Feature Request: Structured inbound/outbound message schema beyond plain text

[yidianyiko]

Summary

ClawScale currently appears to route messages primarily as plain text content. For many production messaging use cases, a gateway needs a richer canonical message schema.

Problem

A text-only abstraction is too narrow for modern IM gateway use cases.

Important message features usually include:

• images
• audio / voice notes
• files / documents
• quoted replies / references
• group metadata
• thread metadata
• sender metadata
• platform-specific delivery fields

Without a structured canonical schema, each adapter ends up handling message features ad hoc, and external backends cannot reliably consume or produce richer message types.

Proposed solution

Introduce a canonical message envelope for both inbound and outbound flows.

Example top-level fields:

• id
• channelId
• conversationId
• sender
• recipient
• chatType (direct / group)
• threadId
• replyTo
• parts[]
• metadata

Where parts[] can support:

• text
• image
• audio
• video
• file
• location
• contact
• sticker
• system

Each part should include normalized metadata plus adapter-specific raw metadata where necessary.

Expected behavior

• Adapters map provider-native messages into a canonical schema
• Backends can receive structured messages, not only flattened text
• Outbound sends can specify structured content
• Quoted replies and media attachments are preserved when possible
• Group and thread metadata are available consistently

Why this matters

A richer schema makes ClawScale much more viable as a long-term gateway layer, especially for systems that need multimodal chat, quoted replies, and reliable cross-channel behavior.

Acceptance criteria

• A documented canonical message schema exists
• At least text, image, audio, file, and reply/reference are supported
• Inbound adapters emit structured payloads
• Outbound APIs can accept structured payloads
• Unsupported channel features degrade predictably and are documented