Activity
Mon
Wed
Fri
Sun
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
What is this?
Less
More
38 contributions to Agent Zero
0 likes • 48m
@Mike Doolittle I would be interested in... forking? off and starting a Local LLM agentzero group, but monthly service fees and how to get people to join - the new people I suppose could come from video posts on YouTube.
0 likes • 30m
@Mike Doolittle I seriously would pull something together so long as I could monetize it somehow to offset my time involved. As you are probably aware Local LLM setups and running them every day is an expensive hobby! Would you join me on YouTube? I could probably start a free community there and it would be more appropriate for me to share my screen and give walk-throughs. I could post links here so others can find us.
Aug 20 •
I’m putting A0 on hold for myself, but I won’t stop believing in a better future for it!
Hi everyone! It looks like we’re reaching a point where local development tools are actually improving, rather thanJust seeing the emergence of high-quality agent frameworks like DeepSeek Harness makes it clear why they have become popular. No matter how often I tried using Agent Zero for local development, I invariably hit a dead end. Consequently, this tool will likely remain of little relevance unless its architecture undergoes radical changes—specifically, the removal of the auxiliary ("utility") model; in my view, that is Agent Zero's main problem. Competent memory plugins could surely solve this issue, thereby making A0 more flexible.
0 likes • 1h
@Mike Doolittle See my other response if you haven't already.
0 likes • 1h
@Renat Yaraev Would be happy to invest the time to share, but it is too much to type here. Give a like if you are interested, that's the way I operate on here. I elaborate when I see people show interest. Happy to help, but this will need it's own separate post.
4d •
Proof You CAN Run Local LLMs
I see people on here all the time saying that you cannot run AgentZero with Local LLMs. People think that somehow this is impossible - or that you have to sacrifice intelligence - NOT TRUE. I have been running strictly offline local LLMs since day 1 with AgentZero, I pay ZERO api fees! What makes this possible is is having a good engine to prefill the prompt and a good LLM server. Do not use Ollama, do not use vLLM and although I do like it - do not use LM Studio. You need to run a model server with what is called model streaming - this is how you can get around size limitations on your VRAM. I am getting crazy fast speeds with big & intelligent models now, because I take all this a step further and use speed projection, MTP, and the right Quant-size! All of these matter to get quality intelligent output WITH speed. Stop belly-aching that you can't run local LLMs with AgentZero, you can!
0 likes • 2h
See my other post, but to reiterate: Strata with 1 model (Qwen3.8-flash-next from ISTA-DASLab)
4d •
Smarts & Speed Local LLM for AgentZero
There is a way now. I do completely offline local development with my AgentZero. Granted yes, I do have a 24GB 3090, plus 96GB of RAM, but with that aside you can still do coding with less - yes it IS possible! For example, I ran a coding setup with AgentZero and it only required about 16GB of VRAM and I was getting over 100 tp/s and quality outputs - I did 2 small web apps and the code was flawless! This measures up to the performance of GPT Codex. And I don't pay anything for use! The secret is a really good engine to run your LLM with prefill AND model streaming. I have been waiting for these techniques to be available on private systems for a long time - and now they are here! You CAN run big smart models locally, with speed. So, you can have pro-level coding offline with AgentZero - I am doing it now!
0 likes • 2h
@Mike Doolittle I gave you the engine name on another post, but I will reiterate here. The secret is to use the Strata engine as your LLM server. Currently it has Qwen3.6 and Qwen3.8-flash-next to select as the models. I went with Qwen3.8-flash-next since it is a 125B model and great for code. In order to have a model that retains it's intelligence you want to select the one from ISTA-DASLab (look them up they have done excellent work to compress models and maintain quality). For me, I went with the IQ3_S which is 3.5 bit and in my experience is living up to what was advertised - it is indeed fast and smart. Tell ya what. I will give you a complete walk-through with all the steps and configuration and how to get around setup blocks/issues, but it is too much to type here. I only elaborate here on these forums when people like my posts. If people don't like my stuff I take that as they are uninterested so I don't bother to waste my time. I have a lot I could share - far more than this. Local LLM setups is all I work with. So If you want more, please "vote" for it. I want to share and I want to help, just trying to be conscious of my time and share what people really want on here.
2d •
Unwanted Secrets Placeholder
I typed into AgentZero "local directory" and what it provided in a response was §§secret directory. While this may be helpful at times, this is unwanted behavior for me. I am running local models and trying to debug local web applications. Another example is when in the output the agent was giving me codeblocks and instructions and I got this: Compare §§secret(API_KEY_OTHER) vs remote I never gave an API key, all I typed was "local directory" and now the agent is giving output that I have to modify by hand. This really slows down production. Is there a way to turn secrets off? Again, I am using local models and everything is local on my computer. I do not need these guardrails. Using latest AgentZero version 2.14 Other than that, thanks for building an awesome AI agent framework! @Alessandro Frau @Jan Tomášek
0
0
1-10 of 38
@daniel-p-4970
Certified IT professional and Linux user. Investigative researcher.
Online now
Joined Mar 2, 2026
USA
Powered by