cvat-deploy
Skill8 starsChanged 8 months agoMITView on GitHub
Deploy a trained YOLO model as a Nuclio serverless function for CVAT auto-annotation.
In this repository15 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- active-learningSkill
- analyze-resultsSkill
- annotateSkill
- auto-experimentSkill
- benchmarkSkill
- compare-modelsSkill
- cvat-pullSkill
- cvat-pushSkill
- explain-resultsSkill
- monitor-trainingSkill
- review-annotationsSkill
- review-datasetSkill
- setup-projectSkill
- training-loopSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Use the best model from the latest experiment, or ask the user
- Very specific: line 21
Check experiments/summary.md for the best-performing model path
- Very specific: line 33
Read the generated function.yaml and confirm:
- Distinctive: line 48
Check that the function is running.
- Distinctive: line 67
NEVER deploy without confirming class mapping matches CVAT project labels
- Distinctive: line 68
ALWAYS verify the ONNX export succeeded before deploying
- Distinctive: line 69
Document the deployed model version and metrics in experiments/
- Distinctive: line 70
If deployment fails, show the Nuclio dashboard URL for debugging (http://localhost:8070)
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Deploy Model to CVAT
Deploy a trained YOLO model as a Nuclio serverless function so CVAT users can auto-annotate images directly from the CVAT UI.
Pre-Flight Checklist
- Trained model exists (best.pt or user-specified)
- Nuclio is running (check port 8070)
yolo-project.yamlhas class definitions- Docker is running (required for Nuclio builds)
Workflow
1. Identify the Model
- Use the best model from the latest experiment, or ask the user
- Check
experiments/summary.mdfor the best-performing model path
2. Generate Nuclio Function
yolo-cvat deploy --model <path_to_model> --name <detector_name>This generates:
serverless/<name>/function.yaml— Nuclio config with class specserverless/<name>/main.py— Inference handlerserverless/<name>/best.onnx— Exported model
3. Verify Generated Files
Read the generated function.yaml and confirm:
- Class names match
yolo-project.yaml - Image name is unique
- ONNX model was exported correctly
4. Deploy to Nuclio
Provide the command for the user to run:
nuctl deploy --path ./serverless/<name> --platform local5. Verify Deployment
nuctl get functionsCheck that the function is running.
Decision Tree
Model is .pt format?
├── Yes → Export to ONNX first (yolo-export)
└── Already .onnx → Copy directly
Nuclio reachable on port 8070?
├── Yes → Ready to deploy
└── No → Warn user, suggest checking Docker and CVAT stack
Function with same name exists?
├── Yes → Ask user: overwrite or use different name?
└── No → Deploy normallyGuardrails
- NEVER deploy without confirming class mapping matches CVAT project labels
- ALWAYS verify the ONNX export succeeded before deploying
- Document the deployed model version and metrics in experiments/
- If deployment fails, show the Nuclio dashboard URL for debugging (http://localhost:8070)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

