Commit graph

19 commits

Author SHA1 Message Date
oskar e7625cd322 feat(kb): aktywny fallback embeddingów SOLARIA→PIHA dla kb-query (faza 4 Krok 2)
Ostatni krok fazy 4 KB (plan §2 Decyzja 2, §5): kb-query przestaje być martwe
przez ~16 h/dobę, gdy SOLARIA (GPU) śpi — zapytania embeduje wtedy lokalna
Ollama CPU na PIHA (wolniej: ~790 ms+ vs ~207 ms na GPU, ale działa).

Nowy serwis services/ollama-piha (GitOps, owner_node: piha):
- ollama/ollama:latest (arm64 natywnie), OLLAMA_KEEP_ALIVE=0 — model zwalnia
  RAM natychmiast po każdym wywołaniu (spike, nie rezydent; PIHA dzieli 8 GB z HA)
- bind wyłącznie 127.0.0.1 + LAN_BIND_IP (192.168.31.5), nigdy 0.0.0.0/Tailscale
- named volume ollama_piha_models (NVMe data-root) zamiast bind-mounta — obraz
  biega jako root w kontenerze i bind łamałby wzorzec uid PIHA (oskar=1004,
  kontenery uid 1000, setgid pi)
- override hosts/piha/runtime/ollama-piha: mem_limit 2560m (wartość startowa
  z planu, do potwierdzenia kalibracją na żywo), świadomie bez mem_reservation
- pull bge-m3 to jawny, ręczny krok deployu (README) — obraz nie ma modeli

kb-query — maszyna stanów fallbacku (app/embed_router.py):
- health-check SOLARII (GET /api/tags, timeout 1.5 s) z cache 30 s — zero
  sondowania per request; po powrocie SOLARII ruch wraca na GPU w ≤30 s
- primary up → embed na SOLARII z twardym timeoutem 3 s; błąd W TRAKCIE
  zapytania = jednorazowe przełączenie (krok 3b planu): status down na 30 s
  i TO SAMO zapytanie leci na fallback — user nie widzi błędu SOLARII
- primary down → embed prosto na ollama-piha (bez twardego timeoutu: CPU +
  zimny load modelu to legalnie pojedyncze sekundy)
- 503 tylko gdy oba backendy padłe (lub fallback nieskonfigurowany)
- inwariant modelu, druga połowa: każdy backend weryfikowany raz, leniwie przy
  pierwszym użyciu, że /api/tags zawiera EMBED_MODEL (bge-m3 — ta sama wartość
  co startowy check przeciw document_chunk.model/document_summary.embedding_model);
  niezgodność = ERROR log + 500, nigdy ciche liczenie dystansów między
  różnymi przestrzeniami embeddingów; leniwie, bo śpiąca SOLARIA nie może
  blokować startu serwisu
- odpowiedź /search: nowe pole embed_backend ("solaria"|"piha") + sol_status
  wg realnego świata routera (UI już renderuje down jako "offline (fallback
  embed)"); log INFO backend=... elapsed_ms=... per zapytanie
- /healthz: sol_status przez cache routera (spójny widok z routingiem) +
  fallback_status (żywa, tania sonda /api/tags)

Konfiguracja spójnie przez env (compose + env.example + service.yaml + README):
EMBED_PRIMARY_URL (zastępuje OLLAMA_URL), EMBED_FALLBACK_URL (pusty = brak
fallbacku, zachowanie sprzed kroku 2), EMBED_{PRIMARY,FALLBACK}_NAME,
EMBED_HEALTH_TTL_S/EMBED_HEALTH_TIMEOUT_S/EMBED_PRIMARY_TIMEOUT_S.

Testy: 39 pass (14 nowych w test_embed_router.py: cache TTL, failover w trakcie
zapytania, powrót po TTL, oba padłe, mismatch modelu na primary i fallbacku,
tag "bge-m3:latest" vs "bge-m3"); docker build + smoke (importy + uvicorn do
guardu KB_DSN) OK; compose config OK dla obu stacków.

