Hello Folks, I am trying to create a simple Agent that downloads a paper and some of the key citations from Arxiv and trying to create a presentation. However, I do see that the agent keeps going in a loop. # Agent: You are a Junior Researcher ## Thought: Thought: I have attempted the same extract…

If you only want local models, can you try Qwen 2.5 models, I saw someone mentioning that they’re great at tool calling. You might have to evaluate various local models until you get one that suitable for your use case.

Hi, I am also seeing the same problem. I am trying to create a simple agent that analyzes the data provide in a csv file. Looks like that the agent goes into a loop with the following output repeated. Can someone please help? ----------------------- start of output -------------------------------…

I encountered the same problem. When I set allow_code_execution=True , an infinite loop occurred. I guess this is because the prompt prepared by crewai highlighted tool usage, so the model thought it was necessary to use the tool.

Still no answer from your tests ? I’m having the same issue with summarizing a pdf.

Can you give a bit more detail on what you are doing and the errors you are getting. The exact same ones mentioned above?

Yes this exact message : “I tried reusing the same input, I must stop using this action input. I’ll try something else instead.” I wrote a tool to extract some text from a PDF. The first time, the agent uses the tool correctly and output the text but do nothing else (I asked for a summary). Then, t…

I’m using ollama loacally with llama3.2

so Ollama models are a hit or miss with tool calling. can you try the same with gpt-4o-mini and let me know if it still throws the same errors. beyond this, we can try different techniques of processing the paper like using knowledge in crewai ( Knowledge - CrewAI ). you can write a function to downl…

I will not use gpt-4o-mini (sorry) as it is non-local and cost money while I already have the computing capabilities in house. I’m trying to do a proof of concept for local agent based LLM for people here at institut Pasteur. We do not have the right to send data outside of the company, so, even fo…

Agents keeps going in a loop

Crews

maxmoura April 3, 2025, 3:01pm 10

Is the returned text longer than 2048 tokens? If that’s the case, this issue was discussed in this thread.

Topic		Replies	Views
Agent preparing response using output from last invocation of same tool Crews tools_issues , agent	2	121	March 9, 2025
Agent is repeating its iteration even after the final answer CrewAI Community Support tools_issues , agent , crewai	1	650	February 18, 2025
Kickoff_for_each - Can I have same crew execute tasks sequentially for each input Crews agent , crewai	0	90	June 16, 2025
Agent stops in the middle and never goes to the end of the input list CrewAI Community Support openai , agent , task	7	548	December 12, 2024
Creating a Conversable Agent in CrewAI with Human-in-the-Loop Interaction CrewAI Community Support	21	10730	June 30, 2025

Agents keeps going in a loop

Related topics