Per-sub control — pick which sub-orchestrators run
Each sub gets its own enable checkbox + LLM-per-sub slider (0..100). Disabled subs contribute 0 to the LLM tier total. Click only the boxes for the sub-orchestrators you actually need — saves cost, reduces fan-out noise.
Enabled subs
7
LLM sub-spawn
700
LLM tier total
708
Binary pool
2,124
Grand total
2,832
Enable all
Disable all
Crypto only (Uriel)
Lead-gen mode (Zadkiel + Michael)
Research team (Uriel + Malika + Katan HaShem)
Set all to 50
Set all to 25
Set all to 10
Smart auto-allocate
When enabled, every LLMStep queries the SmartAllocator to recommend N workers (1..sub_cap) based on task complexity. Short factual = 1 worker. Heavy research = many. Saves cost on simple tasks; ramps up automatically for hard ones.
Smart auto-allocate
Heuristic-based by default (deterministic, no LLM cost). Optional LLM-judge mode when a router is configured.
Auto-allocate workers per task:
disabled — uses static llm_per_sub cap
Test the allocator
Click RECOMMEND to see the SmartAllocator decision.
Optional · Fill all slots with one provider
Quick path: paste one OpenRouter key, click Fill all 8. Or skip this if you already used a hardware profile.
Fill all 8 slots
Same as the Pro dashboard — one provider, all 8 slot positions.