Diary

Diary

Structured diary for users and Agents, based on the Unified Entity Model diary_entry container plus child content_block (text) bricks.

vs memory system

Capability	Diary	Memory bricks (limbic / narrative / dream)
Source	User / Agent writes actively	Sleep pipeline / memory tools
Storage	entities + diary_entry + child content_block	Same diary containers + semantic tags (limbic / narrative / dream)
Editing	Plain text blocks updateable	Semantic bricks append-only / soft-deprecate
Namespace	subject default private world	Agent private world for sleep-written bricks

Diary UI labels semantic bricks（梦境 / 情绪 / 自传）as read-only sections in the day entry.

User / Agent isolation

• User and Agent diaries live in the default private world of subjects resolved from boot-time ResolvedWorldContext / habitat_runtime_config.worlds (user_subject_id / agent_subject_id).
• Shell header User / Agent toggle selects which subject’s diary to view (see entity-model.md global Subject scope).
• LLM tools (ToolSet diary) default to the caller subject’s private world (conversation LLM → agent world); optional world_id overrides (e.g. user_world_id from system prompt).

Data shape

Container (diary_entry):

• body.entry_at — ISO 8601, when the diary entry occurred (unique per day per subject private world)
• tag_ids — optional tags（顶层 entities.tag_ids，指向同 World 的 tag entity）
• title / summary — entity columns
• container content is not used for body text (kept empty after migration)

Blocks (content_block, block_type: text):

• body.parent_id → diary entry id
• body.sort_order — view order (no semantic precedence)
• text lives in the block’s content column
• optional title (entity column) — user-editable block heading; empty means no heading weight in UI
• optional tag_ids (entity column) — unified entity tags（与日记容器同一模型）
• semantic component tags on the block (dream / limbic / narrative) render as read-only labels（梦境 / 情绪 / 自传）beside the title; they are not the block title

日记块模板 (diary_block_template entity, subject private world):

• entities.title = 模板名称（列表/管理显示）
• entities.content / entities.tag_ids stay empty — do not store insert payload there
• body.preset = { title, content, components, tag_ids } applied when inserting a new text block
• SAP: diary.templateList / templateCreate / templatePatch / templateDelete
• Empty world lazy-seeds「今日回顾」「运动」

UI also persists per-block expand/collapse in localStorage (default expanded).

One-shot migration: Habitat runMigrations moves legacy diary content into the first text block and clears the container column.

Agent tools (ToolSet diary)

Load via toolset_load with diary. Tools locate entries by date (YYYY-MM-DD)；default today (CST noon …T12:00:00+08:00)；no diary_create — use diary_append (creates empty shell for the day if missing, then adds a new text block). All tools accept optional world_id.

Tool	Description
diary_append	New text block for date; create shell if missing; tags/tag_ids only on shell creation
diary_update	Update entry metadata (title/summary/tags/tag_ids) by date — not body text
diary_get	Read entry + blocks for date; error if not found
diary_delete	Delete entry for date (cascades child blocks); returns { ok, action, date }
diary_list	List by entry_at DESC (default limit=20, offset); optional date/tags/tag_ids
diary_search	Hybrid search over text blocks, returns matching diary entries

Fine-grained block CRUD / reorder: ToolSet content-block.

vs SAP/UI: Shell /diary uses SAP diary.* / diary.block* located by entity id; Agent ToolSet uses date uniformly.

SAP methods

UI satellite @freeanima/satellite-diary (/diary) calls SAP:

• diary.list / diary.create / diary.append / diary.patch / diary.delete / diary.get / diary.search
• diary.blockCreate / diary.blockPatch / diary.blockDelete / diary.blockReorder (text only; create/patch accept optional title / tag_ids / components)
• diary.templateList / diary.templateCreate / diary.templatePatch / diary.templateDelete（日记块模板；name ≠ preset）

All methods require subject_kind: user | agent.

• diary.list defaults to entry_at DESC, limit=20, supports offset pagination
• diary.create optional content → first text block (container content stays empty)
• diary.append → new trailing text block
• diary.patch → container metadata only
• diary.delete → cascade-delete child blocks
• diary.template* → CRUD for diary_block_template（模板名 name + 插入载荷 preset）