Append-only
Events are immutable entries with parents, timestamps, and actors. Trails never rewrite history; they only extend it.
Trails engineSpec · DraftGlyphd Labs project
Tongbuku 同步库 is a trails engine for synchronized event streams you can replay and reuse as process templates.
Tongbuku 同步库 stores append-only trails of events you can replay, analyze, and reuse as templates. It keeps sync verifiable, views model-ready, and every turn aligned with ChatCard identity and EarthCloud meaning.
Trusted by Glyphd Labs for EarthCloud meaning trails, Buoychong worlds, and future VibeCherry/Funwae experiments.
Trail visual
Fractal tower · stacked event shards
Trail inspector
trail_room_alpha
turn.user#32
“Kickoff sync notes?”
room:alpha · 20:03:12Z
agent.plan#33
plan_summarizer()
glyphir:block_129 · 20:03:15Z
view.summary#34
LLM context window refreshed
server_head mkh_main
Sync snapshot
heads: 1 • lag: 0 events • views refreshed 3.2s ago
What is a trail?
History you can trust—and reuse.
A trail is the ground truth history for a document, room, or workflow. Append-only events capture every turn, transformation, and model output. Trails are verifiable via hashes, merge safely across devices, and stay linked to the semantic layers above them.
Contextful
Each event carries tags, channel hints, and stack metadata so you can filter by room, persona, tool, or tenant without bespoke joins.
Model-ready
Views and summaries stay aligned with the underlying trail so EarthCloud, Buoychong, and your own agents can ingest clean context.
Core concepts
Trail · Event · View / projection
Append-only sequence of events for a conversation, workflow, or document. Branches reconcile later through Merkle heads.
Immutable JSON envelope describing one action or observation. Carries actors, payload, parents, and tags for routing.
Derived representation of one or more trails—summaries, snapshots, or indexes that stay ready for models and audits.
Use cases
Where Tongbuku 同步库 fits
Chat trails & rooms
Capture every human and agent turn in one trail so support leads, compliance teams, and models can replay the same story.
Product design workflows
Link design reviews, approvals, and prototype demos as events in a trail so directors can inspect decisions and fork new runs.
Analytics & research notebooks
Query trails to study behavior over time, tag moments of interest, and hand clean projections to data scientists.
Training data & supervision
Export views from curated trails as labeled corpora for fine-tuning and eval, with traceable provenance for every example.
Data model & spec teaser
One event envelope, many projections.
Events, trails, and views are spelled out in the spec. JSON stays compact but expressive: actors, tags, payloads, and merkle metadata fit into a single envelope so you can ship quickly without inventing another event format.
- Events append via
POST /v1/events/append. - Clients sync using
POST /v1/syncwith Merkle heads. - Views render with
GET /v1/viewsscopes.
Event snippet
trail_room_alpha
{
"event_id": "evt_01JK9WQ3",
"trail_id": "trail_room_alpha",
"kind": "turn",
"timestamp": "2025-11-08T20:03:12Z",
"parents": ["evt_01JK9WQ2"],
"actors": {
"user_id": "user_en",
"app_id": "buoychong",
"card_id": "chatcard_room_alpha"
},
"payload": {
"role": "user",
"text": "Kickoff sync notes?",
"lang": ["en", "zh-CN"]
},
"tags": ["room:alpha", "world:earthcloud"]
}Example flows
How teams ship on trails
1
Write events
Clients append JSON events via `/v1/events/append` with dedupe + actor metadata.
2
Sync via heads
Devices and services exchange Merkle heads to request only the events they are missing.
3
Build views
Tongbuku 同步库 maintains snapshots, indexes, and GlyphIR-aligned summaries tuned for AI workloads.
4
Feed models
Views stream into EarthCloud, ChatCard, or your own pipelines as ready-made context.
Status & stack
Spec · v0.1 draftTrails engineCompatible with EarthCloud, ChatCard, Buoychong
Where it fits in the stack
ChatCard anchors who is acting, EarthCloud captures what each turn means, and Tongbuku 同步库 records when and where it happened as a reusable trail. Glyphd Labs ships the bundle so Funwae, Buoychong, and VibeCherry teams can share the same append-only history.
ChatCard (identity)
Cards establish who is acting. Events reference card_id so permissions, tone, and provenance flow through every trail.
EarthCloud (meaning)
GlyphIR semantics annotate each event before it lands in Tongbuku 同步库, keeping context reusable for models and audits.
Funwae & Buoychong (worlds)
Experiences, ops surfaces, and research sandboxes emit trails that Tongbuku 同步库 stores and VibeCherry/Funwae render as views.
FAQ
What people ask about Tongbuku 同步库
Tongbuku 同步库 complements your primary datastore. It handles append-only history, sync, and replay while you keep transactional state wherever you like.
Contact
Co-design the trails engine.
We are pairing with teams who want trustworthy history for AI-native systems—especially those integrating EarthCloud, Buoychong, ChatCard, VibeCherry, or Funwae stacks. Bring your requirements; we will share reference flows and help shape the spec.
Glyphd labs
info@tongbuku
Include a short note about your product, current stack, and what you need from trails.