Skip to content

Incorrect outputs and exceeded tokens errors #21

@ProMicke

Description

@ProMicke

When I run with deepseek-reasoner for the reasoner_agent, and gpt-4o-mini for the other agents, the context window always goes > 128k, so it results in an "exceeded tokens" error. The test was done with the example prompt "search the web" agent.

And when I replace gpt-4o-mini with gemini-2.0-flash (which has much larger context window), it manages to generate an answer. However, I get incorrect and not so long answers. The code for the pydanticAI agent that it tries to form is completely incorrect and not as all as good as the answers generated in the youtube video for Archon.

Do I have to use the o3-mini as reasoner to make this work as good as in the youtube video? Or what can the issue stem from?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions