Self-evolving scientific research workbench

Key features

• Scientific research memory - each round of dialogue is automatically precipitated into structured memory, seven categories of classification + hybrid retrieval, and relevant memory packages are automatically injected before the model is called.
• Autonomous agent workbench for scientific research
• Where does the data live?
• Long-term goals - complex tasks are automatically broken down into target contracts, which are advanced based on evidence, auditable, and recoverable
• Project workspace - each project has an independent directory and independent git warehouse, and data is migrated with the project