New realism_config table (uuid PK + unique key) + two repo methods
(get/set) backs an admin-only GET/PUT /api/v1/realism/config surface.
The planner now exposes apply_payload(payload) / current_payload() /
reset_to_defaults() and reads its weights through mutable module
globals; pick() resolves the live values each call. Validation
catches negative weights, zero totals, out-of-range canary_probability,
unknown content_class names, and silently drops cross-list entries
(canary class on the user list, etc).
The orchestrator worker calls _refresh_realism_config(repo) on
startup and every 5 ticks (~5min at 60s interval). Operator changes
land within one refresh window with no bus signal — the simpler path
for a knob whose latency tolerance is minutes.
Surfaces realism subsystem state on the existing worker heartbeat
extra hook (system.orchestrator.health) — no new bus topic. Payload
carries {llm_enabled, llm_backend, llm_model, llm_breaker_state}, so
the dashboard's worker panel renders a live LLM badge with a colored
breaker-state dot:
closed (green) — LLM healthy
half_open (amber) — cooldown elapsed; next call is a probe
open (red) — short-circuiting to deterministic templates
Heartbeat is the canonical worker self-report channel; piggybacking on
extra(...) avoids a new topic family while keeping the snapshot
recomputed each beat (30s).
The orchestrator worker clipped last_body at write time, but the repo
didn't enforce. A future caller that forgot the clip would write the
full body. Move the clip to record_synthetic_file and
update_synthetic_file via SYNTHETIC_FILE_BODY_LIMIT in
decnet/web/db/models/realism.py. Worker now passes the full body and
trusts the repo. Tests retargeted to assert repo enforcement.
Stage 6 of the realism migration. User-class file bodies (note,
todo, draft, script) optionally get LLM-authored content; system
classes (cron / daemon logs, /tmp caches) stay template-only because
formulaic *is* the right look for them.
New surface:
- realism.llm.circuit.LLMCircuitBreaker — process-local sliding-window
breaker. 3 consecutive failures trip open; 60s cooldown to half-open;
half-open success closes, failure re-opens. Protects the orchestrator
tick from sustained Ollama wedges (per-call timeout already covers
one-shot hangs).
- realism.prompts._style — em-dash suppression lifted from the
email prompt. Persona.uses_llms_heavily opts out per the
feedback_em_dash_llm_tell.md memory. Includes strip_em_dashes
belt-and-braces sub for output that slipped past the prompt rule.
- realism.prompts.filebody — class-conditioned prompts (note / todo
/ draft / script) with persona context, language pinning, output
shape rule.
- realism.bodies.make_body_with_llm — async wrapper around make_body
that calls the LLM when one is provided AND the breaker allows.
Falls back to template on timeout / error / empty / system-class.
Wiring:
- scheduler.pick_file accepts optional llm + llm_breaker + llm_timeout.
When the planner picks a create action and the content_class is a
user-class, the body_hint is replaced with the LLM-authored body
(or falls back to the deterministic body_hint).
- orchestrator.worker constructs get_llm() at startup gated by
DECNET_REALISM_LLM env var (any non-empty value enables; empty /
"off" / "none" / "0" disables). Passes llm + breaker through every
tick.
- decnet orchestrate gains --llm/--no-llm flag overriding the env var.
Stage 3b of the realism migration. A TODO.md planted on Monday gets a
checkbox flipped on Tuesday; a notes file grows a follow-up line; a
cron log gets a fresh entry tacked on. The synthetic_files row's
edit_count, last_modified, and content_hash advance.
New surface:
- EditAction dataclass (peer of FileAction in scheduler.py): carries
decky, path, persona, content_class, previous_body, mtime, and
synthetic_file_uuid for the worker's update path.
- realism.bodies.next_iteration(cls, persona, prev, rng): per-class
deterministic mutators. TODO flips an unchecked box and/or appends;
notes/drafts/scripts append; logs are append-only (mirroring real
log behaviour). Canary, cache_tmp, email raise KeyError —
unsupported.
- realism.planner.pick gains an edit branch: 60% create, 30% edit
(when an edit_candidate is supplied), 10% leave-alone. Returns
None on leave-alone — quiet ticks are realism too.
- scheduler.pick_file pre-fetches a single edit candidate via
repo.pick_random_synthetic_file_for_edit ~50% of ticks; the
planner decides whether to use it.
- SSHDriver._run_edit: turns next_iteration output into a
plant_file call (mtime-bumped, mode 0o644). Stashes new_body in
result.payload so the worker can hash it for synthetic_files.
