- classifyAndGate: escalate read-only commands on lxc: targets that
touch /opt/, /etc/, /var/lib/ to config_mutation. The classifier
scores command text only, not the SSH transport layer — SSH-ing into
a container to read config is riskier than pct exec from the host.
- ontology: standalone-server monitoring override from inherited
[ping, resource, updates] to [http]. VPS-like machines may not be
SSH/ICMP-reachable from the scheduler; HTTP is the LCD liveness
signal. Entities with full SSH can override per-entity.
- get_health_summary: filter out state=destroyed entities (was noise
from 20+ destroyed test LXCs, deprecated services, etc.)
- create_entity: document the monitoring footgun in the tool description
(creating a type=check entity does NOT wire a check_def; the correct
path is update_entity_attributes with monitoring + url attributes)
Adds a new 'dns' semantic monitoring kind that probes whether a DNS name
resolves. Uses net.LookupNS (NS records) with fallback to net.LookupHost
(A/AAAA). Supports an explicit server config for split-horizon resolution.
Changes:
- seeds/ontology.yaml: dns-zone monitoring: none → [dns] (was deferred
since 2026-06 with a comment 'no dns checker exists yet')
- seeds/inventory.yaml: host:netbird-vps monitoring: [http] (was none;
VPS was invisible for 7 days during the 2026-07-29 outage)
- internal/checkdefaults/defaults.go: add KindDNS, buildKind case for 'dns'
that creates a check_def at 5-minute intervals
- internal/scheduler/scheduler.go: add checkDNS probe + wire in executeCheck
The DNS checker catches stale/unreachable zones (e.g. matrix.hubris.network
pointing to a dead VPS IP). The VPS HTTP check probes the public endpoint
every 60s, closing the 7-day monitoring gap.
get_relations now accepts an optional 'types' (comma-separated) parameter
to filter relationship types — filters out the noisy exec/targets edges
that previously drowned useful host/provides edges.
get_health_summary now accepts an optional 'health' (comma-separated)
parameter to return only entities in specific health states (e.g.
'health=down,stale') instead of the full 100+ entity list.
ping_service now:
- Falls back to e.attributes->>'public_host' when 'url' is not set
(covers LXCs that only have public_host in the graph)
- Performs a live HTTP HEAD probe against the resolved URL, returning
the actual status code instead of just the scheduler's stale health
state
Also: fixed matrix.hubris.network DNS record (was pointing to dead VPS),
pruned 6 dead graph edges, wired url attributes on 7 LXCs, added VPS
HTTP monitoring check, and resolved the 18k-occurrence unmonitored signal.
This session's audit is documented as
document:nomos/2026-08-05-dns-monitoring-improvements-for-strong-hosted-services.
- Add MCP tool — lightweight connectivity check returning server
identity, no DB hit (resolves agent connection-test friction)
- Tighten 6 tool descriptions (get_relations, get_health_summary,
query_metrics, get_trend, get_event_timeline, ping) to be searchable
in the first 8-12 words
- Document Hermes MCP client setup in ADR-0012 with token security caveat
- Move completed plan to plans/done/
Backend:
- Add isThinking flag to agentEvent for text before tool calls
- Separate thinking from response text in runChatTurn and continue.go
- Persist thinking in a dedicated field in message content
Frontend:
- Add thinking field to MessageContent, ChatMessage, ChatTextEvent types
- Create ThinkingBlock.svelte — collapsible block with brain icon
- SSE handler moves text_delta content to thinking on isThinking flag
- Render thinking block between tools and response in ChatThread
- Fix chat window scroll reset on focus change (stable windowKeys order)
- Remove redundant #key id wrapper in WindowLayer
- Enlarge sidebar rail (24→32 default, 40→60 max)
- Remove glyph from sidebar, square graph at top
- Replace AgentTrace/ToolCallCard/UnifiedTimeline with TurnTrace/ToolLine
The "Working — message will queue" strip lived inside the sized input Pane, so
appearing/disappearing ate the textarea's fixed height and clipped it, forcing
a resize. Move it into the message Pane alongside the connection/error banners
— those correctly consume transcript space (flex-1) rather than the input's
fixed height. The input Pane is now stable whether or not a background turn is
running. Styling/idiom unchanged.
