Activity
Mon
Wed
Fri
Sun
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
What is this?
Less
More
Memberships
4 contributions to Agent Zero
3d •
Latest Agent Zero Version?
Is the the latest version still the 0.9.8.2? I had thought I seen somewhere that there was an update that not has Telegram and a bunch of other stuff added?
0 likes • 3d
That is the latest release. there is quite a lot of talk about the dev version which you can cheque out on Github
6d •
Mistral API impossible to use in Agent Zero?
Hello, I tried to use payed Mistral-API for Agent Zero. In simple Chat-Situation no problem, but if I gibe Agent Zero a complicated job, it crashes. AI tell me, it is because there are too many requests and Mitral clode the door (I AM PAYIMNG!), so I try, that the reuqests had to run slowly each by each, but I Cannot set this in Agent Zero. Sad, because in europe, I want not use an API which is connected to US-Cloud-Act. Had someone clear this issue? Thanks for a tipp!!! Chris
2 likes • 5d
Hi Christoph, I was interested in this issue so I asked to support but in agent zero discord. Is what came back.. It is a common frustration to encounter rate limits (429 errors) even on a paid tier, especially with European providers like Mistral who maintain strict traffic management. When Agent Zero performs a "complicated job," it often spawns multiple processes or sub-agents simultaneously, which can exhaust your API quota in seconds. Based on the current architecture of Agent Zero and its integration with LiteLLM, here is how you can resolve this and implement the throttling you need. ### 1. Throttling via Environment Variables (Most Reliable) Agent Zero uses a configuration system that allows you to inject settings directly into the underlying model handler. You can force the agent to slow down by setting specific "Rate Limit" variables in your `.env` file. Add or modify these lines in your **`.env`** file: * **`A0_SET_chat_model_rate_limit_rpm=3`** (This limits the agent to 3 Requests Per Minute. Start low and increase it until you find the breaking point.) * **`A0_SET_chat_model_rate_limit_tpm=50000`** (This limits the Tokens Per Minute.) By setting the **RPM (Requests Per Minute)** to a low number, Agent Zero is forced to queue its requests and wait for the timer to clear before sending the next one, effectively running them "slowly each by each." ### 2. Adjusting LiteLLM Retries Agent Zero relies on **LiteLLM** to communicate with Mistral. LiteLLM has a built-in "exponential backoff" feature, but if the "complicated job" is too large, the default retries might not be enough. You can increase the number of times the system tries to send a request before crashing by adding this to your `.env`: * **`LITELLM_NUM_RETRIES=10`** This tells the system: "If Mistral closes the door, wait a moment and try again up to 10 times before giving up." ### 3. Managing Subordinate Agents The reason "complicated jobs" cause crashes is often **parallelism**. Agent Zero might try to:
0 likes • 5d
@Christoph Baumgarten awesome
5d •
"Vector DBs: Agent Zero Never Forgets—Unlike Me When My Wife Says 'Are you listening ...'"
Ever wondered how AI agents—like in Agent Zero—can "remember" lots of information the LLM's produce and the data held in documents we feed it, grab the right info instantly, and skip bloating the model's context? Checkout vector databases and semantic search with embeddings. Basically, it gives your AI agents a fast, meaning-based memory that actually works. I just watched this clear Computerphile vid that explains it well (with examples and code): https://youtu.be/YDdKiQNw80c?si=8XWOyWkBJZDhm60h I had no clue about vector databases before looking into agent zero and this video really helped me understand how storage and retrieval in agent zero works since Agent Zero uses **FAISS** for quick, local similarity searches across main memories, conversation chunks, and proven solutions. Worth a look—short, no fluff. Anyone messing with vector DBs like **Chroma**, **Pinecone**, or **FAISS** in their agents? Curious what you're running! 🚀
1
0
6d •
Using Brave Search API in Agent Zero
Hey everybody, so happy to be part of this project, I am a newbie, don't know much, about 2 months in learning about agents but with the help of chatgbt guiding me successfully setup everything, got my brain agent setup with kimik 2.5, my question is about using brave search via API and if this is possible? If so how would I configure this in agent zero? Looking forward to your response 😀👍
1 like • 5d
I tend to just use Agent Zeros built-in search tools. Have you tried them? Also, I think I read Kimi K2.5 had it's own web search to built in. As long as your prompt includes the instructions to use specific tools A0 should comply
1 like • 5d
Watch the A0 execution flow in the main chat window to see which tools it uses and analyse the results
1-4 of 4
