feat(vllm_performance): Add experiment to test geospatial endpoint with a custom dataset and remove geospatial experiments from BETA features

[christian-pinto]

No description provided.

[DRL-NextGen]

No vulnerabilities found.

[michael-johnston]

@christian-pinto FYI plugins/actuators/vllm_performance/ado_actuators/vllm_performance/experiment_executor.py need black run on it.

[christian-pinto]

@christian-pinto FYI plugins/actuators/vllm_performance/ado_actuators/vllm_performance/experiment_executor.py need black run on it.

Weird, the pre-commit hooks had all passed when I pushed those changes. I will give it another pass as there are more changes coming.

[michael-johnston]

Ready for review?

[christian-pinto]

Ready for review?

Waiting for @mgazz to confirm he could replicate with guidellm on his side

[mgazz]

Works for me. For context, the only thing I had to do to switch between vllm bench and guidellm was to swap the experiment section in the discovery space

from

experiments:
  - actuatorIdentifier: "vllm_performance"
    experimentIdentifier: "test-geospatial-deployment-custom-dataset-v1"

to

experiments:
  - actuatorIdentifier: "vllm_performance"
    experimentIdentifier: "test-geospatial-deployment-guidellm-custom-dataset-v1"

Full discovery space definition: discoveryspace_geospatial.yaml

[christian-pinto]

@michael-johnston ready for review.

[DRL-NextGen]

Checks Summary

Last run: 2026-04-07T08:22:56.796Z

Code Risk Analyzer vulnerability scan found 5 vulnerabilities:

Severity	Identifier	Package	Details	Fix
🔷 Medium	CVE-2026-34756	vllm	vLLM: Unauthenticated OOM Denial of Service via Unbounded n Parameter in OpenAI API Server GHSA-3mwp-wvh9-7528 vllm:0.18.0	0.19.0
🔷 Medium	CVE-2026-34753	vllm	vLLM: Server-Side Request Forgery (SSRF) in download_bytes_from_url GHSA-pf3h-qjgv-vcpr vllm:0.18.0	0.19.0
🔷 Medium	CVE-2026-34755	vllm	vLLM: Denial of Service via Unbounded Frame Count in video/jpeg Base64 Processing GHSA-pq5c-rjhq-qp7p vllm:0.18.0	0.19.0
◻ Unknown	CVE-2026-34450	anthropic	Claude SDK for Python has Insecure Default File Permissions in Local Filesystem Memory Tool GHSA-q5f5-3gjm-7mfm anthropic:0.86.0->anthropic:0.86.0,vllm:0.18.0	0.87.0
◻ Unknown	CVE-2026-34452	anthropic	Claude SDK for Python: Memory Tool Path Validation Race Condition Allows Sandbox Escape GHSA-w828-4qhx-vxx3 anthropic:0.86.0->anthropic:0.86.0,vllm:0.18.0	0.87.0