https://localusecases.com/Real people running AI on their own machines.Mon, 19 Oct 2026 03:39:24 +0000https://localusecases.com/use-cases/home-rag-assistant-for-embedded-systems-docs-on-proxmox/https://localusecases.com/use-cases/home-rag-assistant-for-embedded-systems-docs-on-proxmox/Ollama, llama.cpp, Open WebUI, BGE-M3 and SearXNG in a GPU-passthrough VM. Moving from 6 GB to 16 GB of VRAM changed which models could run.https://localusecases.com/use-cases/auto-tagging-documents-in-paperless-ngx-with-ollama/https://localusecases.com/use-cases/auto-tagging-documents-in-paperless-ngx-with-ollama/Since version 3.0, Paperless-ngx suggests titles, senders, document types and tags, and can chat with your documents, using a local Ollama model.https://localusecases.com/use-cases/self-hosted-bookmarks-that-tag-themselves-with-local-ai-karakeep/https://localusecases.com/use-cases/self-hosted-bookmarks-that-tag-themselves-with-local-ai-karakeep/Karakeep (formerly Hoarder) runs on a Debian LXC, and gemma3:4b in Ollama tags each bookmark. Getting reliable tags from a small model takes one key setting.https://localusecases.com/use-cases/offline-meeting-minutes-with-whisper-cpp-ollama/https://localusecases.com/use-cases/offline-meeting-minutes-with-whisper-cpp-ollama/Turns a 1-hour meeting into structured minutes in under 4 minutes on an RTX 3060, with zero network requests. Bash and Python pipeline included.https://localusecases.com/use-cases/paperless-ai-on-an-8-gb-rtx-3060-ti-shared-with-frigate/https://localusecases.com/use-cases/paperless-ai-on-an-8-gb-rtx-3060-ti-shared-with-frigate/Gemma 3 4B with embeddinggemma answers over 1,318 documents in about 5 seconds, on a GPU that's also running Frigate and Plex.https://localusecases.com/use-cases/overnight-video-to-rag-pipeline-for-2-000-saved-clips/https://localusecases.com/use-cases/overnight-video-to-rag-pipeline-for-2-000-saved-clips/A local Qwen3.6 35B with ffmpeg, scenedetect and Whisper indexes saved TikToks and YouTube videos, so you can later find 'that one plumbing tutorial'.https://localusecases.com/use-cases/home-assistant-voice-pe-with-a-local-llm-instead-of-echo/https://localusecases.com/use-cases/home-assistant-voice-pe-with-a-local-llm-instead-of-echo/A $60 Voice Preview Edition, an N100 box and an optional RTX 3060 replace every Echo and Nest in the house. It keeps working with the internet unplugged.https://localusecases.com/use-cases/a-private-chatgpt-for-a-family-of-four-on-a-mac-mini/https://localusecases.com/use-cases/a-private-chatgpt-for-a-family-of-four-on-a-mac-mini/One always-on M4 Mac mini serves Open WebUI to every phone and laptop in the house. Kids get a tutor preset, and three AI subscriptions are gone.https://localusecases.com/use-cases/hybrid-workflow-local-for-private-and-repetitive-cloud-for-hard/https://localusecases.com/use-cases/hybrid-workflow-local-for-private-and-repetitive-cloud-for-hard/A Linux sysadmin stopped trying to replace cloud LLMs. Ollama now handles private docs, overnight article summaries and VS Code autocomplete.https://localusecases.com/use-cases/read-only-personal-admin-assistant-over-work-email-and-calendar/https://localusecases.com/use-cases/read-only-personal-admin-assistant-over-work-email-and-calendar/A project manager juggling a dozen projects has a local bot that builds a wiki from email, surfaces to-dos, sends reminders and drafts replies.https://localusecases.com/use-cases/fully-local-home-assistant-voice-assistant-on-a-gpu/https://localusecases.com/use-cases/fully-local-home-assistant-voice-assistant-on-a-gpu/Wake word to spoken answer in 2-3 seconds, with zero audio leaving the house. Built from faster-whisper, Piper and Ollama, with a cloud fallback.https://localusecases.com/use-cases/offline-rag-over-268-obsidian-notes-queryable-from-a-coding-agent/https://localusecases.com/use-cases/offline-rag-over-268-obsidian-notes-queryable-from-a-coding-agent/nomic-embed-text, ChromaDB and mistral-nemo:12b give cited answers from a vault of internal docs, and an MCP server lets the coding agent search it too.https://localusecases.com/use-cases/replacing-google-nest-minis-with-local-assist-on-llama-cpp/https://localusecases.com/use-cases/replacing-google-nest-minis-with-local-assist-on-llama-cpp/After a year of Nest Minis getting dumber, a household moved to fully local Assist. Higher-quant GGUF models from Hugging Face are what finally fixed tool calling.https://localusecases.com/use-cases/self-hosted-ai-email-assistant-on-linux-n8n-ollama/https://localusecases.com/use-cases/self-hosted-ai-email-assistant-on-linux-n8n-ollama/Classifies IMAP mail with llama3.1:8b, routes it, saves reply drafts and extracts action items as JSON. Reports 85-95% classification accuracy.https://localusecases.com/use-cases/extracting-clean-data-from-old-sec-filings-with-small-models/https://localusecases.com/use-cases/extracting-clean-data-from-old-sec-filings-with-small-models/7B-13B models running locally turn wordy 1990s SEC text filings into structured tables, about 97% accurate on a random sample. Lots of tiny prompts, because tokens are free.https://localusecases.com/use-cases/mdrag-a-rust-cli-to-chat-with-an-obsidian-vault/https://localusecases.com/use-cases/mdrag-a-rust-cli-to-chat-with-an-obsidian-vault/A self-contained RAG tool built on SQLite with sqlite-vec and gemma3:1b. It