SPID Protocol Specification
The SPID Protocol defines a structured, open standard for publishing, discovering, and interacting with Smart Packets via unique identifiers (PulseIDs).
This spec outlines how systems store, resolve, and route Smart Packets using the SPID identity and registry system.
1. Resolution Overview
Smart Packets are resolved by querying a SPID-compatible resolver using a PulseID or intent tag.
Input:
{
"query": "spid:creator:elena-podcast"
}
Output:
{
"packets": [
{
"id": "spid:creator:elena-podcast:intro",
"voice": "...",
"transcript": "...",
"intent": "intro/show",
"cta": [ ... ],
"meta": { ... }
}
]
}
The SPID Resolver MUST support:
Exact ID resolution (e.g. spid:brand:xyz-support)
Intent-based filtering (e.g. support/faq)
Tag- or metadata-based fallback
2. Identity Registry
Each PulseID maps to a JSON record that includes:
{
"id": "spid:brand:acme",
"displayName": "Acme Corp",
"inboxUrl": "https://api.acme.com/inbox",
"packetsUrl": "https://cdn.acme.com/spid/packets.json",
"metadata": {
"verified": true,
"created": "2025-05-28",
"tags": ["insurance", "mortgage", "quote"]
}
}
This allows systems to:
Fetch all published Smart Packets for a given identity
Route new voice messages to inboxes or bots
Determine if the source is verified
3. Trust & Security
SPID-compliant systems SHOULD:
Sign packets with a verifiable token or public key
Verify authorship before displaying packets to users
Respect expiration metadata (expires field)
Optional:
Use DNS-based TXT records to validate ownership of spid:domain:* IDs
Include HTTPS-based signatures for secure handoff between agents
4. Intent Structure
Intent strings follow a predictable path-based format:
[category]/[action] or [topic]/[context]
Examples:
insurance/quote
onboarding/welcome
product/demo
AI systems can use these intents for retrieval, clustering, or handoff logic.
5. Compliance Criteria
To be SPID-compliant, a system must:
Serve or retrieve packets matching the Smart Packet Format
Accept SPID PulseID queries via API or registry
Respect routing logic based on intent and packet metadata
Support voice + transcript pairing
Limit CTAs to defined schema types
Future Extensions
Planned extensions include:
Cross-agent packet routing
Distributed registry models (using DNS or DHT)
Live voice-to-Smart Packet encoding
Permissioned inboxes and async threads
Schema.org and JSON-LD integration for semantic indexing
The SPID Protocol turns knowledge into a retrievable, trusted, voice-first layer of the web — optimized for AI, but owned by people.
---