F

Flyte enables you to build & deploy data & ML pipelines, hassle-free. The infinitely scalable and flexible workflow orchestration platform that seamlessly unifies data, ML and analytics stacks. Explore and Join the Flyte Community!

Agent Development Execution Issues

Summary

The user is developing a new agent and encountered execution issues in a sandbox related to TaskSpecification and Pod configuration during workflow scheduling. They suspected the correct agent was not being identified and sought advice on updating task and plugin settings. The user planned to review agent pod logs, specifically looking for logs that show only sensors and exclude their custom agent, as the absence of their agent could indicate a problem with their flytekit implementation. They expressed interest in sharing their use case in a blog or community sync if the agent is successful. Ultimately, the user resolved their main issue by installing the agent as a plugin, updating their pyproject.toml and path structure, and ensuring all agent metadata were JSON serializable. They also invited further questions to improve documentation.

Status
open
Tags
    Source
    #flyte-agents