VERSION: 0.17.2 -> 0.17.3
The background-working hint was a plain muted text line bolted above the
textarea — misaligned with the input column and off-idiom. Restyle it as a
terminal status strip: spinner + uppercase fg "Working" label + muted detail,
hairline primary-tinted border (matching .trace.running), square, aligned to
the textarea's max-w-3xl column. Reads as part of the working state now.
VERSION: 0.17.1 -> 0.17.2
Move ten completed plans from plans/ to plans/done/ and update the index:
- 2026-07-18 session-review-three-sessions, 2026-07-20 desktop-mascot,
2026-07-20 session-review-ten-sessions, 2026-07-21 chat-full-polish,
2026-07-29 health-check-reality-and-knowledge-graph,
2026-07-30 session-review-plan-drift, and the four 2026-08-03 chat plans
(changes-review, reliability-and-ux-audit, cyberspace-style-adoption,
working-visibility).
- Refresh two stale statuses: cyberspace-style-adoption ("Draft" -> shipped as
full replacement in v0.16.0/757ef2f) and health-check-reality ("ready for
implementation" -> shipped across the v0.14.x-0.16.x check commits).
- .gitignore: ignore local tooling artifacts (.playwright-mcp/, config-screen.png).
No code change. index.md Active/Done tables now match the filesystem (no orphans).
VERSION: 0.17.0 -> 0.17.1
Make background/long/desynced turns visible and queueable, fixing the four
symptoms that survived the v0.15.0 chat reliability pass.
F1 - status-driven working signal (workspace.ts taskWorking/currentWorking =
streaming OR status in {planning,executing}). Drives the chat trace, indicator,
and activity spinner so a turn with no live stream (background resume, a dropped
SSE, an idle-close mid long turn) still looks alive.
F2 - operator messages sent during an in-flight turn are now QUEUED and
auto-run when the gate frees, replacing the "still finishing a previous step...
send it again" rejection. Per-session in-memory FIFO (messagequeue.go, capped at
20) drained one-at-a-time under the turn gate; a `queued` SSE event drives a
"Queued" hint. drainQueued releases via a per-iteration deferred closure so a
runChatTurn panic can't deadlock the session's gate.
F3 - SSE keepalive (12s `:keepalive` comment) in handleChat so 20-40s
inter-iteration gaps no longer trip a proxy/browser idle close (the desync root
cause). All SSE writes serialized through one mutex.
F4 - generation-aware activity timeline (only the last propose_plan renders;
superseded ones collapse to one "Earlier plan revised" marker; step-attribution
follows only the current generation) + debounced plan refetch on lifecycle
events so a missed plan.proposed self-heals.
Verified against the last session (23da10db: 6m33s turn, operator "status"
deferred at 19:48:05). go test ./cmd/nomos/ green (new messagequeue tests);
web vitest 72/72 (new F4 generation tests); vite build clean.
VERSION: 0.16.0 -> 0.17.0
Rebrand light/dark themes to the cyberspace.online look: warm cream-on-black
palette (light/dark are exact inverses), self-hosted JetBrains Mono + VT323,
square corners, border-driven surfaces with no soft shadows. Adds a terminal
design-system CSS layer (DOS double-border modals with hatched corner, fg focus,
inversion-on-hover), a theme-aware <RasterImage> (Atkinson-dithered canvas with
img fallback), and unifies desktop icons, taskbar, window controls, pills and
links under one idiom. Pins window titlebars to a fixed height and switches chat
auto-scroll off scrollIntoView to avoid titlebar reflow.
VERSION 0.15.1 -> 0.16.0
The Send button's disabled={!input.trim()} threw "trim is not a function" when
input was initialized from a non-string initialDraft (a Svelte 5 prop-init edge
where a null/undefined draft reached $state). Coerce at init so the composer
state is always a string.
VERSION: 0.15.0 -> 0.15.1
The agent could run two turns for one session at once (a reconnect resumed
while the live turn was still going), and their interleaved tool calls
corrupted the activity panel, fabricated a confusing "parallel/nested"
sequence, and made tasks feel stuck/never-ending. Several UX gaps compounded it.
Turn serialization (F1):
- turnGate: at most one in-flight turn per session. Background resume paths
(continuation worker, idle sweep, answer-question, /resume, reconnect)
skip non-blocking when busy; the live chat path waits briefly then bails
cleanly instead of stacking a second turn.
- resumeSession returns whether it ran; continueSession marks an execution
"continued" only after a real run (review P0) so a busy-skip can't lose a
finished-execution result. Idle nudge bumps only after delivery (P1).
