Agents are developing rapidly, but still face problems
How to perceive the environment and have memory.
How to reasonably plan and decompose tasks.
How to make autonomous decisions and take action.
How to implement Agents into real business scenarios in enterprises.
Build Agents like building a team
GPTBots builds multiple Agents to respond to tasks through concrete professional roles.
Set task goals
Set a clear and specific task goal for Agents, which will affect the task planning and output results of Agents.
Define the best team combination
Define the professional roles that need to participate in the task goal for Agents. Different roles will play different roles in the task and will autonomously coordinate and synchronize information.
Equip with effective tools
The tools equipped for Agents determine the ability boundary. The knowledge base, memory, database, code interpreter, Tools and other capabilities supported by GPTBots can help Agents better achieve task goals.
The running environment configured for Agents
GPTBots provides 2 different levels of running environments to meet the needs of different scale businesses.
Standard configuration
Provide an independent code sandbox hardware environment for executing code tasks.
Advanced configuration
Provide a higher performance sandbox hardware environment for advanced users, executing code tasks faster.
The key to letting Agents perceive, plan and execute autonomously
Provide various components, let the truly powerful LLMs and Plugins provide driving force, while making LLMs controllable and stable.
Knowledge base
The knowledge base equipped for Agents is not only compatible with various formats of knowledge, but also has a powerful RAG knowledge recall scheme.
Memory ability
It has long-term memory, short-term memory, user attributes, and timed tasks at the user dimension.
Tools
Not only supports calling plugins for LLM, but also traditional API calls are easy.
Code interpreter
Write code, execute code, build service.
Independent operating environment and database
Thousands of python libraries, independent service running space and MySQL database.
LLMs
Choose LLM autonomously, effectively reducing costs.
Support the choice of professional models in different ability quadrants.
Intelligent routing avoids over-frequency and ensures service stability.
