Multi-Agent
Several agents working on a task, with defined roles or handoffs.
Aperçu
Multi-agent systems split a task across agents that each hold a narrower instruction, and have them exchange results. It solves the problem that one prompt carrying every constraint produces a model that satisfies none of them well. It exists because separating concerns works for prompts as it does for code, with the cost that coordination failures replace reasoning failures.
Graphe des technologies
Comment Multi-Agent se rattache au reste de la pile.
Utilise et dépend de
- Agent Memory technique AI Agents Multi-Agent uses Agent Memory High ✓ source
- Function Calling technique Prompting Multi-Agent uses Function Calling High ✓ source
- LLM concept Models Multi-Agent based on LLM High ✓ source
Outils
Logiciels qui implémentent Multi-Agent ou s'y intègrent.