LocalSutra AI documentation

Know your hardware. Choose your model. Keep AI local. An offline, evidence-first local-AI readiness toolkit.

v0.2.1126 passed · 0 skipped · 0 failed0 runtime dependenciesApache-2.0Honest: loopback only, never installs or downloads
Executive SummaryLocalSutra AI 0.2.1 is an open-source (Apache-2.0), offline-first command-line tool that tells people whether their own computer can run a lFunctional SpecificationStudents and hobbyists with ordinary laptops, privacy-minded individuals, educators, developers using Ollama or llama.cpp, researchers compaArchitectureuser / IDE agent --argv--> CLI --> core --> node:os, statfs, nvidia-smi (measured)Technical ReferenceNode.js 22 or newer. No runtime or development dependencies. Optional: nvidia-smi for GPU facts; Ollama or llama.cpp for probe/infer.User Guidenpx localsutra-ai doctor # no installSecurity and Privacy1. Read-only by default. No installs, downloads, driver changes, listening ports or telemetry.Evidence Model and RulesTwo benchmark reports are compared only when workload, iteration count, platform and architecture match; otherwise the tool refuses. DiffereTest ReportMachine-generated figures from the recorded run. Regenerate with npm run proof.Operations and Release- npm package localsutra-ai (binaries localsutra, localsutra-ai), Apache-2.0.Legal and GovernanceEngineering documentation, not legal advice.Roadmap and FAQCan my computer run a 7B model? Run localsutra fit. If it says possiblyfits, memory looks plausible. Load it in Ollama to find out for real.GlossaryAssurance and EvidenceThis page lists what a skeptical reviewer can check, how to check it, and what remains unproven.Sample audit reportReal HTML output of localsutra audit (37 adversarial checks)Raw test log (TAP)Unedited machine output