Product feature blueprints convert fuzzy requirements into shippable milestones. Use these patterns to map feature intent to agent boundaries, package choices, and readiness checks before launch.
How to use this page
- Pick the blueprint closest to your current product requirement.
- Keep phase 1 limited to one user-visible workflow and one reliability objective.
- Attach one training module and one docs checklist before adding secondary features.
Prerequisites
- Complete one bounded implementation from Quickstarts by Persona or Counter Agent.
- Choose runtime boundaries in Reference Architectures.
- Validate package intent with Package Matrix.
- Review release guardrails in Production Readiness: Supervision, Telemetry, and Failure Modes.
Blueprint format
Each blueprint should include:
- User-facing outcome and explicit non-goals.
- Owning agent boundary and required package set.
- Verification proof route (example or command path).
- Launch gate checklist tied to docs and training assets.
Template:
Feature:
Owner:
Primary workflow:
Non-goals:
Required packages:
Proof route:
Readiness checks:
Blueprint library
Blueprint A: Conversational support assistant
- Outcome: One production support flow that answers with tool-backed context.
-
Required packages:
jido,jido_action,jido_ai - Proof route: Examples index
- Launch checks: Security and Governance, Actions and Validation
- Non-goals: autonomous ticket closure, cross-team escalation routing
Blueprint B: Demand operations workflow
- Outcome: One event-driven demand scoring flow with deterministic state transitions.
-
Required packages:
jido,jido_action,jido_signal - Proof route: Demand Tracker Agent example
- Launch checks: Signals and Routing, Incident Playbooks
- Non-goals: full forecasting pipeline, multi-region replication
Blueprint C: Internal ops co-pilot
- Outcome: One operator workflow that proposes actions and records directive results.
-
Required packages:
jido,jido_signal,jido_ai - Proof route: Jido Documentation
- Launch checks: Production Readiness Checklist, Production Readiness: Supervision, Telemetry, and Failure Modes
- Non-goals: replacing existing incident command process, auto-approval of high-risk actions
Proof check: feature blueprint stays bounded
alias AgentJido.Demos.DemandTrackerAgent
alias AgentJido.Demos.Demand.BoostAction
agent = DemandTrackerAgent.new()
{agent, directives} = DemandTrackerAgent.cmd(agent, {BoostAction, %{amount: 20}})
{agent.state.demand, Enum.any?(directives, &match?(%Jido.Agent.Directive.Emit{}, &1))}
Expected result:
{70, true}
This proof confirms a blueprint can define a narrow workflow with explicit state changes plus side-effect directives.
Selection checklist
- Confirm the first release has one primary workflow and explicit non-goals.
- Confirm package choices match runtime boundaries and signal needs.
- Confirm at least one runnable proof route before adding integration breadth.
- Confirm readiness checks are assigned to specific owners.
Get Building
Pick one blueprint, implement the smallest proof that reaches user value, and document non-goals before sprint planning. Then use Mixed-Stack Integration to connect that blueprint to existing product services.