Immediately after switching the page, it will work with CSR.
Please reload your browser to see how it works.
building agents with DAGs, Graphs, and Chains means you're forcing imperative guardrails instead of allowing the agent to reason about which step to take in what order. i believe these will soon be considered tech debt as models like o1 are widely adopted
building a team of declarative reasoning agents means providing instructions and tools, and allowing them to communicate with each other to accomplish work
i want to make building these multi-agent teams and then deploying and monitoring them in production easy for anyone (not just engineers), so i'm building aster agents
bring your own llm api keys and connect integrations to unlock tool calls
for custom tool calls, you can bring your own API to dispatch them
otherwise, you don't need code
It uses vercel, including the ai sdk, for streaming chats. I also built a python backend for handling tool calls. It can handle webhooks, API calls, and URL scraping for your assistant so you don't have to build those yourself. Add in code interpreter and file search from the native OpenAI functionality and you can make some really powerful assistants.
The best examples that my customers have come up with so far are using Zapier / Make.com to connect their assistant to other tools via webhooks. Anything from chatting with leads and collecting contact info, to generating Google Slides or Docs with information inputted in the chat.