Every step — planning, tool calls, LLM, evidence checks — is traced and inspectable.
When an autonomous agent decides to release capital, every step of that decision is recorded as a trace. This page makes those traces readable — what the agent planned, which tools it called, what the LLM said, and what evidence was checked before the verdict.
service.name = 'weft-daemon' AND weft.milestone_hash = '0xwinningagent2'No live trace yet
Program officers do not need SigNoz accounts. They need a defensible receipt: what evidence was checked, what the agent did, whether the LLM merely narrated, and why capital released or stayed locked.
The boolean verdict is deterministic. The LLM span exists so teams can inspect cost and narrative behavior, not because an LLM is allowed to decide payment.
KeeperHub fallback, degraded peer quorum, and LLM failures become alertable conditions. A failed dependency is no longer hidden inside an autonomous run.
agent/scripts/weft_signoz_demo.sh && agent/scripts/weft_signoz_provision.shWhat you see here is illustrative — real onchain milestones, real SigNoz traces, but staged for demo. Ready to fund a real milestone on 0G Testnet?