mockup v2

AVE

Mastra Operator
dev / MLZero:4111 graph connected

Operator console for real work

Agent status, execution flow, and operator records. Not a toy chat.

context-1 ready cortexadb 2520 chunks
Nano Router
nvidia/nemotron-3-nano · LMStudio :1234
idle
142
routed
98.6%
accuracy
85ms
latency
Orchi Orchestrator
nemotron-orchestrator-64k · Ollama :11434
idle · 0 active runs
87
runs
3.2s
avg time
2
errors
Fifi Writer
phi4-mini · Ollama :11434 · 3.8B
writing report · thread_a7f3
56
docs
1.8k
avg tok
4.1s
avg time
❮/❯
Qwen Coder
qwen3.5-27b · LMStudio :1234
idle · ready
34
tasks
91%
review pass
6.7s
avg time

Execution timeline

last 1h
task_received
thread_a7f3 — sprint review report AI3
14:32:01
nano_routed → Fifi Writer
confidence 0.94 · nemotron-3-nano
14:32:01
agent_started — Fifi Writer
phi4-mini · generating draft · ~4s
14:32:02
review_started — iteration 1/3
reviewer checking output quality
14:32:06
review_approved
score 0.88 · passed iteration 1
14:32:07
response_built
envelope ready · action: none · 6.4s
14:32:08

Operator channel

4 agents
A
Aidar 14:31
Sprint review AI3 — what shipped, what's in progress, blockers.
N
Nano Router 14:31
Writing task → routing to Fifi Writer. Confidence 0.94

Operator response

memory used

Sprint review drafted. Context-1 recall confirmed 3 prior decision trails. Orchi stays on process view, Fifi handled the text generation in one pass.

phi4-mini 4.1s 4 steps memory yes
S
System 14:32
Task completed in 6.4s · 1 review iteration · action: none
Context-1 recall
Stack rule Context-1 :3000 is canonical shared memory endpoint
Topology Orchi → OpenWebUI :8081 · Phi-4 → Mastra lane
Model Access Layer SSH tunnels + alias registry (AI3-124 approved)
AVE LangGraph v0.3.0
Mastra Studio · MLZero:4111
Graph ML1:8765
Flowise ML1:3001
Context-1 ML1:3000 · uptime 8.4d
CortexADB · 2520 chunks