Skip to content
FaradAiLab Docs
Enterprise Beta

Frequently Asked Questions (FAQ)

Frequently Asked Questions (FAQ) ❓

How much latency does FaradAiWire add to tool calls?

For typical tool invocations evaluating fast-path rules (Prefill block, Rate Limiting, Schema Validation, and DLP Regex Masking), FaradAiWire adds less than 1 millisecond (< 0.5 ms) of P99 latency overhead. When the Deep-Path Neural SLM Guard is engaged for ambiguous or untrusted content, local inference adds approximately 15 to 40 ms, which is negligible compared to standard LLM network latency (1000–3000 ms).

Do I need a GPU to run FaradAiWire?

No. FaradAiWire’s core engine, regex DLP, schema pinning, and loop protectors are written in compiled Go machine code and require minimal CPU and < 30 MB of RAM. If you choose to enable the optional Deep-Path SLM Guard, lightweight models like prompt-guard (86M parameters) or qwen3guard-0.6b execute with sub-20ms latency on modern CPUs and Apple Silicon (M1/M2/M3/M4) unified memory without requiring dedicated discrete NVIDIA GPUs.

Does FaradAiWire transmit my code or tool data to any third-party cloud?

Absolutely not. The open-source FaradAiWire Core binary operates under a strict Zero-Egress Guarantee:

  • All regex pattern matching is performed in local memory.
  • Small Language Model inference runs on your machine via Ollama or local Docker sidecars.
  • Audit logs and analytics are stored in local files (logs/faradai_audit.jsonl and ~/.faradaiwire/telemetry.duckdb).
  • Zero bytes of prompts, parameters, or metadata are ever transmitted across the internet.

Why is FaradAiWire distributed as a native daemon instead of Docker by default?

As established in ADR-0051, over 95% of developer MCP tools (e.g. filesystem access, git, sqlite, local scripts) run as local stdio subprocesses (npx, python, git). Running FaradAiWire in a Docker container severs access to host directories and runtimes without complex bind mounts and virtualization slowdowns. Running as a native binary preserves native OS filesystem speed and seamless tool integration.

Can I run FaradAiWire in containerized environments?

Yes. If you are deploying in headless Linux environments or container clusters, you can build the container from deploy/docker/ and bind listeners with faradaiwire run --host 0.0.0.0 --standalone.

What is the difference between FaradAiWire Core and FaradAiLab Enterprise Edition?

  • FaradAiWire Core (Open Source): Apache 2.0 single-tenant Policy Enforcement Point (PEP) running on local developer workstations with fast-path deterministic filters.
  • FaradAiLab Enterprise Edition: Commercial multi-tenant control plane providing centralized Open Policy Agent (OPA) ABAC policy distribution, KMS BYOK crypto-shredding, SSO, and real-time SIEM audit streaming. Learn more in the FaradAiLab Enterprise Edition Docs.