answers 'how did I fix that Git submodule issue?' from your own notes.https://localusecases.com/use-cases/connecting-a-private-obsidian-vault-to-ollama-on-macos/https://localusecases.com/use-cases/connecting-a-private-obsidian-vault-to-ollama-on-macos/Meeting notes stay off public AI services. llama3.1 and nomic-embed-text in Ollama, wired up through the Local GPT and AI Providers plugins.https://localusecases.com/use-cases/mac-studio-as-an-always-on-llm-driven-smart-home-brain/https://localusecases.com/use-cases/mac-studio-as-an-always-on-llm-driven-smart-home-brain/Phrases like 'it's stuffy in here' become real device calls, run by Ollama with qwen3:14b on a Mac Studio. €15 ESP32 satellites keep all audio on the LAN.https://localusecases.com/use-cases/receipt-extraction-and-bank-matching-with-papercortex/https://localusecases.com/use-cases/receipt-extraction-and-bank-matching-with-papercortex/An open-source local AI layer for Paperless-ngx. It answers plain-English questions, extracts receipt line items, matches bank transactions and exports to DATEV.https://localusecases.com/use-cases/obsidian-vault-q-a-with-mini-rag-and-llama-cpp/https://localusecases.com/use-cases/obsidian-vault-q-a-with-mini-rag-and-llama-cpp/An Obsidian plugin that chats with a note or folder, using separate local llama.cpp servers for chat and embeddings (Ollama works too).https://localusecases.com/use-cases/security-camera-descriptions-and-search-with-frigate-genai/https://localusecases.com/use-cases/security-camera-descriptions-and-search-with-frigate-genai/Frigate sends camera frames to a local vision model (qwen3-vl via Ollama or llama.cpp) for searchable event descriptions, review summaries and chat.https://localusecases.com/use-cases/email-triage-and-morning-digest-with-a-local-ollama-model-n8n/https://localusecases.com/use-cases/email-triage-and-morning-digest-with-a-local-ollama-model-n8n/Classifies each email as URGENT, REPLY_TODAY, FYI or JUNK using gemma4:e4b, sends urgent alerts and builds a grouped morning digest.https://localusecases.com/use-cases/auto-label-gmail-and-draft-replies-with-a-local-model-n8n/https://localusecases.com/use-cases/auto-label-gmail-and-draft-replies-with-a-local-model-n8n/Sorts unread Gmail into six categories with Ollama, applies labels, drafts replies for urgent mail and logs everything to Sheets, with zero paid API costs.https://localusecases.com/use-cases/local-first-meeting-notes-app-whisper-cpp-ollama/https://localusecases.com/use-cases/local-first-meeting-notes-app-whisper-cpp-ollama/A macOS app that records your mic and the call's audio as two tracks, transcribes locally and writes a Markdown note with 'You vs. Them' speaker labels.https://localusecases.com/use-cases/14-months-local-first-what-still-breaks/https://localusecases.com/use-cases/14-months-local-first-what-still-breaks/An honest report from a three-machine home lab. Local handles about 85% of tasks, but four kinds of work still fail, and that cut cloud spend from $187 to about $25 a month.https://localusecases.com/use-cases/xygen-a-private-desktop-chat-app-for-ollama/https://localusecases.com/use-cases/xygen-a-private-desktop-chat-app-for-ollama/Conversations, files and settings live in a file in your home directory. It talks only to Ollama on 127.0.0.1 and adds optional embeddings, vision, voice and image generation.https://localusecases.com/use-cases/branching-conversation-graphs-with-openyoke/https://localusecases.com/use-cases/branching-conversation-graphs-with-openyoke/A Tauri and Rust desktop app for Ollama where every exchange is a node you can branch from, and sibling branches never bleed context into each other.https://localusecases.com/use-cases/offline-chat-on-desktop-or-android-via-termux-with-prefrontal/https://localusecases.com/use-cases/offline-chat-on-desktop-or-android-via-termux-with-prefrontal/A privacy-first chat UI that works with Ollama on desktop or llama.cpp anywhere, including Android through Termux. No telemetry, no subscriptions.https://localusecases.com/use-cases/leaving-google-photos-for-immich-with-local-ai-search/https://localusecases.com/use-cases/leaving-google-photos-for-immich-with-local-ai-search/CLIP smart search and face recognition run on the author's own NAS. A GTX 1070 turned days of indexing into one night.https://localusecases.com/use-cases/offline-llm-on-a-raspberry-pi-5-for-classification-and-extraction/https://localusecases.com/use-cases/offline-llm-on-a-raspberry-pi-5-for-classification-and-extraction/Phi-3 mini at Q4_K_M runs at 4-6 tokens per second on an 8 GB Pi 5 with no network at all. Under GDPR, data that never leaves the board isn't transferred anywhere.https://localusecases.com/use-cases/aider-qwen2-5-coder-14b-as-a-local-claude-code-alternative/https://localusecases.com/use-cases/aider-qwen2-5-coder-14b-as-a-local-claude-code-alternative/A tic-tac-toe test on an RTX 5060 Ti 16 GB. The game worked on the first pass, but the tests and input validation needed several rounds of prompting.https://localusecases.com/use-cases/copilot-style-coding-assistant-from-a-headless-rtx-3090-box/https://localusecases.com/use-cases/copilot-style-coding-assistant-from-a-headless-rtx-3090-box/qwen2.5-coder:7b on a home server, reached from a laptop over an SSH tunnel and wired into VS Code through Continue for chat, edits and autocomplete.