orbynt-protocol 1.0.0

OMP - portable AI memory passport

Orbynt Memory Protocol (OMP)

OMP (Orbynt Memory Protocol) is an open standard for portable AI memory.

It lets a user save identity, preferences, goals, values, skills, interests, relationships, and memories into one .orb file. That file can be read by another AI system, app, tool, or library.

Simple idea:

User memory -> .orb file -> another AI system

What Is OMP?

OMP (Orbynt Memory Protocol) is an open protocol and binary file format for portable AI memory.

OMP = Protocol
.orb = File Format
orbynt-protocol = Python Package
omp = Python Import

The official Python reference implementation is distributed as the orbynt-protocol package.

An Orbynt Passport stored inside a .orb file can carry useful context such as:

• who the user is
• how the user wants answers
• what the user is trying to do
• what the user cares about
• what the user knows
• important facts, notes, projects, and experiences

Without OMP, every AI system starts from zero. With OMP, a user can bring structured memory with them.

Why OMP Is Useful

OMP provides a portable, readable, and structured standard for AI memory.

Developers can use OMP to:

• create .orb memory files with Python
• read .orb files back into apps
• turn memory into AI prompt text
• search memory records locally
• export memory to JSON or Markdown
• access bundled protobuf schemas
• build tools around one shared memory format

File Diagram

Inside the file:

.orb
|-- Header
|   |-- magic bytes: ORB1
|   |-- protocol version
|   |-- flags
|   `-- chunk table
|
|-- Chunks
|   |-- protocol metadata
|   |-- identity
|   |-- profile
|   |-- preferences
|   |-- goals
|   |-- values
|   |-- skills
|   |-- interests
|   |-- relationships
|   |-- memories
|   |-- graph
|   |-- provenance
|   `-- export policy
|
`-- Footer
    |-- total file length
    |-- SHA-256 checksum
    `-- footer magic: ORBF

Install

pip install orbynt-protocol

The package installs the official Python implementation of OMP.

Quick Start

import omp

omp.write(
    "user_memory.orb",
    identity={"display_name": "User"},
    preferences={"communication.style": "clear and simple"},
    goals=[{"title": "Learn AI tools"}],
    values=[{"statement": "Build useful things"}],
    memories=[
        {
            "memory_type": "fact",
            "summary": "The user is learning OMP",
            "full_content": "The user wants to create portable AI memory files.",
            "tags": ["omp", "ai", "memory"],
            "importance": 1.0,
        }
    ],
)

passport = omp.read("user_memory.orb")
print(passport.identity.display_name)
print(passport.memories[0].summary)

That creates a real .orb file.

Full Example

import omp

passport = omp.write(
    "complete_memory.orb",
    identity={
        "display_name": "User",
        "username": "user",
        "primary_language": "English",
        "timezone": "UTC",
    },
    profile={
        "biography": "A person building with AI tools.",
        "occupation": "Builder",
        "organization": "Independent",
        "website_links": ["https://example.com"],
    },
    preferences={
        "communication.style": "concise but complete",
        "formatting": "markdown",
        "technical_depth": "high",
    },
    goals=[
        {
            "title": "Build a memory tool",
            "description": "Create a useful AI memory workflow.",
            "priority": "high",
            "status": "active",
        }
    ],
    values=[
        {
            "statement": "Build ethically",
            "priority": "high",
            "category": "work",
        }
    ],
    skills=[
        {
            "name": "Python",
            "description": "Writing scripts and tools",
            "level": "intermediate",
            "years_experience": 2,
        }
    ],
    interests=[
        {"name": "AI", "description": "Practical AI systems"},
        {"name": "Protocols", "description": "Portable data formats"},
    ],
    relationships=[
        {
            "person_name": "Team",
            "relationship_type": "collaborator",
            "notes": "People working on the same project.",
        }
    ],
    memories=[
        {
            "memory_type": "project",
            "summary": "The user is testing OMP",
            "full_content": "The user is creating a .orb file and reading it back.",
            "tags": ["project", "omp"],
            "confidence": 1.0,
            "importance": 0.9,
        }
    ],
    compress=True,
)

What Is Stored In A Passport?

OMP stores memory in a clean passport structure:

Orbynt Passport
1. Protocol Metadata
2. Identity
3. Profile
4. Preferences
5. Goals
6. Values
7. Skills
8. Interests
9. Relationships
10. Memories
11. Memory Graph
12. Embeddings
13. Adapters
14. Provenance
15. Export Policy
16. Integrity Metadata

In simple words:

• Identity means who the user is.
• Preferences means how the AI should respond.
• Goals means what the user wants to achieve.
• Values means what matters to the user.
• Skills means what the user can do.
• Interests means topics the user cares about.
• Relationships means important people, groups, or roles.
• Memories means facts, notes, projects, and experiences.
• Provenance means where information came from.
• Integrity Metadata means checksum and file integrity information.

Memory Types

Each memory has a memory_type.

fact
goal
preference
value
skill
interest
relationship
event
project
note
conversation
episodic

Memory Record Fields

Each memory is saved as a MemoryRecord:

memory_id
memory_type
summary
full_content
namespace_id
tags
confidence
importance
created_at
updated_at
provenance
related_memory_ids
embedding_refs
graph_node_refs
custom_attributes

Read A .orb File

import omp

passport = omp.read("complete_memory.orb")

print(passport.protocol_metadata.protocol_name)
print(passport.identity.display_name)

for memory in passport.memories:
    print(memory.memory_type, memory.summary)

Open .orb files always read as:

passport.verified  # False
passport.tier      # "open"

Turn Memory Into An AI Prompt

import omp

passport = omp.read("complete_memory.orb")

generic_prompt = omp.inject(passport)
provider_prompt = omp.inject(passport, target="provider")
custom_prompt = omp.inject(passport, target="custom")

The generated prompt includes a trust block saying the memory is open and unverified.

