System 05 Creator economy operations
Campaign decisions,
with receipts.
Turn a campaign brief and creator evidence into a defensible roster, budget, risk record, and human decision.
13 / 13Golden scenarios
18 / 18Evaluation measures
0External writes
Evidence becomes a decision here.
The council will show the roster, budget, exceptions, agent trace, and approval boundary.
Council decision
Ready for reviewRoster is ready for accountable human review
Campaign name
0Roster score
Planned creator spend
$0 / $0
Ranked roster
Creator evidence
Exceptions
What needs attention
Agent trace
How the result formed
Human gate
Campaign or talent lead required
Approval records a receipt. It performs no external action.
Run containedThe campaign run could not complete.
Release proof
A working system,
not a scripted screen.
Every output comes from the public API, typed campaign data, deterministic policy, and a checkpointed LangGraph run.
Audience, budget, conflict, capacity, rights, safety, evidence, and security cases.
Decision accuracy, evidence binding, schema safety, policy, and input sensitivity.
HTTP, graph, agent, risk, audit, approval, budget, policy, and error monitoring.
PostgreSQL state, LangGraph checkpoints, idempotency, and signed serverless receipts.
System architecture
Twelve bounded nodes. One human decision.
Built with
Python
FastAPI
LangGraph
PostgreSQL
OpenTelemetry
PythonFastAPILangGraphPostgreSQLOpenTelemetry