\n
\nA local-first, self-hosted assistant for your home. One stateful “brain” runs a local LLM on
\nhardware you own, and every window into it \u2014 your phone, a terminal, the kitchen mic, Home
\nAssistant \u2014 talks to that same brain. Nothing runs in the cloud, nothing is exposed to the
\ninternet, and your data never leaves the house.
\nThe idea it’s built on. Most “AI for the home” points the model at the things it’s worst
\nat: remembering a schedule, watching a threshold, firing a reminder at the right minute. Hestia
\ndoes the opposite. Anything deterministic \u2014 a chore is due, the soil is dry, the trash goes out
\nTuesday \u2014 is handed to something dumb and reliable: a timer, a record, a row in a database. The
\nLLM is left to do the one thing it’s genuinely good at, which is judgment and conversation. The
\ngoal was never a smarter brain. It’s a more reliable one. (ARCHITECTURE.md is
\nthe long version; MEMORY-DESIGN.md covers the memory plan.)
\nWhat it actually is.<\/p>\n
A brain (brain\/) \u2014 an OpenAI-compatible endpoint (POST \/v1\/chat\/completions) wrapping a
\nlocal LLM (Ollama, qwen3:14b) with an agent loop. Every client speaks one dialect.
\nEight scoped tools \u2014 home (control Home Assistant), media (Plex + *arr), memory,
\nrecords, reminder, search, status, weather. There is deliberately no shell tool:
\nthe brain can act in your house but cannot run arbitrary commands.
\nMemory that grows \u2014 markdown soft-facts plus a SQLite record of the things in your life
\n(pets, garden, wildlife, chores), and a background note-taker that proposes durable facts for
\nyou to approve rather than writing them silently.
\nA media appliance \u2014 Plex + the *arr stack + Bazarr subtitles + qBittorrent behind a
\nfail-closed VPN kill-switch.
\nVoice \u2014 talk to it through Home Assistant’s Assist pipeline or the browser.<\/p>\n
What it isn’t. A cloud service, a wrapper around someone else’s API, or anything you should put
\non the public internet. It runs rootless on your own box and never phones home.<\/p>\n
\u26a0\ufe0f Read SECURITY.md before running it. The brain has no built-in
\nauthentication and can control your devices, so it must stay on a private network (Tailscale or
\nLAN). That’s a deliberate trade-off, not an oversight \u2014 the doc explains the trust model.<\/p>\n
Hestia is part of the Forager \/ Homesteader Labs constellation, alongside forager_ml,
\nforager-field-station, and the Homesteader Labs site.<\/p>\n
Phase 0 \u2014 Reach + brain \u2705 \u2014 talk to your home model from your phone (details below).
\nPhase 1 \u2014 Media appliance \u2705 \u2014 Plex + qBittorrent + gluetun VPN kill-switch (verified) + the
\n*arr automation layer (Prowlarr\/Sonarr\/Radarr + FlareSolverr + Bazarr subtitles). Full loop:
\nsearch \u2192 download (via VPN) \u2192 hardlink \u2192 Plex.
\nPhase 2 \u2014 House (Home Assistant) \u2705 \u2014 HA running; lights and devices reachable via the home tool.
\nPhase 3 \u2014 Voice \u2705 \u2014 speak to Hestia through HA’s Assist pipeline and a browser voice loop.
\nPhase 4 \u2014 The seam (memory + tools) \u2705 core in place, still growing \u2014 the brain is a
\ntool-calling agent with the eight tools above plus deterministic skill injection, and HA’s
\nconversation agent points at Hestia, so Assist and voice route through the brain (which can
\ncontrol HA back). It also gets smarter over time via the note-taker (see Memory & learning).
\nNext: vision (Eyes).<\/p>\n
Phase 0 \u2014 Reach + brain<\/p>\n
Win: talk to your home model from your phone.<\/p>\n
The brain (brain\/) is a thin OpenAI-compatible proxy onto Ollama. Every client \u2014
\nterminal, phone, kitchen mic \u2014 speaks one dialect (POST \/v1\/chat\/completions).
\nIn Phase 0 it forces the chosen model, injects Hestia’s system prompt (persona +
\nthe hardened safety rules from the benchmark A\/B), and streams the reply back.
\nMemory and tools land in Phase 4 behind this same URL.<\/p>\n
Service
\nWhat
\nBind
\nGPU<\/p>\n
hestia-ollama
\nOllama inference engine
\n127.0.0.1:11434 (localhost only)
\nRTX 5080 only<\/p>\n
hestia-brain
\nHestia \/v1 proxy
\n0.0.0.0:8730 (reachable over Tailscale)
\n\u2014<\/p>\n
Both are user systemd services (no root), defined in deploy\/systemd\/ and
\ninstalled into ~\/.config\/systemd\/user\/. Linger is enabled, so they survive
\nlogout\/reboot. Ollama is pinned to the 5080 (CUDA_VISIBLE_DEVICES), leaving the
\n4060 Ti free for Phase 3 (Whisper\/Piper) per the benchmark verdict.
