Make AutoGen Consistent: CONTROL your LLM agents for ACCURATE Postgres AI Data Analytics


Prompt Engineering is only the FIRST STEP. What comes next is learning to CONTROL multi-agent conversations. To building agentic software that works on your behalf, we must guide the conversations of our agents. Microsoft’s AutoGen takes a brief shot at this with the GroupChat and GroupChatManager python classes but this is not enough for all use cases.

If you’re an engineer looking to build your own multi-agent applications like our Postgres AI Data Analytics tool, then ensuring consistency with your LLM agents is vital. In this video we dive into how we can CONTROL agent conversations and orchestrate agent function calls with precision. Building on our previous two AutoGen postgres data analytics video and codebase, we reveal two architectural patterns to help you get control of your agents. The key? A robust python orchestrator class which guides agentic engineering by helping us manage our LLM agents. With this, we guide agent conversations and ensure seamless function calls. We establish principles of LLM agent orchestrations and agent conversations aided by gpt-4, openai function calls, and AutoGen.

Discover, experiment, and learn with us. From mastering prompt engineering to multi-agent conversations powered by AutoGen we push our Postgres AI Data Analytics tool further. For those seeking insights into ai engineering and aspiring to build their own agentic software, this video was made for you. Let’s steer AI agents with precision, and learn to harness the power of agentic engineering.

Huge shout out for all the likes and subs on the previous two videos. Feel free to join the journey – we’re just getting started.


βœ… Watch Part Two – Using AutoGen to code a multi-agent postgres ai tool

πŸ€–πŸ’» AI Engineering Resources
Microsoft’s Autogen:

Autogen group chat example:

Free Postgres Hosting With Neon:

πŸ€– ZERO Touch coding with AIDER? YUP

πŸ“˜ Chapters
00:00 The channel has exploded BUT
00:45 What this channel is about
01:30 Recap our Postgres Multi-Agent Tool
03:43 Control Multi-Agent Conversations
05:24 Running the Data Eng Agent Team
09:45 The Sequential Conversation
11:20 The Broadcast Conversation
16:05 Multi-Agents with unique functions
17:55 Running the Data Viz Agent Team
20:19 Compose our two Multi-Agent Teams
24:50 Engineering is all about building blocks
25:20 The Order of Agent conversations matters
27:15 Where we’re headed next

πŸ› tags
#postgresql #agentic #promptengineering