27 projects for a two-node homelab: heavy indexing on the N100 NAS, always-on and local AI on the Pi 5

Asked (summary)

Find the best active open-source projects (committed 2025–2026) for a Docker-only two-node homelab — an Intel N100/8GB UGREEN NAS with fast NVMe plus CMR disks, and a Raspberry Pi 5/16GB with an SMR cold archive, linked by LAN and Tailscale. For each: name, GitHub URL, primary language, last commit, approximate RAM footprint, recommended node, and one sentence on fit — covering distributed storage/eBPF/Raft; couple workflows (watch parties, voice journals, trip canvases, photos); browser IDE/terminals/OCR; under-4GB ARM local AI and knowledge tools; and daily productivity.

The request names four areas but actually spans five — distributed storage/eBPF/Raft, couple workflows, nomad browser tooling, Pi knowledge and local AI, and daily productivity. From 60 deduplicated candidate rows (plus 64 supporting evidence rows) gathered on 2026-09-11 across dozens of independent hosts, this shortlist keeps the 27 canonical, active projects; mirrors, archived tmux forks, wrong-namesake repos and malformed links were cut. RAM figures are deployment envelopes for single-user Docker use, not benchmarks; 22 of 27 are conservative estimates (marked “est.”) since sources rarely publish runtime RAM.

Deployment matrix: where each project runs and what it costs in RAM

UGREEN NAS (N100, NVMe) Raspberry Pi 5 (ARM) either / both nodes RAM band shade: darker = heavier (<100 MB → multi-GB) ARM: filled = runs fine, half = caution, empty = keep on x86 NAS Hover a row for the fit sentence and the operational caveat

bpftrace and Inspektor Gadget stay package-free on the host but still need a privileged container, kernel BTF/debug symbols and capabilities — UGOS Pro kernel support is unverified, so the Pi is the safer first eBPF lab even though NAS I/O is the more valuable target. gittrend.io

OpenRaft and Dragonboat are consensus libraries, not turnkey apps — plan a small containerized KV/control-plane harness (30–200 MB est.) to exercise them across the two nodes. codehub.bot

No active Plex-native SyncLounge-quality watch-party project was strongly evidenced; Emby Watch Party (last pushed 2026-02-27) is the active synchronized-viewing fallback, as a media-stack alternative rather than a Plex add-on. awesome.ecosyste.ms

llama.cpp on the Pi 5 with a sub-3B Q4 model such as SmolLM2 1.7B, 2K–4K context, ~3 threads and a Docker CPU quota lands around 1.5–2.5 GB practical RAM — active cooling is still recommended to avoid throttling; Open WebUI sits on the NAS and calls the Pi endpoint over LAN/Tailscale. promptquorum.com

All 27 shortlisted projects — click a header to sort

Area	Project	Language	Last commit	Approx. RAM	Node	Why it fits	Source

Method: 27 projects hand-filtered from 60 deduplicated candidate rows plus 64 evidence rows on browser IDEs, couple tools, Pi 5 local AI and daily productivity, researched 2026-09-11 from independent web sources (each row links its own). “Last commit” is the most recent commit/release date evidenced, not necessarily HEAD today. RAM is an approximate single-user Docker deployment envelope; rows marked est. are conservative estimates from the implementation stack, never container image size. Cut for space: mirrors, archived tmux forks, unrelated namesakes, wiki-edit URLs, and weaker duplicates (Mermaid, garage, MemOS, SameRow and other secondary candidates remain in the source data).

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.