ibfs: # Initial strategy generation prompt initial_strategies: system: | You are an assistant that helps users explore different ways to answer their questions. Generate exactly the requested number of distinct strategies to approach their query. Number each strategy clearly (1, 2, 3, etc.) and ensure each one starts with "I can answer by". Make each strategy different from the others - be creative and consider various angles. Each strategy should be substantively different in approach, not just variations on the same idea. Provide enough detail for each strategy that the user can understand the approach. user: | For the following query: {query} Please generate exactly {k} different numbered strategies for how you might approach answering this question. Format each strategy like this: 1. I can answer by [approach description] 2. I can answer by [approach description] ... and so on. Each strategy should represent a substantially different approach to answering the question. # Continuation strategy generation prompt continuation_strategies: system: | You are an assistant that helps users explore different ways to answer their questions. The user previously said they liked a specific strategy, and now your job is to generate sub-strategies. Generate exactly the requested number of distinct sub-strategies to approach their query. Number each strategy clearly (1, 2, 3, etc.) and ensure each one starts with "I can answer by". Make each strategy different from the others - be creative and consider various angles. Each strategy should be substantively different in approach, not just variations on the same idea. Provide enough detail for each strategy that the user can understand the approach. But make sure that each sub-strategy is a distinct variation of the original strategy. Again, make sure that each sub-strategy is a distinct variation of the original strategy. user: | Original query: {query} The user previously said they liked the strategy: {selected_strategy} Now, let's be more specific. Generate exactly {k} different numbered sub-strategies of this strategy. Each sub-strategy should be a distinct variation of the original strategy. Format each strategy like this: 1. I can answer by [approach description] 2. I can answer by [approach description] ... and so on. Each strategy should represent a different way to implement or approach this strategy. # Final answer generation prompt final_answer: system: | You are a helpful assistant. A user has asked a question and selected a specific approach to answer it. Please provide a thorough and detailed answer using exactly the approach specified. Your response should directly implement the strategy without restating it or introducing new approaches. user: | Query: {query} Selected strategy: {final_strategy} Please answer the query using precisely this approach. # UI messages ui_messages: welcome: | I'll help you explore different ways to answer this question using Interactive Best-First Search (IBFS). We'll go through {m} iterations, with {k} strategy options at each step. generating: | Generating initial strategy options... select_strategy: | Step {current_step}/{max_steps}: Please select a strategy by entering a number ({valid_options}): final_processing: | Generating final answer based on your selected strategies... final_result: | Based on your selected strategy: '{final_strategy}', here's my answer: {final_answer} saved_result: | Results saved to {save_path} User ID: {user_id} error_missing_choice: | Please select an option by entering a number. error_invalid_choice: | Invalid choice. Please select a number between 1 and {max_option}. error_invalid_number: | '{choice}' is not a valid number. Please enter a number like 1, 2, 3, etc. error_no_strategies: | No strategy options available. Please start a new query. error_general: | Error: {error_message} # Zero-shot prompts zero_shot: # Direct answer without interactive search direct_answer: system: | user: | {query} # UI messages for zero-shot ui_messages: generating: | Generating direct answer to your query... result: | Here's my answer to your question: {answer} saved_result: | Results saved to {save_path} User ID: {user_id} error_general: | Error: {error_message}