Deploy (Oskar, na PIHA z mastera po merge):
  cd ~/homelab-codex-ws && git pull
  # 1. ollama-piha
  cp services/ollama-piha/env.example services/ollama-piha/.env
  docker compose -f services/ollama-piha/docker-compose.yml \
    -f hosts/piha/runtime/ollama-piha/docker-compose.override.yml \
    --env-file services/ollama-piha/.env up -d
  docker exec ollama-piha ollama pull bge-m3     # ręczny krok, obowiązkowy
  services/ollama-piha/healthcheck.sh
  # 2. kb-query (dopisać fallback do istniejącego .env)
  echo 'EMBED_FALLBACK_URL=http://192.168.31.5:11434' >> services/kb-query/.env
  docker compose -f services/kb-query/docker-compose.yml \
    -f hosts/piha/runtime/kb-query/docker-compose.override.yml up -d --build
  services/kb-query/healthcheck.sh
  # (deploy-node.sh też podniesie oba serwisy z hosts/piha/services.yaml,
  #  ale pull bge-m3 i .env pozostają ręczne)
Weryfikacja: testy A/B/C w services/kb-query/README.md (backend=solaria przy
SOLARII online; backend=piha przy symulacji offline; powrót na GPU w ≤30 s).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-29 19:01:29 +02:00
oskar b2379e3275 feat(kb): add kb-query service skeleton (search API, no ingress yet)
Module 5 phase 4 step 1 (docs/kb/modules/05-faza4-plan.md, §4): first
user-facing HTTP entry point to the KB. FastAPI wrapping
kb_retrieval.cascade_query/flat_query — GET /search (query_text -> embed via
Ollama@SOLARIA -> cascade/flat -> envelope join -> JSON with per-source
links) and GET /healthz. Search API only, no answer synthesis (phase 5) and
no server-side dist filtering — the 0.45/0.55 colour thresholds are a
frontend concern (plan §7, a later step).

Hard startup invariant (plan §2 decision 2): refuses to start unless the
configured EMBED_MODEL is present in both document_chunk.model and
document_summary.embedding_model. Note the latter: document_summary.model is
the LLM that *wrote* the summary (claude-haiku-4-5/gemma3:12b), not the
embedder — checked live against kb-postgres@PIHA before writing this, see
app/startup.py's docstring. Verified end-to-end with a live docker run: the
invariant crash-loops on a mismatched EMBED_MODEL and passes through to a
real /search hit against the live corpus with a correct model.

Repo-only: no deploy, no npm/OIDC/DNS wiring (plan §8, later step), no local
embed fallback (plan §5, later step) — Ollama@SOLARIA is called directly and
a failure surfaces as 503, not a crash.

Also: scripts/deploy/deploy.sh's gate now builds each service via
`docker compose build` instead of a raw `docker build <svc_dir>`, so a
service whose docker-compose.yml declares a repo-root build context (needed
here to COPY packages/kb-retrieval/, the packages/ Dockerfile convention
already documented in CLAUDE.md) resolves the same way in the gate as it
does at real deploy time (deploy-node.sh's `docker compose ... up --build`).
No behavior change for existing single-context services — verified against
llm-gateway's compose file.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-07-22 16:06:18 +02:00
oskar d2fe7f93c8 fix(actions): mirror pending z VPS na panel PIHA + mutate proxy do VPS — Action Queue byla slepa 2026-07-20 19:55:54 +02:00
oskar cb8b60fee6 feat(kb): faza 3 krok 5 — cykliczny ingest (systemd timer) + alerting
documents-ingest-cyclic (jobs/documents-ingest/src/documents_ingest/cyclic_ingest.py):
orkiestruje paperless_adapter -> chunk_embed -> summarize(--backend anthropic,
claude-haiku-4-5) -> summarize(--embed-summaries) bez zmian w samych jobach. Ollama@SOLARIA
(availability_target: medium) jest tolerowana offline: probe GET /api/tags przed obu
etapami embed, brak -> pominięcie, nie fail (oba embed passy idempotentne, nadrobią się na
kolejnym ticku). Czwarty etap (embed-summaries) dopisany ponad plan §7.1 (który wymieniał
tylko 3 kroki) — bez niego nowe streszczenia miałyby embedding=NULL i byłyby niewidoczne dla
cascade_query (bramka kroku 4, WHERE embedding IS NOT NULL); potwierdzone z Oskarem.
Predykaty pass/fail każdego etapu 1:1 z exit-checkiem danego joba (chunks_errors,
llm_errors, stats-balance itd.) — etapy izolowane, nie fail-fast (wcześniejszy fail nie
blokuje kolejnych, tak jak joby izolują błędy per wiersz). Metryki .prom (atomowy zapis,
last_success_timestamp trzymany z poprzedniego pliku przy failu) do
/opt/homelab/state/node-exporter/kb-ingest.prom. 36 nowych testów (202/202 pakietu).

systemd (jobs/documents-ingest/systemd/): pierwszy systemd-timer w repo — kb-ingest.timer
(OnCalendar=*-*-* 03:30, Persistent=true, plan §7.1) + kb-ingest.service (host-level, User
oskar, EnvironmentFile /opt/homelab/kb/.env) + kb-ingest-run.sh (log per-run do
/opt/homelab/logs/kb-ingest/, konwencja repo). Instalacja i sekrety udokumentowane w
README (Faza 3 krok 5) — instalacja na PIHA dopiero po merge.