Connection state (F2/F3, web):
- humanize/bucket raw errors ("model connection dropped..."); one surface
per drop; a terminal task.status event clears stuck streaming/disconnected
state and dismisses the connection toast. Reconnect no longer spawns turns.
Streaming where you look (F4, web):
- live command output in the global activity timeline and in the inline
tool card (auto-opened, tail-pinned) -- not just the per-window rail.
Other (web): artifact/knowledge deep links (F5); step-first stable
"thinking" headline (F6); stable chat layout, no empty->content reflow (F7);
lazy event sync (P2.2); reconnect skips a terminal session (P2.1).
VERSION: 0.14.2 -> 0.15.0
The plan recorded false history after a re-plan and the activity panel
showed fabricated, churning timestamps. Two bugs compounding on one event
stream.
Plan drift (P0.1):
- proposePlan seq is now 1..N per generation; (session,generation,seq) is
the addressing key. The model's 1-based update_plan_step calls always map
to the CURRENT plan after a re-plan, instead of resurrecting a superseded
`replaced` row as done while the live work went unrecorded.
- updatePlanStep resolves against MAX(generation); a stale/out-of-range seq
returns errPlanStepNotFound (never touches a superseded generation).
- getPlanSteps returns only the current generation by default; ?all=true
keeps the audit/eval view (plan_generations assertion).
- completeTask auto-close scopes to the current gen, stamps started_at, and
emits one plan.step.finished per closed step so the panel converges
instead of freezing on "running" after completion (P1.1).
- propose_plan result enumerates step seqs; writeback detector matches
"write back"/"writeback"/"upsert_knowledge" so a natural-language final
step isn't doubled (P1.2).
- migration 029 renumbers existing seq per generation + unique index.
Activity panel (P0.2 / P1.1, web):
- computeActivityLog uses the real message created_at for tool calls; live
entries fall back to wall-clock frozen on first sight, killing the 3s
poll churn. Steps use real started_at.
- dropped plan-step events warn + count instead of a silent no-op.
Tests: TestProposePlan updated; + generation-relative-seq and auto-close
event-emission regression tests; + web activity purity/timestamp tests.
VERSION: 0.14.0 -> 0.14.1
Graph view: raise the node cap 500 -> 2000 and exclude execution/task audit
rows from the default whole-graph view so the cap is spent on actual topology
rather than ~380 cognition records that crowded out every host/lxc/service.
dns-zone monitoring [dns] -> none: no dns checker exists, so the declaration
only produced unresolvable `unmonitored` noise (requires ontology re-ingest;
coverageSweep now auto-clears the stale signals). Flip back to [dns] when a
checker lands.
Operator tooling: tools/deploy-checks.sh pushes check scripts into guests via
pct push (a pct-exec-routed check runs the script INSIDE the guest), wired
into the post-pull setup-checks hook so guests stay in sync on Proxmox hosts;
scripts/cleanup-orphan-checks.sh (dry-run by default) and
report-stray-test-lxcs.sh retire legacy cruft. VERSION 0.13.0 -> 0.14.0.
Plan: plans/2026-07-29-health-check-reality-and-knowledge-graph.md.
Problem: the mascot's right-click menu was non-interactive — RadialMenu's
root div rendered inside MascotLayer's pointer-events-none root (and the
new DockedLayer wrapper compounded it) without re-enabling pointer-events,
so clicks passed straight through. The desktop right-click menu was a
hand-rolled positioned div, inconsistent with the rest of the UI.
Change: both menus now use the shadcn-svelte context-menu primitive
(bits-ui, portaled to <body>).
- Mascot: MascotMenu.svelte renders the action tree recursively —
children become ContextMenu.Sub (native hover sub-menu navigation,
replacing the manual breadcrumb stack), leaves become ContextMenu.Item
with onSelect. MascotLayer wraps <Mascot> in a ContextMenu.Trigger;
visibility predicates read reactively off ctx.model so items
appear/disappear live. Removed the manual menuPos/openMenu/closeMenu
machinery. RadialMenu.svelte deleted.
- Desktop: the surface's bare-desktop hit area is now a
ContextMenu.Trigger layer (absolute inset-0, pointer-events-auto)
placed before the icons/windows in the DOM. The DOM-structure gate
(icons/windows are pointer-events-auto siblings that paint on top and
intercept their own right-clicks; bare desktop falls through to the
trigger) replaces the old fragile e.currentTarget === e.target check.