- worker._bump_synthetic_file_after_edit: patches edit_count + 1,
last_modified=now, content_hash, last_body for the row UUID.
No-op when the row was pruned mid-flight.
- events.to_row / topic_for / event_type_for now recognise
EditAction (kind="file", action="file:edit").
Stage 3 of the realism migration. Replaces orchestrator/scheduler.py's
hardcoded _FILE_TEMPLATES/_USERS (3 templates emitting epoch-suffixed
filenames like notes-1777315854.txt with identical bodies per
template) with a persona-driven realism engine.
New surface:
- SyntheticFile SQLModel (synthetic_files table, UNIQUE on
decky_uuid+path) — per-(decky, path) state for the future
edit-in-place flow. Pre-v1, no _migrate_* helper.
- BaseRepository methods: record_synthetic_file,
update_synthetic_file, list_synthetic_files,
pick_random_synthetic_file_for_edit (used by stage 3b).
- realism/naming.py: per-content-class filename templates,
persona-conditioned. /var/log/cron.log + logrotate skeleton for
system-class; /home/<persona>/TODO.md, scratch.md, etc. for
user-class. Anti-regression test pins "no 8+ digit decimals in
basenames" (the realism failure today).
- realism/bodies.py: deterministic body templates per content_class.
TODO body uses checkbox markdown, script body has a shebang, cron
body matches syslog cron shape ("CRON[PID]: (user) CMD (...)").
- realism/planner.py: pick(deckies, now, rng) returns a Plan.
Diurnal-gated, weighted user/system content split (70/30 user
bias). Create-only in stage 3; edit branch lands in stage 3b.
Scheduler split:
- scheduler.pick is now traffic-only (sync).
- scheduler.pick_file is async, takes a repo, resolves personas
(Topology.email_personas for topology-source deckies; global
realism.personas_pool otherwise), and maps Plan -> FileAction.
- FileAction gains persona/content_class/mtime fields.
Worker:
- _one_tick rolls 50/50 between traffic and file each tick. After a
successful FileAction plant, _record_synthetic_file persists or
patches the synthetic_files row (catching the unique-constraint
collision on re-plant of the same path).
- SSHDriver._run_file passes action.mtime through to plant_file so
files don't all stamp at wall-clock-now.
Switches _one_tick from list_running_topology_deckies to
list_running_deckies (the union view added in 095500a). Resolves the
permanent "no actionable deckies (running+ssh count=0)" log on hosts
running only unihost MACVLAN / IPVLAN decoys — the orchestrator now
sees fleet_deckies rows alongside MazeNET topology rows and SWARM
DeckyShard rows.
Also fixes the misleading log message: the old "running+ssh count=N"
reported the *pre-filter* total (count of all running deckies, not
the SSH-eligible subset that scheduler.pick actually evaluates). New
line breaks down running, ssh_eligible, and per-source counts so
debugging "why isn't it picking?" no longer requires reading
scheduler internals.
Regression test: orchestrator integration suite now seeds fleet_deckies
rows (not just topology_deckies) and verifies a tick picks them and
records an event with dst="local:fleet-*" — proving the original bug
on the operator's mothership is fixed.
Every 100 ticks, trim per-dst_decky_uuid history down to 10000 rows
(oldest first). Keeps the events table bounded on long-running fleets
without paying the cost on every write.
Adds a new decnet orchestrate worker whose job is to keep the honeypot
ecosystem from looking suspiciously static — a frozen LAN with no
inter-host traffic and no filesystem aging is its own honeypot tell.
MVP scope:
- New OrchestratorEvent table + repo methods (purpose-built sibling
to Log so synthetic events stay separable from attacker-driven ones).
- New orchestrator.{activity,file}.<decky_id> bus topics +
system.orchestrator.health heartbeat.
- SSH-only driver. Traffic action runs python3 inside src container
to TCP-connect dst:22 and read the SSH banner — real on-the-wire
SSH-protocol traffic without shipping creds. File action drops or
refreshes a small file via docker exec on the destination.
- Random scheduler (50/50 traffic/file when >=2 SSH-capable deckies
are running). Diurnal shaping, role-aware pairing, and session-aware
backoff are explicit non-goals for MVP.
- CLI registration, systemd unit (SupplementaryGroups=docker),
worker-registry entry so the dashboard shows orchestrator health.
- 11 tests: scheduler policy, driver argv shape + injection-safety,
end-to-end one-tick integration with FakeBus + SQLite.
