Hello all 
,
Second post of this, as I deleted the first one so that I could fix a packaging bug and verify it. 
As someone who learned best through doing and not reading, that often involves making lots of mistakes and learning from them. Equally, as someone who’s passionate about AI - I’ve seen it equally have issues due to outdated knowledge etc.
To help others like me get quick feedback in terms of mistakes with configuration - I have created a VS Code Extension entitled “CrewAI Lint”. It’s basic as the result of a few hours of Saturday afternoon (and Sunday morning) coding but I thought best to get it out into the world already for feedback.
Functionality
-
YAML: Only works with agents.yaml and tasks.yaml, not supporting Python.
-
Required Attributes: Does basic checks to ensure required attributes are provided
-
Correct Attributes: Makes sure only supported attributes are provided/checks for correct casing
-
Agent Existence: Makes sure the agent specified for a given task actually exists
-
Version Compatibility: Works for the latest 0.102.0
Here’s an example, having intentionally mis-named an agent reference from the crewAI-examples repo:
I’ve tested it locally in a number of different ways, created a small test suite and also verified against all of the YAML examples within the crewAI-examples repo: Which all pass except for landing-page-generator which has a few tasks missing values for expected_output.
How to use it?
-
It’s published on the Marketplace
-
It’s fully open source, so you can package it yourself: GitHub - wsimmonds/crewai-lint: CrewAI Linter for Visual Studio Code
Feedback and contributions are welcome, and I apologise in advance for any bugs but hope this might help a few people along the way!
-Will