Left-click blur moved onto the trigger; Undo/Redo disabled state
snapshotted via onOpenChange (canUndo/canRedo are wmkit methods).
Risk: the blocker that made the mascot menu non-interactive in the first
place — Mascot.svelte's handleContextMenu called e.stopPropagation(),
which would have prevented a ContextMenu.Trigger wrapper from ever
seeing the right-click. Removed that handler; bits-ui now owns
right-click on the mascot, left-click drag/pet passes through. The
context-menu content portals to <body>, escaping the pointer-events-none
mascot and docked layers entirely — the structural fix, not just a
component swap.
Verification: vitest 38/38; svelte-check + tsc clean for changed files;
eslint clean (the shadcn-generated ui/context-menu/* files carry the
same baseline custom_element_props_identifier warnings as the rest of
the ui/ folder, not from this change); vite build green; runtime
confirmed — right-click mascot opens the action tree with hover
sub-menus, right-click bare desktop opens Cascade/Tile/Show/Reset/
Undo/Redo, right-click on an icon or window does not.
Problem: the frontend had an implicit OS+Apps metaphor (desktop, floating
windows, an app registry) but the contract was informal — the mascot was
hardcoded into the shell, all apps were statically imported into one
800KB bundle, and there was no install/uninstall path.
Change: three phases landed.
- Phase 1 (contract + docked kind): AppDef extended with docked/noIcon
and optional geometry; the mascot registered as a docked app via a
generic DockedLayer that replaces the hardcoded <MascotLayer />;
openAppWindow branches on docked → toggleDocked; persisted docked
visibility store (absent key = visible, no APPS import to avoid a
static cycle).
- Phase 2 (lazy loading): AppDef.component is now a dynamic-import
loader; LazyApp renders with a loading skeleton; Vite code-splits
each app (main bundle 800KB→485KB); the LazyMascot wrapper is gone
since the lazy loader breaks the import cycle directly.
- Phase 3 (installable apps, local bundles): AppManifest + catalog +
installApp/uninstallApp + localStorage persistence; reactive apps
store (built-in + installed) and derived appById; App Store page;
Notes demo app; icons.ts and WindowLayer's orphan-close react to
registration so installs appear without a reload.
- Structure: data-table casing unified to PascalCase; the mislabeled
DataTable.svelte.ts (pure types, not runes) renamed to types.ts;
LazyApp colocated with its desktop-shell consumers; app-store moved
under lib/ so the dependency direction is consistent.
Risk: the app registry is now a reactive store, not a static array, so
every consumer (Desktop, DockedLayer, Taskbar, icons, windows) reads
from derived stores. Two static-cycle traps are documented in
docs/mbse/components.md §9: docked.ts must not import APPS (it would
fire a TDZ at init via the apps.ts→pages→windows.ts→here path), and
apps.ts must not statically import the mascot (the lazy loader defers
its module graph). Remote bundle loading, the /api/v1/apps endpoint,
and permission enforcement are deliberately NOT in this commit — they
are security-critical and deferred to Phase 4 with an ADR.
Verification: vitest 38/38; svelte-check + tsc clean for changed files;
eslint clean; vite build green; runtime smoke confirmed (install
Notes → icon appears → open → uninstall → icon + window gone; survives
reload). docs/mbse/components.md Component 9 and the plan updated.
Plan: plans/2026-07-21-frontend-os-apps-architecture.md
Implements plans/2026-07-20-desktop-mascot.md. New code under
web/src/lib/mascot/ (types/sprites/render/state/behavior/actions/
stimuli + Mascot/MascotLayer/RadialMenu/NameDialog components) plus
CC0 sprite sheets at web/public/mascot/ (chicken + Onocentaur egg pack
+ reaction bubbles). MascotLayer is inserted into Desktop.svelte after
WindowLayer; <2-line integration.
Tamagotchi: egg -> chick -> adult lifecycle persisted to
localStorage['oikos-mascot'] (debounced 300ms). Egg hatches on first
naming (no timed incubation per implementation deviation). Chick/adult
wander, peck, sleep, blink autonomously via a weighted-random FSM; the
chicken walks above windows (ground line = highest window top edge
beneath its x, recomputed each tick from wmState; rides the ground when
the window beneath is dragged).
