v1.0.0 — CAIS 2026 Artifact Evaluation

Artifact for the paper: Securing the Agent: Vendor-Neutral, Multitenant Enterprise Retrieval and Tool Use (CAIS 2026)

Quick verification (no API key needed)

git clone https://github.com/varshaprasad96/ogx-evals.git
cd ogx-evals
uv sync --frozen
./run_all.sh --analysis-only

Or with Docker:

docker build -t ogx-evals .
docker run --rm ogx-evals

Contents

• 6 experiments covering security, latency, throughput, adversarial resilience, retrieval quality, and predicate pushdown scaling
• 9 scripts, 5 configs, 300 synthetic documents, pre-computed results
• 80 pytest tests with deterministic synthetic embeddings
• Dockerfile for containerized execution

See REPRODUCING.md for claim-to-command mapping.