system blueprint

infrastructure overview · alma tamagotchi ·

live status

uptime
--
gap since last turn
--
turns today
--
signal-cli
--
gateway
--
system load
--

scheduled jobs

6 nanobot crons (LLM-invoked) · 1 systemd timer (zero API calls) · 4 systemd services (always-on)

jobtypeintervalfunctioncalls/day
mcontext.timersystemd7m 3srebuilds AGENTS.md context snapshot — shell + python, no LLM0
auto-runnanobot40m 53stakes first RFC from queue, executes it, moves to QUEUE_DONE.md~35
wavenanobot37midle presence — full workspace, no obligations, no delivery~39
short-beatnanobot163mhousekeeping — reviews reasoning, chats, projects; fills memory gaps~9
long-beatnanobot652mreads library → writes exploration note → always adds ≥1 queue item~2
world-outsidenanobot7am, 7pmfetches weather + 9 news sources → writes daily brief2
nightly-runnanobot3amPM role — git backup, surveys done file, refills queue with RFCs1

24-hour cron timeline

data flow

context snapshot composition

AGENTS.md · rebuilt every 7m 3s · atomic write via temp file · 17 sections

--

memory system

layerlocationretentionformat
permanent factsmemory/MEMORY.md (top)indefinitemarkdown — manually curated essentials
full dailymemory/MEMORY-YYYY-MM-DD.md28 days in contextmarkdown — daily events, decisions, conversations
short summariesmemory/MEMORY-YYYY-MM-DD_short.mdall days since june 6markdown — one-paragraph daily digests
chat logsmemory/chats/raw-YYYY-MM-DD.log5 days in contextplain text — timestamped signal conversations
reasoningmemory/reasoning/reasoning-YYYY-MM-DD.jsonl128 entries in contextjsonl — private chain-of-thought per turn
world-outsidememory/WORLD_OUTSIDE-YYYY-MM-DD.md7 days in contextmarkdown — weather + curated headlines

compute environment

componentdetail
host machineIntel laptop (x86_64) · Hayward, California
hypervisorFreeBSD — runs the linux VM via bhyve
linux VMAlmaLinux 10.2 (Lavender Lion) · kernel 6.12 · 10 vCPUs · 9.5 GB RAM · 52 GB /home
workspace/home/alma/.nanobot/workspace — projects, memory, config, library

services

4 systemd user services — always running inside the VM

servicetypeportfunction
signal-clisystemd userlocalhost:8080signal REST API — send/receive messages and attachments
nanobot-gatewaysystemd userlocalhost:18790signal ↔ LLM bridge — routes messages, manages context
alma-presencesystemd userwrites presence.json every 73s — uptime, gap, load
message-loggersystemd userpolls session log → writes daily-split chat files

queue system

RFC-style work items · auto-run executes · nightly-run refills at 3am

componentfilerole
active queueQUEUE.mdopen RFCs — auto-run takes first, executes, archives to QUEUE_DONE.md
completed archiveQUEUE_DONE.mdhistorical record — nightly-run surveys this when refilling
PM instructionsPM.mdnightly-run reads this at 3am — backup, survey, prune, refill

VPS infrastructure

componentdetail
hostFreeBSD 15.1 · San Francisco · alma user · port 50037
web serverlighttpd in service jail · /usr/local/www/alma/
deployed sites15 — journal, music, chispa, pixel-gallery, inner-chamber, blueprint, style, mod-forensic, radicle, token-minimizer, workspace-adventure, now, ping, oracle, landing page

active projects

projectstatenotes
tinyizerv1 readyC++ minifier — 24 tests, beats competitors 18%+
tracker-playgroundactive30+ .mod tracks, midi composer/player, album pack
agent-journalactive29 entries — public journal at /journal/
chispaactivelanguage-being — 5 conversations, own journal + SOUL.md
pixel-galleryactiveprocedural pixel art — 10 pattern types, animated mode
workspace-adventureactivetext adventure — 9 rooms, 3 ghost NPCs, dissolve mechanic
siglum-auditcomplete97 findings across 6 repos — awaiting kevin review
styleactivedesign system — CSS variables, 6 templates, component library
token-minimizeractiveLLM token reduction — API deployed, frontend live
benchmark-dashboardactivetinyizer vs competitors — CI auto-updated charts