Interaction: draggable with flutter-fall physics on release mid-air;
plain click = pet (heart bubble + happy anim); right-click opens a
rounded-button radial menu (Interact/Care/Identity/Debug nested groups)
mirroring the desktop's own right-click menu styling; auto-flips above/
left near screen edges.
Awareness: stimulus bus subscribes to chat.ts streaming, activity.ts
activityLog (knowledge-entry diff), events.ts liveEvents (critical/
signal -> alarmed, execution -> happy), with priority+cooldown gating.
Egg-stage reactions are suppressed. Reaction bubbles are anti-aliased.
Sprite loop runs at ~60fps via setTimeout (not rAF) per GraphBackground
convention, dt clamped to 100ms; position via transform: translate3d
+ will-change: transform for compositor-friendly motion. Z-index
ordering: WindowLayer z-40 < MascotLayer z-[45] < desktop context menu
z-50 < RadialMenu/NameDialog z-[60].
Docs: plan + docs/mascot/README.md (MBSE subsystem model) updated to
Implemented with a deviations note covering hatch-on-naming, PNG-sheet
art, button-column radial menu, 60fps loop, egg-reaction suppression,
and window-walking ground model. VERSION bumped 0.7.13 -> 0.8.0.
Classifier now unwraps pct exec / qm guest exec / bash -c / sh -c / sudo
and env-var assignments before classification, so read-only inspection
wrapped in pct exec no longer escalates to config_mutation. curl GET
(default method, no -d/-F/-T/-o/>) is read-only. Eliminates the three
duplicate rclone sessions (a51e2086, 8acea2e3, cb8c8a4a) that bounced
off the classifier for the same goal.
New classify_command MCP tool: command-scoped preflight that returns the
exact risk class run would assign. Documented in SOUL.md with guidance
to pre-classify before run when the verdict is uncertain.
set_goal surfaces prior partial/failed sessions from the last 24h so the
agent picks up the thread instead of rediscovering it.
completeTask auto-closes in-flight plan steps (pending/running -> done
on success, skipped on partial/failure), so one-step plans no longer
need the per-step running->done dance right before completion.
Migration 021 adds blocker + closed_at to agent_sessions. completeTask
sets closed_at once and derives a structured blocker reason
(approval_timeout, user_abandoned, classifier_overreach, model_refusal,
tool_error, ...) from the last assistant message.
/sessions list now carries message_count, tool_call_count,
duration_seconds (server-side aggregates — no more N+1 transcript
fetches to audit a fleet). GET /sessions/{id} returns both metadata
and messages. New query params filter + paginate: outcome, status,
entity_id, blocker, since (RFC3339 or Go duration), cursor, limit.
Titles now prefer the goal when set; sessions without a goal fall back
to the first assistant text.
New GET /sessions/{id}/tool_calls flat view for audit scripts.
Plan: plans/2026-07-20-session-review-ten-sessions.md. VERSION 0.7.12 -> 0.7.13.
Session-review implementation for the three sessions audited in
plans/2026-07-18-session-review-three-sessions.md. v0.7.11 → v0.7.12.
P0.1 — retry cap + investigate-before-retry (cmd/nomos/retrycap.go,
agent.go): after 3 identical failing run calls in a single turn, refuse
to dispatch the call again and return a directive to investigate *why*
(ps/strace/lsof) or surface the blocker. Per-turn scope so a fresh turn
after the operator responds can retry once more. Session 1e9c7691's 20+
identical chown retries (knfsd held a kernel lock on the exported NFS
dir) is the direct motivation.
P0.2 + P1.8 + P2.10 — SOUL.md guidance: hung command is not a failed
command (investigate before retry); ask before proposing a multi-step
migration; multi-goal sessions summarize the arc not just the last goal.
P1.3 — two new runbook entities in seeds/knowledge.yaml:
- nfs-exported-dir-mutation-hang (the knfsd fchownat lock procedure:
killall → exportfs -u → mutate → exportfs -a → verify)
- netbird-mgmt-oidc-race-after-upgrade (docker restart netbird-mgmt
after ~30s for the traefik/authentik OIDC race)
P1.4 — setGoal emits task.superseded event when prior goal is overwritten
by a different goal (store.go, TestSetGoal_SupersededEvent). Session
55927f0a had two set_goal calls with the first silently abandoned.
