Policy-driven task orchestration and failover (autopilot) that keep AI coding assistants running unattended — observable, replayable and auditable end to end.
Autopilot orchestration — an unattended task pipeline
Autopilot is Dawvia's orchestration layer. It drives task triggering, routing, retries and failover with policies, so AI coding assistants keep running unattended across long jobs. Every run is recorded in full for observability, replay and audit, so problems localize to the exact call and account.
Why move from manual switching to policy self-healing?
With an account pool and quota balancing, a single request already picks the best account. Autopilot goes further and hands whole task sequences to policy:
- Configure triggers, retry counts and fallback rules once — long-running tasks need no babysitting.
- Auto-failover and retry on node failure — the pipeline self-heals without interruption.
- Full traces mean problems localize to the exact call and account.
What are a request's four stages?
A call moves through four stages in Autopilot:
- Trigger: a coding assistant fires a request into the unified orchestration entry.
- Smart routing: pick the best account by remaining quota, rate and priority (see
- Failover: on limits or errors, switch to a standby account seamlessly.
- Audit: record the full call chain and outcome for observability, replay and review.
# Define a flow: trigger → route → retry 3x → fall back to standby pool
dawvia autopilot flow create \
--on request \
--route quota-first \
--retry 3 \
--fallback standby-pool
# Inspect flow state
dawvia autopilot status
How does Autopilot stay observable and replayable?
Running unattended requires "see it when it breaks, replay it to review":
- Live state and call chains are on one screen in the remote console.
- Every run is traced end to end — replay to localize issues, export for compliance.
- Audit and permission model in Security & compliance.
Which plans include Autopilot?
Autopilot orchestration starts on Pro; enterprise security governance and usage insights start on Team. Compare on the features overview and the pricing page.
What should I read next?
- Lay the resource base: Account pool
- Balancing and failover: Quota balancing
- Observe live: Remote console
- A team walkthrough: Build an AI IDE account pool for your team