\nModel: qwen3:14b (resident, thinking off) \u2014 the current pick after the model eval
\n(brain\/eval_models.py; qwen2.5:14b kept on disk as a fallback). See MODEL_EVAL.md.<\/p>\n
Day to day, use deploy\/hestiactl (symlinked into ~\/.local\/bin) \u2014 one command
\nfor the whole estate, run from the GPU box:
\nhestiactl status # brain health + local units + every container on hl-relay
\nhestiactl health # raw \/health JSON
\nhestiactl up|down|restart X # X: brain ollama | arr services | plex qbit ha adguard … | all
\nhestiactl logs X (-f) # journalctl (local) or docker logs (remote)
\nhestiactl vpn # verify the qBittorrent kill-switch
\nall covers only the Hestia-managed pieces (local units + arr stack); core
\ncontainers (AdGuard = house DNS, gluetun, HA) are controlled one at a time and
\nask for confirmation before stopping.
\nThe underlying commands, for when you need them directly:
\n# status \/ logs
\nsystemctl –user status hestia-ollama hestia-brain
\njournalctl –user -u hestia-brain -f<\/p>\n
# restart after editing brain code or a service file
\nsystemctl –user daemon-reload # only if you edited a .service
\nsystemctl –user restart hestia-brain<\/p>\n
# health (Ollama up + model present?) \u2014 brain binds the Tailscale IP, not localhost
\ncurl -s 127.0.0.1:8730\/health | jq<\/p>\n
# talk to it
\ncurl -s 127.0.0.1:8730\/v1\/chat\/completions -H ‘content-type: application\/json’ \\
\n -d ‘{“messages”:({“role”:”user”,”content”:”hello Hestia”})}’ | jq -r .choices(0).message.content
\nIf you edit a deploy\/systemd\/*.service file, re-copy it into
\n~\/.config\/systemd\/user\/ before daemon-reload.
\nReach it from the phone (Tailscale)
\nTailscale is the one piece that needs root, so it isn’t auto-installed. On the GPU
\nbox:
\ncurl -fsSL https:\/\/tailscale.com\/install.sh | sh
\nsudo tailscale up
\nThen on the phone: install the Tailscale app, sign in to the same tailnet. The
\nbrain is then reachable at http:\/\/:8730\/v1 from any app
\nthat speaks OpenAI (set that as the base URL; any API key string works \u2014 Ollama
\nignores it). Nothing is exposed to the public internet.<\/p>\n
brain\/
\n hestia.py # the agent loop: \/v1\/chat\/completions + \/health, tools, memory, note-taker hook
\n config.py # single source of paths + secret loading; makes the brain relocatable
\n prompt.py # SYSTEM_PROMPT \u2014 persona + hardened safety rules
\n records_store.py \/ memory_store.py # SQLite entities+events \/ markdown soft facts
\n note_taker.py # background “gets smarter over time” extractor
\n review_notes.py # CLI to review + promote the note-taker’s proposals
\n tools\/ # home, media, memory, records, reminder, search, status, weather (+ skill router)
\n tests\/ # pytest: stores, dispatch, note-taker (run: uv run –project brain pytest)
\n pyproject.toml # deps + dev (pytest) + pytest config (uv-managed, isolated venv)<\/p>\n
Relocatable. Every path derives from config.py’s own location, so moving or
\nrestoring the repo to a new path needs no edits; HESTIA_ROOT overrides if needed. All
\nservice URLs, tokens, and thresholds stay env-overridable next to the tools that use them.<\/p>\n
Phase 1 \u2014 Media appliance (Dell Micro = hl-relay)<\/p>\n
Win: the media stack runs, independent of the brain.<\/p>\n
Most of this already existed on the Micro before Hestia: Plex (hl-plex),
\nqBittorrent behind gluetun (Surfshark, OpenVPN, NL) with a fail-closed VPN
\nkill-switch, plus AdGuard, MQTT, and Home Assistant. The kill-switch is verified:
\nqBittorrent’s traffic egresses via the VPN datacenter IP, not the host’s. Don’t
\ndocker compose up the existing \/opt\/home\/compose.yml blindly \u2014 its volume paths
\nare literal \/path\/to\/… host dirs that the running containers depend on.
\nHestia added the missing automation layer as a separate, isolated stack
\n(deploy\/media\/compose.yml, deployed to \/opt\/home\/arr\/): Prowlarr (:9696,
\nindexer manager), Sonarr (:8989, TV), Radarr (:7878, movies). All reachable
\nover Tailscale.
\nAlso added FlareSolverr (:8191) so Prowlarr can reach Cloudflare-protected
\nindexers, wired as a Prowlarr indexer-proxy (tag flaresolverr).