P1.5 — inspect_path MCP tool: runs mount/df/ls/stat for one path across
up to 8 targets in one parallel call, replacing the 15+ run-call
fact-gathering fan-out sessions 1 and 2 each spent on cross-target path
tracing (tools.go, server.go: inspectPathAcrossTargets, inspectOneTarget).
P1.6 — vm: target support in run via qm guest exec (no more SSH-hop
with nested quoting). Extracted shared resolveProxmoxHostSlug for
LXC + VM, with hosts-relationship fallback when attributes.host is
absent (server.go, tools.go). Session 55927f0a's SSH-hop workarounds
for vm:zimaos are the direct motivation.
Deferred (documented in plan): P1.7 (approval window auto-extend on
timeout) and P2.9 (long-running command PENDING detection) — both
addressed at lower cost by the retry cap. Session 3's poll-after-timeout
pattern already works; the cap protects against the failure mode.
Rewrote .agents/domains/knowledge/schema.md and .agents/shared/llm-wiki.md
which described the deleted Python substrate (bin/homelab, oikos/cards/,
oikos/ledger.py, root inventory.yaml, knowledge/sources/, get_page/
search_docs MCP tools). Now reflect ADR 0003: Postgres DB is the single
source of truth for structured data and narrative knowledge; seeds/*.yaml
are bootstrap+DR manifests (content-hashed via seed_versions); archive/
knowledge/ is the frozen legacy wiki; MCP search_knowledge/get_entity_
knowledge replace get_page/search_docs.
Swept substrate refs in .agents/shared/{writing-style,page-templates}.md
and .agents/domains/operations/schema.md: bare inventory.yaml ->
seeds/inventory.yaml; knowledge/sources/ -> archive/knowledge/sources/
(historical); get_changelog/oikos/ledger.py -> DB audit trail / structured
document changelog field; HERMES -> Nomos.
Root inventory.yaml (618-line Python-era file superseded 2026-07-07 by
seeds/inventory.yaml) replaced with a deprecation stub pointing to the seed
and DB. Kept as a stub rather than deleted because AGENTS.md §1/§2 still
point clients at /opt/homelab-context/inventory.yaml; full on-client path
reconciliation deferred to R13.
Flagged export gap: oikos export regenerates seeds/{ontology,inventory,
policy}.yaml but NOT seeds/knowledge.yaml — API-added knowledge lives only
in the DB until hand-edited into the seed.
VERSION 0.7.7 -> 0.7.8. Plan R5 marked done.
The auto-complete fired when the agent hit the P5 approval gate — it
queued a config_mutation run for approval, the P5 gate blocked further
runs, the turn ended, and auto-complete closed the session as 'partial'.
The operator's approval would then land on a dead task.
Fix: hasPendingApprovals check — if the session has any executions in
pending_approval state, skip auto-complete. The session stays in
'executing' until the operator approves (or denies).
VERSION 0.7.5 → 0.7.6
The len(pending)>0 path still injected a brief note saying 'execution(s)
are now running' — the model saw this, thought work was being done for
it, and no-op'd (finish_reason=stop, content_len=0). Same confusion as
the len(pending)==0 case, just from the other branch.
Fix: both assent paths are now fully silent. No system note at all. The
model sees 'go ahead' in the replayed history and responds naturally.
Also removed chat_assent tool_use/tool_result emit events. These were
persisted in the transcript and confused the model on replay — it saw
its own 'tool calls' (chat_assent) and thought it had already acted.
VERSION 0.7.4 → 0.7.5
The auto-complete safety net required hadEntityWriteback to be true,
which meant sessions where the agent did the work but forgot to call
update_entity_attributes stayed stuck in 'executing' forever.
Relax: auto-complete fires if the agent did discovery (ran run),
regardless of writeback. If writeback happened → success; if not →
partial (honest: work was done but knowledge graph not updated).
VERSION 0.7.3 → 0.7.4
When the chat handler approves a pending execution via chat-assent, the
execution completes in ~2s. The continuation worker detects the completed
execution and calls resumeSession — while the chat handler is still
processing 'go ahead'. Two concurrent LLM calls for the same session cause
empty responses (finish_reason=stop) and race conditions.
Fix: mark the execution as continued immediately after chat-assent grants
it, so the continuation worker skips it. The chat handler will drive the
continuation itself (the model sees 'go ahead' and executes the plan).