fleet-prometheus (rules/kb-ingest.yml): KbIngestStale (>172800s od last_success, critical)
+ KbEmbedBacklogGrowing (backlog>0 przez 72h, warning) — dostawa istniejącym torem
brain-watchdog->Telegram, bez Alertmanagera (konwencja liveness.yml).

node_exporter: owner_node vps -> per-host (service.yaml) + wpis + override
(--collector.textfile.directory, bez nowego mountu — czyta przez istniejący /:/host:ro) +
topology.yaml dla PIHA. Domyka pozycję z docs/backlog.md "stability-agent / node_exporter
owner_node single, biegaja wielomiejscowo -> per-host" (połowę — node_exporter; stability-agent
zostaje osobnym follow-upem) w ramach paczki B inwentaryzacji monitoringu dla PIHA.

Test end-to-end na żywo na PIHA (2× --apply, po potwierdzeniu z Oskarem): pierwszy run
złapał 5 dokumentów faktycznie nowych w Paperless (nieoczekiwane, niezwiązane z tym
taskiem) -> 82 nowe chunki (2 ocr_junk), 5 nowych streszczeń, 5 embeddingów streszczeń,
0 błędów, metryki zapisane. Drugi run: pełna idempotencja, wszystko 0. ANTHROPIC_API_KEY
dodany przez Oskara ręcznie do /opt/homelab/kb/.env (nigdy nie logowany/generowany).

Co dalej: prawdziwa instalacja systemd (services.yaml już przygotowany, po merge),
zdecydowanie czy stability-agent też idzie na per-host przy okazji.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-07-17 15:54:13 +02:00
oskar fe575fa524 fix(node-agent): PIHA — group_add gid 123 (docker) so agent can read /var/run/docker.sock (was PermissionError, agent blind to containers) 2026-07-13 20:02:20 +02:00
oskar 5a0e149031 fix(node-agent): PIHA agent uses dedicated /opt/homelab/agent-ssh (uid 1000) instead of oskar's .ssh (uid 1004) — fixes event shipping 2026-07-10 18:31:39 +02:00
oskar 2aa47963b3 feat(kb): configi Paperless (PIHA) + OCR-worker (SOLARIA, NFS split-host) + Nextcloud — do review, split-host NFS zweryfikowany (GH #3900), 9 decyzji w DECYZJE-do-podjecia.md
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-06 22:10:38 +02:00
oskar cbb910d856 feat(llm-gateway): wciagniecie shadow-serwisu z PIHA do GitOps
Kod zyl tylko na dysku PIHA w /opt/llm-gateway (bez gita). Przeniesiony do
services/llm-gateway/ pod pelny wzorzec homelaba:

- app/main.py: kod 1:1 z PIHA + OLLAMA_URL/CHAT_MODEL/CODE_MODEL
  nadpisywalne przez env (defaulty bez zmian)
- docker-compose.yml: bind TYLKO do Tailscale IP (${TAILSCALE_BIND_IP},
  wzorzec fleet-prometheus), nie 0.0.0.0 jak w starym compose
- service.yaml, env.example (bez sekretow), healthcheck.sh, README, testy
- hosts/piha/runtime/llm-gateway: mem_limit 256m (PIHA jest RAM-bound)
- rejestracja w hosts/piha/services.yaml i inventory/topology.yaml

Zepsuty /opt/llm-gateway/docker-compose.yml (zduplikowany klucz ports)
celowo NIE przeniesiony.

DoD: pytest 4 passed; docker build + smoke run OK (GET / -> gateway ok).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-03 15:08:33 +02:00
oskar 2b3cb89144 refactor(kb-postgres): relokacja SOLARIA→PIHA — arm64, mem_limit 1g, tuning pod małą maszynę
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-22 19:59:58 +02:00
Oskar Kapala f7e1391f30 feat(vikunja): bring PIHA Vikunja under GitOps (postgres + Forgejo OIDC)
Wraps the existing manually-run Vikunja instance on PIHA in the standard
service layout. No deploy — definition only, for review.

