🌙 CONSOLIDATION ENGINE (5 min interval · sleep-spindle bursts)
passes run: 0
last pass: never
dreaming now: no
pass interval: 5 min
📚 Dictionary API + Vocabulary
🌐 API SMOKE TEST
status: pending
fetch available: —
💾 CACHE
size: 0 / 10000
hits: 0
errors: 0
rate-limited: 0
in-flight: 0
📖 VOCABULARY (K→PhD)
prefetched: no
defs taught: 0 / … (0%)
🧠 Cortical Microstructure
🎯 WIRING ASSERTION
status: unknown
🏛 MICROCOLUMNS
columns: 0
columnSize: 80
📑 LAYERS
L1: 0
L2/3: 0
L4: 0
L5: 0
L6: 0
⭐ HUBS + OSC
hubs: 0 (0%)
θ phase: 0.0
γ scale: 1.0
Φ proxy: 0.0
🌟 GlobalWorkspace + Predictive Coding
🌟 IGNITION (GW broadcast)
current: subthreshold
strength: 0.000
rate: 0.0% ticks ignited
🔮 PREDICTIVE ERROR (Friston)
last: 0.000
trend: →
📖 DEFINITION LEARNING RATE
defs/hour: 0.0
background-trickle during dream cycles
✨ Compositional Emergence
📜 VERBATIM
count: 0
rate: 0%
🌟 NOVEL
count: 0
rate: 0%
first ms: —
↗ PARTIAL
count: 0
rate: 0%
🎤 MAX NOVELTY
—
score: —
🗣 Word-Creation Tip-of-Tongue
No candidates yet — emission rejections haven't produced a tip-of-tongue compound. Either chat hasn't fired enough OR signal floor is too lenient.
🗣 Speech Health (SPEAK) — separability + coherence floor
No speech-health data yet — populates after the first word-emission teach phase.
HER VOICE — measured, not inferred
unmeasured
💬 Chat-Time + Dream-Time Learning
💬 CHAT-HEBBIAN
turns: 0
pairs: 0
last: —
errors: 0
🌙 DREAM-RECOMB
dreamed: 0
consolidated: 0
last: —
🧮 Neuron Utilization
language cortex lifetime fired: —
weight recruitment (CSR rows): —
📡 WS Backpressure (GPU client)
🔌 CONNECTION
compute.html: —
buffer: 0.0 / 500 MB
📤 DROPS (after timeout)
total: 0
rate: 0.00 / sec
last: never
shadow: clean
⏳ ABSORBS (drained)
total: 0
no Hebbian update lost
⚠ ENOBUFS (OS-level)
total: 0
first 3 logged then silenced
🎤 Emit Diagnostic (silent-fail tracer)
Surfaces cortexCluster._lastEmitRejection — set by emitWordDirect when no word bucket clears the adaptive signal floor.
⛔ LAST REJECTION
reason: —
age: —
📊 SIGNAL VS FLOOR
bestMean: 0.000
floor: 0.000
📈 ADAPTIVE EMA
EMA: 0.000
samples: 0
🎚 LIVE FLOOR
current: 0.000
noise-floor=0.001 · adaptive=ema×0.5
Cluster Activity
Conversation Activity (aggregate)
Per-session activity counters only — user text is private and never displayed here. See docs/WEBSOCKET.md "Privacy Model".
🗣 Messages this session: 0
⏱ Last activity: —
👥 Active users last 60s: 0
Process Log
⚡ Community Compute & Auto-Scale (admin · all donor GPUs)
⚠ INSUFFICIENT COMPUTE — connected GPUs can't hold the current neuron count.
donors: 0
batch round-trip: —
community VRAM: 0 MB
size driver: 0 MB
capacity: 0
tier qualifies: 0
tier running: 0
pending: none
replicas: 0
last merge: —
DOWNSCALE RECTIFY — only when compute genuinely can't hold the running size (buffers for the buffers)
deliberate — retrains immediately at a fitting smaller size
Weight-SAFE. Forces the cortex GPU mirror to re-upload from the intact CPU master to the currently-connected donor — clears a gpuShadowDirty flag that's stuck because no donor has respawned. Touches no weights.
Inside the stable band [downscale-floor … upscale-gate] the brain just keeps running at its current neuron count — donors come and go freely. A single donor (or 10) leaving NEVER downgrades the running brain unless compute stays collapsed below the floor past the downscale hold window.