VERSION 0.7.2 → 0.7.3
The assent pre-processing injected verbose system notes ('the operator
approved... they are now running... you MUST continue...') on top of the
replayed user message ('go ahead'). The model saw both, latched onto
'now running', concluded the work was being done for it, and no-op'd
(finish_reason=stop, content_len=0) — leaving the session stuck in
'executing'.
Root cause: the model already sees 'go ahead' in the replayed history
(the user message is saved to the DB before chat() is called, and
getRecentMessages replays it). The system note was redundant AND
confusing — it told the model work was 'running' when it wasn't.
Fix:
- len(pending)==0 (plan-proposal approval): open assent window silently.
No system note. The model sees 'go ahead' and responds naturally.
- len(pending)>0 (actual pending executions): brief note naming the
specific execution IDs that were approved ('don't re-request those').
No 'continue the plan' directive — the model knows to continue.
VERSION 0.7.1 → 0.7.2
P1: add docker compose (logs|ps|top|config|images|port|cp) to read-only
allowlist. docker compose logs was classified as config_mutation, causing
individual approval cards for read-only inspection commands.
P2: remove approval entries from activityLog. They were always status=running
and never transitioned to done (the derived store builds from tool-call
text, not execution status), causing AgentIndicator to latch onto a stale
'Approval: ...' entry and never clear — even after the session completed.
P3: remove InlineApproval from Chat.svelte. The green 'Completed in 1s on
lxc:...' boxes were noise in the chat stream. Approval UX belongs in the
Operations page (already has it via Ops.svelte), not inline in the chat.
P4: stale execution cleanup. Startup sweep (mark >1hr non-terminal as
cancelled) + 5-min periodic sweep (mark >10min non-terminal as cancelled).
98 orphaned executions accumulated from eval testing (39 running from
apt_upgrade:audit timeouts, 19 pending_approval, 3 approved).
P5: refuse second config_mutation run when an approval is already pending
for the session. Without this, the agent queues N individual approvals
before the operator can respond — confirmed in session 20757eb9 (two
approval cards for what should have been one plan-level approval).
VERSION 0.7.0 → 0.7.1
P1 plan-first: run handler refuses without propose_plan (structural gate,
not SOUL.md prose). Plan window decoupled from set_goal — config_mutation
auto-run only on operator approval (assent window). Closes the approval-free
config_mutation hole confirmed in session d0d562e0.
P2 iteration: reopenSession flips terminal→executing, marks prior plan steps
replaced, clears outcome. proposePlan excludes replaced from in-flight check,
bumps generation. A follow-up on a completed session starts a new sub-task
with a fresh plan — no more errPlanInFlight dead end.
P3 reasoning: accumulate per-iteration text into the persisted row instead
of overwriting with the last text event. Reload shows intermediate thinking,
not just the final summary.
P4 read-only allowlist: add find, tree, locate, systemctl list-timers/
list-unit-files/show, timedatectl, hostnamectl, systemd-analyze, rclone
ls/lsl/md5sum/check/cryptcheck. Fixes the find misclassification from
d0d562e0.
P5 eval harness: new assertion kinds (proposes_plan, plan_before_run,
plan_generations), multi-turn followups, fetch /sessions/{id}/plan. Four
manifests under evals/.
P6 SOUL.md: strip degenerate-case carve-out, add ITERATE step, update
set_goal guidance.
VERSION 0.6.0 → 0.7.0
- Terracotta (light) and Carbon (dark) themes with toggle
- Inknut Antiqua headings, DM Sans body
- Dot grid background on EntityGraph and GraphBackground
- Theme-adaptive graph colors on EntityGraph
- Art Nouveau chat styling (borders, underlines, blockquote quotes)
- Bullet point styles in chat prose
- Task goal in header, rename Overview→Tasks, New Task labels
- Logo uses var(--primary) for theme awareness
The unrooted graph endpoint caps at 500 entities with ORDER BY e.slug, which fills the cap with exec:* rows and excludes every host/lxc/service/vm entity. Since edges require both endpoints in the node set (ANY/ANY), 99.9% of edges were dropped — 500 nodes but only 1 edge survived.
Fix: select the 500 most-connected entities (by relationship count descending) so the topology is preserved. Result: 500 nodes, 900 edges across all relationship types.
Ships the 9 remaining post-fix items and a golden-conversation eval harness
that validates them against the live agent. All 4 evals pass.
SOUL.md (F.1, C.2, E.1):
- Consolidated three overlapping task-flow sections (MANDATORY TASK FLOW,
'Every chat is a task', 'AFTER EVERY TASK: WRITE BACK') into one. ~50
lines shorter. The operator's 'be more crisp' feedback.
