• Ollama and Open WebUI
• Knowledge/RAG setup for team documents
• RAG Retrieval Quality Audit fast path with fixed questions and cited-answer review
• Qwen3-Embedding 0.6B/4B and Qwen3-Reranker 0.6B/4B first, 8B only after fit checks
• Visual RAG and evidence search for screenshots, scans, diagrams and product images
• Qwen3-VL-Embedding and Qwen3-VL-Reranker benchmark option
• User roles, source citations and private domain setup
• Benchmark-first rollout plan with prioritized support

For teams that need source-backed internal knowledge and visual evidence workflows with managed operations, role boundaries, source citations, and a measured retrieval audit before monthly production scope. Live inference claims stay gated until the actual server passes GPU, Ollama, and target-model smoke tests.