Test 1

Prompt

Analyze, plan, reason, and execute a complex technical project: ## Objective Design and build a functional prototype of a **personal AI productivity system** that runs locally on a Linux computer. The system should be capable of: 1. Accepting natural-language commands from the user. 2. Breaking complex requests into smaller tasks. 3. Creating and executing a task plan. 4. Using local computer resources when necessary. 5. Searching the web when current information is required. 6. Reading and analyzing local files. 7. Running Python scripts and other safe commands. 8. Remembering relevant information during the project. 9. Recovering from errors instead of stopping immediately. 10. Clearly reporting what it is doing and why. ## Constraints * Assume the computer runs an Arch-based Linux distribution. * Assume 16 GB RAM and an AMD GPU. * Prefer open-source and free software. * Do not require paid APIs unless absolutely necessary. * The system should be designed to run locally whenever practical. * Do not blindly execute destructive commands. * Never claim that something works unless you have actually verified it. * If a dependency, API, library, or command may have changed, verify it using current documentation or reliable sources. * Keep the architecture reasonably lightweight. ## Your Task Do not immediately start writing random code. First: ### Phase 1 — Analysis Identify: * The requirements * The major technical challenges * Necessary components * Possible architecture * Security risks * Resource limitations * What should run locally vs. externally ### Phase 2 — Architecture Design the complete architecture. Explain: * Components * Data flow * APIs * Storage * Agent/task system * Tool execution system * Error handling * Security model Provide a directory structure for the project. ### Phase 3 — Implementation Implement the prototype. Provide all necessary code and configuration files. The prototype must contain at minimum: * A command-line interface * An AI agent loop * Tool/function calling * Task planning * Persistent task state * File-reading capability * Python execution capability * Web-search capability or a clearly defined interface for it * Error recovery * Logging ### Phase 4 — Testing Create tests that demonstrate: 1. A simple request. 2. A multi-step request. 3. A request requiring file analysis. 4. A request requiring code execution. 5. A request where a tool fails. 6. A request where the AI must recover from an error. 7. A request where the AI must refuse an unsafe operation. Do not merely describe the tests. Actually reason through the expected execution and identify likely failure points. ### Phase 5 — Optimization After the prototype is complete, critically evaluate your own implementation. Identify: * Bugs * Bottlenecks * Security vulnerabilities * Poor architectural decisions * Unnecessary complexity * Possible improvements Then provide a second, improved version of the most important components.