- Added anti-patterns: don't re-execute on UI/sidebar complaints (C.2);
don't re-run fleet-wide audits when same-day knowledge exists (E.1).
- Updated approval vocabulary in step 4 to match tasks.go (approved/yes/
go/proceed/continue/ok/go ahead).
Tool-result strings (F.2):
- set_goal: tightened to 'Goal set. NEXT: pre-plan (read-only tools only).
Then propose_plan. Do not call run.'
- update_plan_step: added '(Advance with update_plan_step + run; do not
re-propose.)'
C.1 — completeTask rejects re-completion of a terminal session:
- Returns errTaskAlreadyComplete when status is already done/failed.
- The tool result directs: 'Task is already complete. Do not call
complete_task again. If the operator pointed out a UI/sidebar
inconsistency, fix it with update_plan_step...'
B.4 — Surface real model error text:
- chatWith's error event now includes finish_reason + refusal text:
'Nomos returned an empty or unusable response (finish_reason=length).
Retry or rephrase.' instead of generic 'empty response'.
- The resume-failed note already carried errText (B.3), which now has
the real context.
B.5 — Back off between resume retries (4s, 8s):
- resumeSession now sleeps before attempts 1 and 2 (exponential backoff).
A transient provider issue gets time to clear instead of 3 identical
calls in 3 seconds.
B.6 — Don't persist the empty placeholder as a visible bubble:
- If a chat turn ends with no text and no tool calls (model empty-response'd
and all retries failed), delete the placeholder row instead of persisting
an empty bubble. The error was already streamed via done+error=true.
E.2 — list_lxcs last-audited hint:
- The list_lxcs result now includes last_audited_at — the most recent
knowledge entry (tagged audit/update, or titled audit/update) linked
via an 'about' edge. The agent can see 'nextcloud — last audited today'
and skip re-running it.
Tool-call doubling bug fix (found by the eval harness):
- main.go + continue.go: the tool_use and tool_result events were both
appending separate entries to the persisted tool_calls array, doubling
every tool call in the transcript. Confirmed pre-existing (d9cdcee1,
v0.3.x era). Fixed: tool_use creates the entry, tool_result merges the
result into the same entry (matched by id). One entry per tool call.
Golden eval harness (cmd/nomos/eval/):
- A standalone Go program that loads YAML manifests of golden conversations
+ assertions, sends prompts to the chat endpoint, drains the SSE stream
(keeping the agent's context alive), and scores structural assertions
against the persisted transcript.
- 4 golden conversations covering: trivial read-only (degenerate case),
plan + proceed (the original duplication bug), UI complaint (no re-exec),
fleet audit (knowledge preferred over re-execution).
- Structural assertions only (tool-call sequences, plan steps, writeback,
completion) — text quality is model-dependent and not scored.
- Run: go run ./cmd/nomos/eval -gateway http://localhost:8092 -manifest
cmd/nomos/eval/evals/*.yaml (~$0.10/run in OpenRouter credits).
Eval results (4/4 passed):
trivial_readonly: 2 tool calls, no plan, no run
plan_advances_on_proceed: 13 tool calls, propose_plan x1, writes back
ui_complaint_no_rerun: 12 tool calls, propose_plan x1, writes back
knowledge_preferred_over_rerun: 7 tool calls, search_knowledge x1, 0 run
Version 0.5.2 -> 0.5.3 (minor: eval harness + structural hardening).
The overview background graph and the Knowledge Base graph both rendered
empty because the SPA's OIDC access token expired (~5 min TTL) and was
never refreshed. fetchWithAuth called getToken() synchronously (no refresh);
ensureToken returned the stale token without refreshing; storeTokens
discarded expires_in; the resulting 401 made fetchGraph return null and
both graphs drew nothing, with no error surfaced.
- oidc.ts: track expiresAt from expires_in; getToken() returns null within
30s of expiry; ensureToken/initOIDC refresh instead of returning stale
tokens; isOIDCConfigured no longer claims configured on expired-only state
- config.ts: fetchWithAuth awaits ensureToken (refresh on demand), falls
back to static token if OIDC can't yield one, flushes OIDC session on 401;
sseUrl is async + refreshes before constructing the EventSource
- stores/events.ts: connect() awaits the now-async sseUrl
