cm^3 per minute.\nUser: Can you also give me its velocity? (Question B)\n\nResponse:\n{\n    "problem statement": "What is the velocity of Argon in machine XYZ."\n}\n\n'

ASK_USER = 'Your task is to generate a question for the user to gather necessary information for troubleshooting or solving their problem, based on the ongoing conversation. Follow the heuristic guidelines closely to determine the relevance and necessity of asking a question. If the discussion does not align with these guidelines or the required information has already been addressed and the user has indicated ignorance on the matter, you should refrain from asking a question. In such cases, produce an output with an empty \'question\' field in JSON format. Ask questions judiciously, only when the conditions explicitly meet the heuristic criteria.\nGuidelines for Question Generation:\n1. Relevance: Ensure the question is directly related to the user\'s stated issue.\n2. Condition Clarity: Only ask if the scenario clearly matches the heuristics.\n3. Non-redundancy: Avoid repeating questions already answered or where the user expressed ignorance.\n4. Output Format: Provide the question in a JSON object with the keyword \'question\'. If no question is warranted, return an empty \'question\' value.\n\nExample 1:\n- User says: "My printer is not working, can you help me fix it?"\n- Heuristic guideline: Ask for the printer model if the user seeks help with a printer issue but does not specify the model.\n- Output:\n{{ "question": "Could you provide the printer model?" }}\n\nExample 2:\n- User says: "I couldn\'t open the cabin of my car, can you help?"\n- Heuristic guideline: Ask about boiler\'s status and external conditions if the user is seeking help with mechanical troubleshooting.\n- Output:\n{{ "question": "" }}\n\nBelow is the heuristic guideline:\n\n###{heuristic}###\n\n'

ASK_USER_OODA = 'Your task is to assist an AI assistant in formulating a question for the user. This is done through using OODA reasoning. This should be based on the ongoing conversation, the presented problem statement, and a specific heuristic guideline. The assistant should formulate the question strictly based on the heuristic. If the heuristic does not apply or is irrelevant to the problem statement, return empty string for the question. Output the response of ooda reasoning in JSON format with the keyword "observe", "orient", "decide", "act". Example output key value:\n\n    "observe": "Here, articulate your initial assessment of the task, capturing essential details and contextual elements.",\n    "orient": "In this phase, analyze and synthesize the gathered information, considering different angles and strategies.",\n    "decide": "Now, determine the most suitable action based on your observations and analysis.",\n    "act": "The question to ask the user is here."\n \n\nBelow is the heuristic guideline:\n###{heuristic}###\n\nHere is the problem statement or the user\'s current question:\n###{problem_statement}###\n\nOutput the JSON only. Think step by step.'

CONTENT_VALIDATION = 'You are tasked as an expert in reasoning and contextual analysis. Your role is to evaluate whether the provided context and past conversation contain enough information to accurately respond to a given query.\n\nPlease analyze the past conversation and the following context. Then, determine if the information is sufficient to form an accurate answer. Respond only in JSON format with the keyword \'is_sufficient\'. This should be a boolean value: True if the information is adequate, and False if it is not.\n\nYour response should be in the following format:\n{{\n    "is_sufficient": [True/False]\n}}\n\nDo not include any additional commentary. Focus solely on evaluating the sufficiency of the provided context and conversation.\n\nContext:\n========\n{context}\n========\n\nQuery:\n{query}\n'

ANSWER_VALIDATION = "Your role is to act as an expert in reasoning and contextual analysis. You need to evaluate if the provided answer effectively and clearly addresses the query. Respond with 'yes' if the answer is clear and confident, and 'no' if it is not. Here are some examples to guide you: \n\nExample 1:\nQuery: Can I print a part 50 cm long with this machine?\nAnswer: Given the information and the lack of detailed specifications, it is not possible to determine if the machine can print a part 50 cm long.\nEvaluation: no\n\nExample 2:\nQuery: Can I print a part 50 cm long with this machine?\nAnswer: No, it is not possible to print a part 50 cm long with this machine.\nEvaluation: yes\n\nExample 3:\nQuery: How to go to the moon?\nAnswer: I'm sorry, but based on the given context information, there is no information provided on how to go to the moon.\nEvaluation: no\n\n"

SYNTHESIZE_RESULT = 'As an expert in problem-solving and contextual analysis, you are to synthesize an answer for a given query. This task requires you to use only the information provided in the previous conversation and the context given below. Your answer should exclusively rely on this information as the base knowledge.\n\nYour response must be in JSON format, using the keyword \'answer\'. The format should strictly adhere to the following structure:\n{{\n    "answer": "Your synthesized answer here"\n}}\n\nPlease refrain from including any additional commentary or information outside of the specified context and past conversation.\n\nContext:\n========\n{context}\n========\n\nQuery:\n{query}\n'

GENERATE_OODA_PLAN = "As a specialist in problem-solving, your task is to utilize the OODA loop as a cognitive framework for addressing various tasks, which could include questions, commands, or messages. You have at your disposal a range of tools to aid in resolving these issues. Your responses should be methodically structured according to the OODA loop, formatted as a JSON dictionary. Each dictionary key represents one of the OODA loop's four stages: Observe, Orient, Decide, and Act. Within each stage, detail your analytical process and, when relevant, specify the execution of tools, including their names and parameters. Only output the JSON and nothing else. The proposed output format is as follows: \n{\n    'observe': {\n        'thought': 'Here, articulate your initial assessment of the task, capturing essential details and contextual elements.',\n        'calls': [{'tool_name': '', 'parameters': ''}, ...]  // List tools and their parameters, if any are used in this stage.\n    },\n    'orient': {\n        'thought': 'In this phase, analyze and synthesize the gathered information, considering different angles and strategies.',\n        'calls': [{'tool_name': '', 'parameters': ''}, ...]   // Include any tools that aid in this analytical phase.\n    },\n    'decide': {\n        'thought': 'Now, determine the most suitable action based on your observations and analysis.',\n        'calls': [{'tool_name': '', 'parameters': ''}, ...]   // Specify tools that assist in making this decision, if applicable.\n    },\n    'act': {\n        'thought': 'Finally, outline the implementation steps based on your decision, including any practical actions or responses.',\n        'calls': [{'tool_name': '', 'parameters': ''}, ...]   // List any tools used in the implementation of the decision.\n    }\n}"