stello-agent

StelloAgent 创建配置教程。完整说明 createStelloAgent 的每个配置项，包含 sessionDefaults、storage、tools、skills、forkProfiles、session 层接入、orchestration 等。

Engine 职责定义：per-session-round 生命周期管理器。驱动 tool call 循环，管理单个 Session 的多轮对话。不感知树结构，不感知调度。

Fork 机制完整说明。覆盖 ForkProfile 与 EngineForkOptions 的字段对齐、四层 fallback 合成链（sessionDefaults → parent → profile → forkOptions）、systemPrompt 合成三种模式、skills 三态语义、持久化边界（SerializableSessionConfig 只固化 systemPrompt/skills）。任何涉及 fork / profile / stello_create_session / 配置合成 的工作都应读这个。

Stello 框架内所有 LLM 调用位置的消息结构速查。覆盖 Session 对话、compress、consolidate；应用层 reflection 调用由 orchestrator 自行决定。

Consolidation 触发机制：自动触发通过 Factory 配置内联，手动触发通过 StelloAgent API；全局 reflection 由应用层在 SDK 之上自行实现。

Server 层设计：传输层架构决策、StelloAgent 映射原则、连接态管理模式。存储层见 server-storage，Engine 细节见 engine-design。

@stello-ai/server 的 PG 持久化层设计决策和实现模式。触发条件：修改或引用 server 存储层。

Session 对话单元的设计理念、上下文组装规则、memory / insight 槽位语义、单一 Session 模型与跨 Session 通信模型。

Use when an agent should behave like KitKit Kitty: maintain global space awareness, read shared memory early, treat it as possibly stale, inspect conversation-tree digests, push useful insights, update durable memory, or create forks through kitkit-cli.

Use when an agent needs to work with KitKit through kitkit-cli: authenticate, choose spaces, understand the KitKit space/session/topology/shared-memory/insight/fork mental model, parse JSON output, or prepare for Kitty-like global conversation-tree work.

Use when an agent needs to treat a KitKit or Stello conversation tree as a document library: inspect topology, choose relevant session nodes, read targeted digests, compare branches, or build space-level context without loading full transcripts.

Use when an agent needs to create or plan KitKit fork sessions with kitkit-cli: choose source sessions, select none/inherit/compress context, apply fork profiles, attach skills, pass prompts, or place topology nodes.

Use when an agent needs to push, update, or reason about KitKit insights: cross-session findings, contradictions, warnings, reusable context, target-session selection, overwrite safety, and concise insight content.