the idea is to have a text-generative GPT AI be employed as a whole software company.
For this, several roles and specific actions need to be defined for each role which can be acted on.
The text-generative ai can then manage it's own tasks and work on them iteratively.
- available tools need to be text based. The AI needs an interface to read / edit text files and interact with the console.
- The amount of data possible per action may be limited. workarounds need to be found here
- knowledgeebase: (this) wiki.js
- flowcharts: mermaid (wiki.js default)
- research: c# browser integration
- devops: build tools
- system interaction: cmd, folder browsing, file interactions
responsible for managing tasks, prioritizing workflows, and allocating resources.
- Create new task lists
- Assign tasks to self or other roles
- Update task status (e.g., in progress, completed)
- Generate reports on task progress
responsible for gathering information, conducting research, and providing insights.
- Search online databases and knowledge bases
- Extract relevant information from sources
- Summarize findings in concise reports
- Update the knowledge base with new information
responsible for writing code, testing, and deploying software.
- Write code in various programming languages
- Test code using automated testing frameworks
- Deploy code to production environments
- Debug and troubleshoot issues
responsible for analyzing the AI's performance and identifying areas for improvement.
- Analyze performance metrics (e.g., accuracy, efficiency)
- Identify bottlenecks and areas for optimization
- Develop and implement improvements to the AI's architecture or training data