my-brain

name

my-brain

description

Give any agent ITS brain. First call routes the agent's purpose over the registered brains (bandit-tempered) and binds the best fit, or births a new specialized brain (charter + seeds, digested, registered). Later calls return the same bound brain. Compositions the agent produces become nested sub-brains — specialization accrues. Use when an agent needs persistent domain knowledge it will query repeatedly.

my-brain — the optimal brain configurator, as a callable skill

# first call: route-or-birth, then bind
python -m brain_agent.my_brain AGENT "what this agent is for" [--seed name=/path/file]

# every later call: the same brain, optionally queried
python -m brain_agent.my_brain AGENT --query "QUESTION"

From Python (or the RLM shell, where these are preloaded):

b = await my_brain("research-dept", task="qualify leads against the ICP")
answer = await b.query("does ACME fit tier A?")
await compose("research-dept", "acme_review", {"verdict": text, "evidence": quotes})
# -> the agent's brain now contains compositions/acme_review/ — a nested
#    specialist brain. Subdirectories ARE sub-brains; nothing else needed.

Requires HEAVEN_DATA_DIR (registry, bindings ledger, newborn brain dirs) and a heaven-routed model key for routing/digesting. Bindings persist in $HEAVEN_DATA_DIR/registry/agent_brains.json.