Search Memories

import omp

passport = omp.read("complete_memory.orb")

matches = omp.query(
    passport,
    text="AI memory",
    kind="project",
    tags=["omp"],
    limit=5,
    min_confidence=0.7,
)

for match in matches:
    print(match.score, match.memory.summary)

The query engine runs locally. No server is required.

Export Memory

import omp

passport = omp.read("complete_memory.orb")

data = omp.export(passport, format="dict")
json_text = omp.export(passport, format="json")
markdown_text = omp.export(passport, format="markdown")

omp.export(passport, format="json", path="memory.json")
omp.export(passport, format="markdown", path="memory.md")

Merge Memory Files

import omp

old_passport = omp.read("old.orb")
new_passport = omp.read("new.orb")

merged = omp.merge(old_passport, new_passport, strategy="latest")
omp.write("merged.orb", passport=merged)

Merge strategies:

• latest
• union
• source_a
• source_b

Inspect A File

import omp

omp.inspect("complete_memory.orb")

Verify A File

omp verify complete_memory.orb

Open .orb files use checksum verification. They do not include issuer signatures.

Proto Files

The official Python implementation (orbynt-protocol) includes bundled protobuf schema files for OMP.

List proto files:

import omp

print(omp.list_proto_files())

Read one proto file:

memory_proto = omp.get_proto_source("memory.proto")
passport_proto = omp.get_proto_source("passport")

Get all proto files:

all_sources = omp.get_all_proto_sources()

Get one schema text block:

bundle = omp.get_proto_bundle()

Get an AI-friendly schema block:

context = omp.get_ai_schema_context()

Export all proto files to one text file:

omp.export_proto_bundle("omp-protos.txt")

Import generated protobuf modules:

from omp.proto import memory_pb2, passport_pb2

memory = memory_pb2.MemoryRecord()
memory.summary = "Created directly with protobuf"

passport = passport_pb2.OrbPassport()
passport.memories.append(memory)

Proto namespace:

package orbynt.memory.protocol.v1;

Python API (orbynt-protocol)

omp.write(path, ...)

Creates a .orb file and returns the generated passport object.

Common inputs:

• identity
• profile
• preferences
• goals
• values
• skills
• interests
• relationships
• memories
• adapters
• provenance
• export_policy
• compress
• version

omp.read(path)

Reads a .orb file.

omp.inject(passport, target="generic")

Converts an Orbynt Passport into AI-readable context text.

Targets:

• generic — Portable AI context format.
• provider — Provider-optimized context format.
• custom — User-defined injection format.

omp.query(passport, ...)

Searches memories.

Inputs:

• text
• kind
• tags
• limit
• min_confidence

omp.inspect(path)

Prints a readable summary.

omp.merge(passport_a, passport_b, strategy="latest")

Combines two passports.

omp.export(passport, format, path=None)

Exports to:

• dict
• json
• markdown

omp.list_proto_files()

Lists bundled .proto files.

omp.get_proto_source(name)

Returns one .proto file.

omp.get_all_proto_sources()

Returns all .proto files.

omp.get_proto_bundle()

Returns all .proto files as one text bundle.

omp.get_ai_schema_context()

Returns schema text that can be given to an AI system.

omp.export_proto_bundle(path)

Writes the proto bundle to a file.

Constants

omp.__version__
omp.PROTOCOL_VERSION

Exceptions

omp.OMPError
omp.InvalidOrbFile
omp.VersionMismatch
omp.UnsupportedFeature

CLI

Create:

omp write --name "User" --out memory.orb \
  --memory "kind:fact|summary:Learning OMP|tags:ai,memory" \
  --pref "communication.style=simple"

Read:

omp read memory.orb
omp read memory.orb --format json
omp read memory.orb --format markdown --out memory.md

Inspect:

omp inspect memory.orb

Query:

omp query memory.orb --text "memory" --limit 5

Merge:

omp merge a.orb b.orb --out merged.orb --strategy latest

Export:

omp export memory.orb --format json --out memory.json

Verify:

omp verify memory.orb

Proto:

omp proto list
omp proto show memory
omp proto bundle --out omp-protos.txt

Version:

omp version

.orb Files

A .orb file is a binary container, not a raw protobuf dump.

The header identifies the file and lists the chunks. The chunks hold the passport data. The footer stores the file length and SHA-256 checksum so a reader can detect corruption.

Trust And Safety

Current .orb files are open and unverified.

That means:

• anyone can create them
• anyone can read them
• checksum validation can detect file corruption
• there is no encryption
• there is no issuer signature
• the file does not prove who created it

AI systems should treat open .orb files as user-provided context.

Package Info

Package name:

orbynt-protocol

Install:

pip install orbynt-protocol

Import:

import omp

File extension:

.orb

MIME type:

application/x-omp-memory

Tags

ai
memory
protocol
llm
passport
portable-memory
protobuf
binary-format
open-protocol
orb

License

Apache-2.0.

Copyright © Orbynt