Linux and Homelab

Linux, Docker, Proxmox, NAS, backups, monitoring, containers, virtualization, and home-server guides written for recoverable operations.

AI
Secure A LAN Model Server: Ollama And LM Studio Across VLANs

Define allowed clients, authentication and resource limits before exposing Ollama or LM Studio across VLANs. Evaluate each product's controls separately and check that direct backend access cannot bypass the intended path.

Read this guide
AI
Can a Local Coding Model Run an Agent? Test Tools, Patches, and Safety

Use a disposable repository to test a self-hosted coding model's API, tool calls, patch quality, refusal boundaries, and rollback before giving it real code.

Read this guide
AI
Is Your Self-Hosted AI Actually Private? Audit Every Data Path

A practical audit for local AI prompts, cloud model selection, web and vision sidecars, exposed APIs, logs, and backups before sensitive data enters the system.

Read this guide
AI
Local LLMs in the Codex App: OpenCodex Setup and Privacy Limits

Compare the native Codex CLI local-model path with OpenCodex's third-party proxy for the desktop app. Verify actual routing, hosted sidecars and recovery before treating a local model as a private workflow.

Read this guide
AI
Coding Agents In Docker Sandboxes: Test The Boundary

Distinguish Docker Sandboxes from ordinary containers before assessing a coding agent. Review mounts, credentials, sockets and network controls with harmless boundary tests, without claiming isolation makes escape impossible.

Read this guide
AI
Why Your Local Model Server Fails On User Two

Diagnose overlapping model requests through latency, queueing, failures and memory use. Compare consistent prompt workloads and cache conditions before setting admission limits from single-user throughput or another server's benchmark.

Read this guide
Linux and Homelab
Trusted Proxies Are an Authentication Boundary: Stop X-Forwarded-For Spoofing

Define the intended reverse-proxy chain, reject untrusted forwarding headers and block unauthorized direct-origin access. Plan allowed and denied path tests while keeping client IP attribution separate from user authentication.

Read this guide
Homelab
Rebuild A Dead Proxmox Host: What PBS Does Not Restore

Identify the host networking, storage, credentials and recovery records that guest backups do not establish. Plan an isolated Proxmox rebuild around the scope of your actual backup jobs.

Read this guide
Linux and Homelab
Media API Keys in Access Logs: Check for Exposure and Rotate Safely

Trace where media API keys may persist in URLs, logs, support bundles and backups. Search without exposing secrets, restrict original evidence, share redacted copies and rotate affected credentials across dependent integrations.

Read this guide
AI
Ollama vs llama.cpp vs LM Studio vs Lemonade: Test the Runtime, Not the Logo

A local-AI runtime comparison framework for Ollama, llama.cpp, LM Studio, and Lemonade: hardware fit, context, APIs, concurrency, privacy, and evidence.

Read this guide