\nWired via API: root folders point at the existing Plex library
\n(\/data\/TV Shows, \/data\/Movies); a remote-path mapping (\/downloads \u2192
\n\/data\/downloads) lets Sonarr\/Radarr hardlink from qBittorrent’s downloads into
\nthe library (instant, no copy \u2014 both are one filesystem under \/mnt\/media); Prowlarr
\nis connected to Sonarr + Radarr (fullSync). Five reputable public indexers added
\n(The Pirate Bay, Knaben, LimeTorrents, plus 1337x + EZTV via FlareSolverr) and synced
\ndown to the apps. YTS deliberately excluded (history of feeding user data to copyright trolls).
\nqBittorrent is wired as the download client in both Sonarr (category tv-sonarr)
\nand Radarr (radarr), tested OK. The full loop works: search \u2192 download through the
\nVPN \u2192 hardlink into the Plex library. Both apps report no health warnings.
\n\u26a0\ufe0f Media currently lives on the Micro’s 98 GB root disk (~66 GB free). Fine to start;
\nplan a dedicated disk or NAS before the library grows.<\/p>\n
cd \/opt\/home\/arr
\ndocker compose ps
\ndocker compose pull && docker compose up -d # update *arr<\/p>\n
Phase 4 \u2014 the seam: HA conversation agent \u2192 Hestia
\ndeploy\/ha\/custom_components\/hestia\/ is a thin custom HA integration: it registers a
\nconversation agent (conversation.hestia) that forwards each utterance to Hestia’s
\n\/v1 and speaks the reply. Hestia owns the loop (memory + tools, incl. controlling
\nHA back); HA is just input + a tool. This is the architecture’s keystone made real.
\nWiring on hl-relay (not in this repo \u2014 lives in HA’s config):<\/p>\n
Integration files installed to \/opt\/home\/ha_config\/custom_components\/hestia\/.
\nA config entry points it at http:\/\/127.0.0.1:8730\/v1\/chat\/completions (Hestia
\nover Tailscale; the HA container can reach it).
\nThe preferred Assist pipeline’s conversation_engine is set to conversation.hestia,
\nso the Assist chat and voice satellites route through the brain.<\/p>\n
Verified: via HA’s conversation API, “turn on the TV light” drove the real light and
\n“what coffee should I buy?” recalled a memory \u2014 HA \u2192 Hestia \u2192 HA round trip.
\nMemory & learning \u2014 it gets smarter over time
\nTwo stores back the brain: memory_store (markdown soft facts\/preferences, git-auditable)
\nand records_store (SQLite entities + a uniform event log: pets\/lineage, wildlife, chores,
\nservice reminders, the garden). Both are injected into the system prompt per request, scoped
\nto what the request implies.
\nThe brain also learns passively. After each exchange \u2014 once the answer is already on the
\nwire \u2014 a background note-taker (note_taker.py) reads the turn and proposes durable
\nfacts it heard (“trash pickup is Tuesday mornings”). True to propose, don’t dispose, those
\nland in a review inbox (memory\/inbox\/), not straight into live memory:
\nuv run –project brain python brain\/review_notes.py list
\nuv run –project brain python brain\/review_notes.py promote id> | –all
\nuv run –project brain python brain\/review_notes.py discard id> | –all
\nIt reuses the resident model by default and never blocks or breaks a request. Tuning knobs:
\nHESTIA_NOTETAKER=0 disables it; HESTIA_NOTETAKER_AUTOWRITE=1 skips the review queue and
\nwrites durable memories directly; HESTIA_NOTETAKER_MODEL points it at a cheaper model (e.g.
\na second Ollama on the free 4060 Ti) to take the load off the brain.<\/p>\n
Hestia is licensed under the GNU Affero General Public License v3.0 \u2014 see LICENSE. A local-first, self-hosted assistant for your home. One stateful “brain” runs a local LLM on hardware you own, and every window into it \u2014 your phone, a terminal, the kitchen mic, Home Assistant \u2014 talks to that same brain. Nothing runs in the cloud, nothing is exposed to the internet, and your data never leaves […]<\/p>\n","protected":false},"author":1,"featured_media":6260,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[676],"tags":[],"class_list":["post-6259","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tech-ai"],"_links":{"self":[{"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=\/wp\/v2\/posts\/6259","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=6259"}],"version-history":[{"count":0,"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=\/wp\/v2\/posts\/6259\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=\/wp\/v2\/media\/6260"}],"wp:attachment":[{"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6259"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6259"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/daiilynews.cu.ma\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6259"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\nThe AGPL is deliberate: Hestia is built to be self-hosted, so the copyleft keeps it open even for
\nanyone who runs a modified version as a network service, while imposing nothing on you for running
\nit at home.
\nBefore running it, read SECURITY.md: the brain has no built-in authentication
\nand can control your Home Assistant devices, so it must stay on a private network (Tailscale\/LAN)
\nand must never be exposed to the public internet. It deliberately has no shell tool.
\n\u00a9 2026 TheFullNacho and contributors.
\n
\n
Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"