Repository navigation
ADR 0004 Pause seam
pavelravvich edited this page Sep 25, 2026
·
3 revisions
- Status: Accepted
- Issue: #32 (epic #26)
The core ends a run at a terminal node; a follow-up is a new run in the thread. Some projects need
a human to approve actions (write tools) mid-run. Added ad hoc, interrupt would conflict with
spend recording while streaming, budgets and Terns.
- The core ships an explicit, optional pause seam (
RunDeps.pause: checkpointer +needsApproval), off by default, andresumeAgent. - The pause happens at the graph level (
approvalnode), not inside the agent loop: the call is held, the agent node ends normally, theninterrupt. This keeps all spend before the pause in state and in the ledger. - Five consistency rules (ledger, budget, Tern, limits, rejection) are specified and proven with
LangGraph's
MemorySaver; durable checkpointers and the approval channel stay in projects.
- Adding approval to a project is configuration, not core changes.
- The agent runs again after a decision; tools must be safe to see a stored result for an identical call (they are not executed twice).
AGBoss
- Home
- Delivery Pipeline
- Architecture
- Configuration
- Workflow — the graph
- Advanced Routing (DSL)
- Routers
- Guards
- CLI — agboss
- App — createApp
- Context and Authorization
- Testing
- Suites and Comparisons
- Components
- Standard DTOs
- Example: job-scout
- Agents — the loop
- Model providers
- Tools
- Knowledge bases (RAG)
- FinOps
- Tracing
- Observability
- Terns and evaluation
- Profiles and comparisons
- Pause seam
- Decisions (ADR)