Spice 1.0.0 Roadmap

Spice 1.0.0 is the first “all-in” release built around the unified SpiceMessage + GraphRunner execution model. All of the legacy Flow/Swarm/DAG hybrids have been replaced with a single, testable runtime that ships with Redis/Kafka eventing, HITL automation, and Spring Boot wiring out of the box.

This page captures the high‑level direction for 1.0.x and how the five primary modules fit together.

What’s New in 1.0.0

• SpiceMessage everywhere – agents, tools, graph nodes, checkpoints, and event buses exchange a single typed payload with explicit ExecutionState.
• DefaultGraphRunner – validates graphs, manages retries/idempotency, records vector embeddings, and emits lifecycle events.
• Distributed EventBus – Redis Streams and Kafka implementations sit beside the in-memory bus so production deployments can choose their transport.
• Spring Boot Starters – modern auto-config configures GraphRunner, Redis-backed caches, EventBus backends, and HITL queue/arbiter helpers with simple application.yml switches.
• State Machine Extension – HITL pause/resume, tool retry/backoff, checkpoint persistence, metrics, and actuator visualizations ship as a dedicated starter.
• Spring Boot AI Integration – bridges Spring AI ChatModels to Spice agents with factories, DSL builders, and registries.

👉 Dive into the full breakdown in Spice Framework v1.0.0 Guide.

Module Focus

Module	Core Responsibility	Notes
spice-core	Graph runtime, state machine, caching, idempotency, event bus, arbiter/queue primitives	Pure Kotlin/JVM – no Spring dependency
spice-agents	Standalone OpenAI/Anthropic agents + mock tooling	Ships its own Ktor client stack
spice-springboot	Auto-config for GraphRunner, Redis pools, EventBus backends, HITL queue/arbiter wiring	Enables property-driven deployment
spice-springboot-statemachine	HITL pause/resume automation, retries, event emission, metrics, visualization	Designed to plug directly into Kai-core
spice-springboot-ai	Spring AI → Spice adapter (factories, DSL, registry)	Lets existing Spring AI apps opt into Spice orchestration

Need a refresher on each module? See Modules & Responsibilities.

1.0.x Priorities

1. Operational Hardening
   • Richer metrics on EventBus backends (lag, consumer health)
   • Redis queue/arbiter adapters that reuse the shared Jedis pool
2. Observability & Tooling
   • OpenTelemetry exporters for GraphRunner and state machine listeners
   • Developer tooling for visualizing checkpoints and node transitions
3. Integration Kits
   • Spring Boot auto-config for Redis Streams/Kafka EventBus selection
   • Optional Spring Messaging bridge for existing queues
4. Docs & Examples
   • Expand the Getting Started and Examples sections with multi-module samples
   • Migration notes focused solely on 1.0.0 (legacy migration guides have been retired)

How to Adopt 1.0

1. Start with the Quick Start to wire a graph + state machine.
2. Configure the new properties outlined in the 1.0.0 Guide to enable Redis/Kafka/HITL when needed.
3. For Spring AI users, jump to spice-springboot-ai to re-use your existing ChatModels.

1.0.x is where Spice becomes opinionated and production ready—no more juggling parallel runtimes or bespoke queue scripts. Let us know what’s missing and we’ll fold it into the roadmap! 🚀