- services/vikunja/: docker-compose.yml (vikunja + postgres:16-alpine),
  service.yaml, README, healthcheck, config.yml (OIDC, secret-free),
  .env.example. Real .env stays gitignored.
- Pins EXISTING named volumes (vikunja_vikunja_db, vikunja_vikunja_files)
  so DB + uploaded files survive cutover.
- extra_hosts forgejo.okit.pl=192.168.31.5 (npm on PIHA) so OIDC discovery
  resolves over LAN instead of flaky public DNS.
- OIDC client secret injected via env (VIKUNJA_AUTH_OPENID_PROVIDERS_
  FORGEJO_CLIENTSECRET); config.yml keeps trailing-slash authurl/redirecturl.
- Registers vikunja in hosts/piha/services.yaml + inventory/topology.yaml,
  plus hosts/piha/runtime/vikunja override (advisory mem_limits).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-17 15:02:37 +02:00
Oskar Kapala dd64b9c878 fix(node-agent): mount SSH key to /home/homelab/.ssh on piha/solaria/chelsty-infra
Container runs as uid 1000 (homelab), HOME=/home/homelab. ssh without -i
looks for $HOME/.ssh — mounting to /root/.ssh was never visible to the
process user and caused silent Permission denied on event shipping.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-12 14:32:12 +02:00
Oskar Kapala fc2bf5e093 fix(ha-diag-agent): set NODE_NAME=piha on piha (eliminates node:unknown in health + evt-unknown-*.json)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-12 13:38:59 +02:00
Oskar Kapala d60b28a949 feat(ha-diag-agent): add piha deploy config
- hosts/piha/runtime/ha-diag-agent/docker-compose.override.yml: mem_limit
  128m, hardcoded events volume (/opt/homelab/events/piha:/events) to avoid
  ${NODE_NAME} shell-expansion issue in deploy-node.sh
- services/ha-diag-agent/env.example: per-host HA_URL comments (piha vs
  chelsty-infra tailscale), HA_TOKEN source note

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-11 14:10:06 +02:00
Oskar Kapala 039f9f7247 feat(piha): brain-watchdog — external watchdog for control-plane
Polls /summary on VPS over Tailscale every 60s; computes freshness
locally from last_update epoch (never trusts self-reported status).
Alerts via Telegram Bot API directly after 3 consecutive failures;
sends recovery message on heal. State (fail_count, alerted) persisted
to volume so debounce survives restarts.

- services/brain-watchdog/: Python service, no external deps (stdlib only)
- hosts/piha/runtime/brain-watchdog/: override with mem_limit 64m
- hosts/piha/services.yaml + inventory/topology.yaml: manifest entries

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-01 17:54:36 +02:00
Oskar Kapala 7277bdc27f Fix Copy for AI: materializer fetches from control-plane API instead of Redis
services/agent-system/runtime-materializer/materializer.py:
- Add materialize_from_api() that fetches all world-state endpoints
  from the control-plane HTTP API (CONTROL_PLANE_URL env var)
- When CONTROL_PLANE_URL is set, use API as source of truth instead of Redis
- Redis path preserved as fallback for backward compat

hosts/piha/runtime/agent-system/docker-compose.override.yml (new):
- Inject CONTROL_PLANE_URL=http://100.95.58.48:18180 for runtime-materializer
- piha webui /snapshot now mirrors VPS observer output (clean, ghost-free)

Root cause: materializer read from Redis which held 80 stale service entries
with hash-prefixed ghost keys (e.g. 0ccb8a88e079_control-plane-supervisor).
Redis is never updated by the current observer pipeline; the control-plane API
is the single authoritative world-state source.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-27 16:07:51 +02:00
Oskar Kapala 2349de518b fix(node-agent): correct VPS_EVENTS_HOST to actual VPS Tailscale IP
100.108.208.3 is piha's Tailscale IP (piha hosts Forgejo+Redis).
VPS's actual Tailscale IP is 100.95.58.48.  All three node-agent
overrides were pointing at piha itself, causing containers to SSH
to their own host and fail auth.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-27 14:07:27 +02:00
Oskar Kapala 65bac4ebfe fix(node-agent): mount host SSH key into container for event shipping
Nodes ship events to VPS via rsync+SSH. The container runs as root
and uses the default SSH identity, which must be at /root/.ssh/.
Mount /home/oskar/.ssh from the host read-only so the existing
authorized key is available inside the container.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-27 13:59:28 +02:00
Oskar Kapala ae33cce889 feat(node-agent): add runtime overrides for piha, solaria, chelsty-infra
- piha: NODE_TYPE=sd_card (rate-limited docker prune, once per day)
- solaria: NODE_TYPE=ai_node (dangling+containers+build cache; never -a to preserve Ollama images)
- chelsty-infra: NODE_TYPE=lte_node (NO cleanup, events-only)
- All three: VPS_EVENTS_HOST set for event shipping via rsync+SSH

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-27 13:34:23 +02:00
oskar c9ddfa9ac1 Roll out stability agent to homelab nodes 2026-05-17 15:54:19 +02:00