Spec · GlyphIREarthCloudIntent · Translation · Trails
EARTHCLOUD · MEANING LAYER

A semantic protocol for multilingual AI.

EarthCloud defines a shared format (GlyphIR) for meaning: messages, intents, roles, and trails—across languages, models, and apps.It keeps what was said, what it means, and where it lives aligned for every system that touches it.

Read the specSee it in use
DefinitionGlyphIRStackUse casesSpec

Part of the Funwae ecosystem.

Used with Glyphd Labs, Buoychong, and Tongbuku.

Meaning · Trails · Worlds

GlyphIR is the envelope that keeps surface text, normalized meaning, context links, and quality metadata together—so any model or app can reason over the same intent.

Meaning core

GlyphIR

Spec status

Draft v0.1

Definition · Canon

What EarthCloud is (and isn't)

EarthCloud is the meaning layer for multilingual AI: an open specification (GlyphIR) that captures surface text, normalized meaning, participants, context links, and provenance so every system sees the same intent.

What EarthCloud is

  • An open semantic envelope (GlyphIR) for multilingual turns.
  • A contract for how apps and models describe messages, intents, and roles consistently.
  • A bridge between free-form language and structured trails / worlds.
  • Model-agnostic: any engine can emit or consume GlyphIR objects.

What EarthCloud isn't

  • Not a translation model or LLM.
  • Not a storage engine or analytics database.
  • Not a proprietary, productized SaaS API.
  • Not tied to one language pair—it is meant to be shared across stacks.

Where people use it

Use cases powered by EarthCloud

EarthCloud shows up anywhere you need consistent meaning across languages, channels, or agents. Here are a few places it is already working.

Cross-language chat

Buoychong pairs English ↔ Mandarin teammates. GlyphIR keeps turns aligned so summaries, replies, and trails stay in sync across languages.

Product design trails

Cross-border product teams capture decisions as structured GlyphIR turns, then write them to Tongbuku for replayable audits.

Knowledge & documentation

Docs assistants emit GlyphIR so intent, topic, and stance are searchable—no matter which model produced the answer.

World & kiosk experiences

Funwae worlds and on-site kiosks use GlyphIR to hand conversations between characters, agents, and visitors without losing context.

GlyphIR overview

What a GlyphIR block carries

EarthCloud doesn't decide what the model should say—it describes what was said, what it means, and how to connect it to the rest of your stack. GlyphIR keeps these layers bundled so every service interprets the same envelope.

Surface text

Original utterance, language, script, and any normalization hints so text can be reconstructed faithfully.

Normalized meaning

Intents, entities, topics, sentiment/stance—structured in a way models and tools can reason over.

Roles & participants

Who spoke, who was addressed, which agents or tools are in the loop, plus their tone/register preferences.

Context links

References to threads, rooms, worlds, Tongbuku trail IDs, message IDs, and related turns.

Quality metadata

Confidence scores, model or human provenance, translation quality hints, and safety annotations.

GlyphIR at a glance

Implementers can extend this shape, but the contract stays the same: surface text, normalized meaning, participants, context links, and quality metadata share one envelope. Trails (via Tongbuku or your own log) simply point back to these IDs.

{
  "glyphir_id": "gph_01HX...",
  "surface": {
    "text": "请帮我把这段内容总结一下给同事",
    "lang": "zh-CN",
    "source_lang": "en"
  },
  "meaning": {
    "intent": "summarize_content",
    "entities": [{ "type": "person", "name": "Chinese colleague" }],
    "stance": "helpful"
  },
  "participants": {
    "speaker": { "role": "user", "card_id": "chatcard_875" },
    "audience": ["teammate"]
  },
  "context_links": {
    "trail_id": "trail_8734",
    "room": "funwae/product-design"
  },
  "quality": {
    "confidence": 0.91,
    "source": "llm:glyphd-2025-03"
  }
}

Spec & implementation

Ready to implement GlyphIR?

The spec page covers the envelope (requests, GlyphIR blocks, realizations, context links) plus REST-ish endpoints you can adopt or adapt. Start there when you want to wire EarthCloud into your own stack.

What you'll find in the spec

  • Reference schemas for requests, GlyphIR blocks, realizations, and trail events.
  • Mapping notes for translating common LLM outputs into GlyphIR fields.
  • Examples that pair EN ↔ zh turns with the same semantic envelope.

Implementation notes

You can start simple: emit GlyphIR alongside your current prompts, then pipe trail events into Tongbuku (or any append-only store). As you mature, routers and realization services can speak GlyphIR end-to-end.

Explore the full spec

Stack placement

Where EarthCloud sits in the Glyphd stack

EarthCloud (GlyphIR) is the semantic layer. It pairs with ChatCard for identity, Tongbuku for trails, and the broader Funwae tools for how those conversations show up in the world.

ChatCard

Learn

Defines who is speaking, identity preferences, and provider context before a turn becomes GlyphIR.

EarthCloud / GlyphIR

Learn

Captures the meaning layer: surface text, normalized intent, participants, context links, and provenance.

Tongbuku

Learn

Stores append-only trails that reference GlyphIR IDs so every turn is auditable and replayable.

Funwae worlds / Buoychong

Learn

Places where people experience the conversation layer—cross-language chat and shared worlds.

VibeCherry / CherryOS

Learn

Templates, personas, and OS-level execution that apply styles to GlyphIR-driven actions.

Questions

Frequently asked questions

Contact

Get involved

EarthCloud is for teams who care about cross-language collaboration, structured meaning, and real trails—not just prompts and logs. If that's you, we'd love to hear from you.

Who should talk to us?

  • •Model providers building cross-language systems
  • •Platforms integrating AI-native workflows
  • •Standards folks shaping the future of AI protocols
  • •Teams building at the frontier of structured meaning

Contact: info@earthcloud.co