RepoPicks Radar

Updated every day around 23:30 UTC

Guides

Local AI Packaged

coleam00/local-ai-packagedReviewed by AI

Checked by RepoPicks Radar on

What is Local AI Packaged?

Self-hosted AI Package is a ready-made Docker Compose set-up that starts a whole local AI kit in one go: Ollama for models on your own machine, Open WebUI to chat with them, n8n for low-code automation with example agent workflows included, Supabase as the database, Qdrant and Neo4j for search and graphs, SearXNG for web search, Langfuse for monitoring and Caddy for HTTPS. You install Docker, Python and Git, fill in a long file of passwords, then run one script. A graphics card makes the models usable; there is also a slower processor-only option.

Local AI Packaged is free and open source (Apache-2.0) and can replace ChatGPT, n8n Cloud, and Zapier.

How easy is it to install Local AI Packaged?

D · terminal, several tools or a specific graphics card · Windows

  • Windows D
  • macOS D
  • Linux D

Windows, Mac and Linux are all a D and there is no release to download. On each one you install Docker, Python and Git, clone the repository, copy `.env.example` to `.env` and generate about fifteen passwords and secret keys by hand, then run `python start_services.py --profile gpu-nvidia` (or `cpu`, `gpu-amd`, `none`). Macs with Apple chips cannot pass the graphics card to Docker, so they run on the processor or point at an Ollama installed outside Docker.

Coming soon: install in 1 click with the RepoPicks app.

Local AI Packaged on GitHub

Stars
3,790on Nov 10, 2026
Downloads
—all releases
Growth
—no gain in the last reading

Stars over time

3,7903,7913,792 Nov 6Nov 7Nov 8Nov 9Nov 10 3,7903,7913,792 Nov 6Nov 8Nov 10

Facts

Runs on
Windows, macOS, Linux
You need to install
Docker, Python, NVIDIA graphics card
To build from source
Docker, NVIDIA graphics card
Disk space
not determined
Accounts or keys
OpenAI (only for Supabase's SQL editor helper), Google sign-in for Supabase (optional)
Cost
Free · Every piece is open-source software running on your own machine, so there is nothing to pay. The same tools (n8n, Supabase, Langfuse, Flowise) also sell hosted versions, which this package replaces.
License
Apache-2.0 (Permissive)
Last change on GitHub
Feb 26, 2026

Local AI Packaged risk signals (9)

  • Attentionexample passwords that look ready to use

    `.env.example` ships values that work as they are: `DASHBOARD_PASSWORD=this_password_is_insecure_and_should_be_updated`, `NEO4J_AUTH=neo4j/password`, `N8N_ENCRYPTION_KEY=super-secret-key`, a filled-in `SECRET_KEY_BASE=UpNVntn3cDxHJpq99YMc1T1AQgQpc8kfYTuRgBiYa15BLrx8etQoXz3gZv1/u2oq`, and Supabase's public demo tokens in `ANON_KEY` and `SERVICE_ROLE_KEY`. The README's February 2026 note even tells everyone to copy the same `S3_PROTOCOL_ACCESS_KEY_SECRET=850181e4652dd023b7a98c58ae0d2d34bd487ee0cc3254aed6eda37307425907`. Whoever skips the "YOU MUST CHANGE THESE" step is running a database everyone's keys open.

    AI review

  • Attentiondocker control left open without protection

    The troubleshooting section says: "If using Docker Desktop: Go into the Docker settings and make sure 'Expose daemon on tcp://localhost:2375 without TLS' is turned on". That setting hands out full control of Docker, and therefore of the machine, over a network port with no password and no encryption.

    AI review

  • Attentionports the firewall does not cover

    The cloud section carries the author's own warning: "ufw does not shield ports published by docker, because the iptables rules configured by docker are analyzed before those configured by ufw", and `.env.example` adds "you might not want to expose Ollama or SearXNG since they aren't secured by default". On a rented server, services can end up reachable by anyone even after you set the firewall.

    AI review

  • Attentioncontainers always pull the newest version

    The images are unpinned: `n8nio/n8n:latest`, `ollama/ollama:latest`, `neo4j:latest`, `flowiseai/flowise`, `qdrant/qdrant`, `ghcr.io/open-webui/open-webui:main`. What runs on your machine is whatever those projects published last, and the README's own list of breaking changes (new Supabase variables in June and in February 2026) shows how that plays out.

    AI review

  • Infocode fetched from another project at start

    `docker-compose.yml` begins with `include: - ./supabase/docker/docker-compose.yml`, a folder that is not in the repository: `start_services.py` downloads Supabase's own repository when you run it, as the troubleshooting note confirms ("you had a 'bad' pull of the Supabase GitHub repository when you ran the start_services.py script").

    AI review

  • Infoa lot of computer for one package

    A single command starts around twenty containers — n8n, Ollama, Open WebUI, Flowise, Qdrant, Neo4j, SearXNG, Caddy, Langfuse with ClickHouse, MinIO and Redis, plus the whole Supabase set (Postgres, Kong, Studio, storage, analytics, pooler) — and on first start Ollama downloads two models by itself (`ollama pull qwen2.5:7b-instruct-q4_K_M; ollama pull nomic-embed-text`). The repository gives no figure for disk or memory.

    AI review

  • Infoworkflows can run commands on the machine

    The README explains how to switch on the n8n nodes that newer versions disable: "Starting with n8n v2+, the `Local File Trigger` and `Execute Command` nodes are disabled by default for security reasons", by uncommenting `NODES_EXCLUDE=[]`. With that on, any workflow you import can run commands in the container and read the shared folder mounted from your disk.

    AI review

  • Infomonitoring sends usage data out

    Langfuse starts with `TELEMETRY_ENABLED: ${TELEMETRY_ENABLED:-true}`, so by default the monitoring container reports usage to its makers, inside a package sold on being local. n8n's own `N8N_DIAGNOSTICS_ENABLED=false` is turned off, by contrast.

    AI review

  • Infoauthor says it is not ready for real use

    The README's note: "While it's not fully optimized for production environments, it combines robust components that work well together for proof-of-concept projects". The last commit is from 2026-02-26, 252 days before this check, with 61 issues open.

    AI review

Evidence is shown in its original language.

What Radar could not check about Local AI Packaged

  • How much disk and memory the twenty-odd containers plus the two downloaded models really take
  • Whether the whole stack still comes up today, since every image follows the moving `latest` tag and Supabase keeps adding required variables
  • Whether the example n8n workflows work after import, as each needs credentials created by hand
  • What the `start_services.py` script does, since its code was not read here
  • How the models perform without a graphics card, on the `cpu` profile

Similar projects

The RepoPicks app

Coming soon: install in 1 click with the RepoPicks app.

Before you download

See the release page on GitHub