AI
Open weights run. EdgeLance decides.
Open-weight models run on the device. EdgeLance places the work, audits the actions, cites the sources, and serves each role the right view.
Open weights
Model runtime.
Open-source inference baseline.
- Open-weight LLMs (Gemma, Phi, Llama, Mistral)
- Open-weight VLMs and embeddings
- Open ASR and recognition models
- Permissive licensing (Apache, MIT)
- Edge runtimes (MLX, llama.cpp, ONNX, Ollama)
- Quantization for on-device inference
- Pre-trained on web-scale data
- Hardware-agnostic inference
EdgeLance AI
Mission AI fabric.
Place, audit, cite, serve.
- Inference broker with capacity and thermal policy
- Agent orchestrator with action receipts and audit
- Threat analyzer: multi-modal reasoning over alerts
- Mission RAG with source-trusted citations
- Mission narrator and missed-update summaries
- Intel curation and source-trusted answers
- Mission-wide search across chat, Vault, markers, sources
- Model control plane (any model, swappable, route receipts)
Backed by provisional patent applications filed with USPTO in June 2026 covering the scope-bound edge intelligence control plane, policy-gated agent execution with dynamic node placement, mission pack compiler with fit gates, clearance-bound retrieval-augmented generation with no-parametric-fallback refusal, multi-modal edge-tier curation with capacity-policy routing, and the policy-denial evidence ledger.