Python SDK for the Neural Protocol Suite (NPS)
Project description
English | 中文版
NPS Python SDK (nps-lib)
Python client library for the Neural Protocol Suite (NPS) — a complete internet protocol stack designed for AI agents and models.
PyPI package: nps-lib | Python namespace: nps_sdk
Status
v1.0.0-alpha.2 — Phase 1 / Phase 2 synchronized alpha release
Covers all five protocols — NCP + NWP + NIP + NDP + NOP — frame definitions, async client, and Ed25519 identity management.
Requirements
- Python 3.11+
- Dependencies:
msgpack,httpx,cryptography
Installation
pip install nps-lib
For development:
pip install "nps-lib[dev]"
Modules
| Module | Description |
|---|---|
nps_sdk.core |
Frame header, codec (Tier-1 JSON / Tier-2 MsgPack), anchor cache, exceptions |
nps_sdk.ncp |
NCP frames: AnchorFrame, DiffFrame, StreamFrame, CapsFrame, HelloFrame, ErrorFrame |
nps_sdk.nwp |
NWP frames: QueryFrame, ActionFrame; async NwpClient |
nps_sdk.nip |
NIP frames: IdentFrame, TrustFrame, RevokeFrame; NipIdentity (Ed25519 key management) |
nps_sdk.ndp |
NDP frames: AnnounceFrame, ResolveFrame, GraphFrame; in-memory registry + validator |
nps_sdk.nop |
NOP frames: TaskFrame, DelegateFrame, SyncFrame, AlignStreamFrame; async NopClient |
Quick Start
Encoding / Decoding NCP Frames
from nps_sdk.core.codec import NpsFrameCodec
from nps_sdk.core.registry import FrameRegistry
from nps_sdk.ncp.frames import AnchorFrame, FrameSchema, SchemaField
registry = FrameRegistry.create_default()
codec = NpsFrameCodec(registry)
schema = FrameSchema(fields=(
SchemaField(name="id", type="uint64"),
SchemaField(name="price", type="decimal", semantic="commerce.price.usd"),
))
frame = AnchorFrame(anchor_id="sha256:...", schema=schema)
wire = codec.encode(frame) # bytes — Tier-2 MsgPack by default
result = codec.decode(wire) # → AnchorFrame
Anchor Cache (Schema Deduplication)
from nps_sdk.core.cache import AnchorFrameCache
cache = AnchorFrameCache()
anchor_id = cache.set(frame) # stores; returns canonical sha256 anchor_id
frame = cache.get_required(anchor_id)
Querying a Memory Node (async)
import asyncio
from nps_sdk.nwp import NwpClient, QueryFrame
async def main():
async with NwpClient("https://node.example.com") as client:
caps = await client.query(
QueryFrame(anchor_ref="sha256:...", limit=50)
)
print(caps.count, caps.data)
asyncio.run(main())
Invoking an Action Node (async)
from nps_sdk.nwp import NwpClient, ActionFrame
async with NwpClient("https://node.example.com") as client:
result = await client.invoke(
ActionFrame(action_id="orders.create", params={"sku": "X-101", "qty": 1})
)
NIP Identity Management
from nps_sdk.nip.identity import NipIdentity
# Generate and save an encrypted Ed25519 keypair
identity = NipIdentity.generate("ca.key", passphrase="my-secret")
# Load from file
identity = NipIdentity()
identity.load("ca.key", passphrase="my-secret")
# Sign a NIP frame payload (canonical JSON, no 'signature' field)
sig = identity.sign(ident_frame.unsigned_dict())
# Verify
ok = NipIdentity.verify_signature(identity.pub_key_string, payload, sig)
Architecture
nps_sdk/
├── core/ # Wire primitives (FrameHeader, codec, cache, exceptions)
├── ncp/ # NCP frames (0x01–0x0F)
├── nwp/ # NWP frames (0x10–0x1F) + async HTTP client
└── nip/ # NIP frames (0x20–0x2F) + Ed25519 identity
Frame Encoding Tiers
| Tier | Value | Description |
|---|---|---|
| Tier-1 JSON | 0x00 |
UTF-8 JSON. Development / compatibility. |
| Tier-2 MsgPack | 0x01 |
MessagePack binary. ~60% smaller. Production default. |
NWP HTTP Overlay Mode
NwpClient communicates via HTTP with Content-Type: application/x-nps-frame.
Sub-paths per operation:
| Operation | Path | Request Frame | Response Frame |
|---|---|---|---|
| Schema anchor | POST /anchor |
AnchorFrame | 204 |
| Structured query | POST /query |
QueryFrame | CapsFrame |
| Streaming query | POST /stream |
QueryFrame | StreamFrame chunks |
| Action invocation | POST /invoke |
ActionFrame | raw result or AsyncActionResponse |
Running Tests
pytest # all tests + coverage report
pytest -k test_nip # NIP tests only
Coverage target: ≥ 90 %.
License
Apache 2.0 — see LICENSE.
Copyright 2026 INNO LOTUS PTY LTD
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file nps_lib-1.0.0a2.tar.gz.
File metadata
- Download URL: nps_lib-1.0.0a2.tar.gz
- Upload date:
- Size: 85.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.14.0
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c40439f8a08e14136a2747cac1693e1e484e2be1a5a9ffa5da7239d475a422cc
|
|
| MD5 |
b909b4b1b0ca58ae1ea9b415b6af7469
|
|
| BLAKE2b-256 |
ce9cc337db3d0c7f44963c523006544eac8a59871acbb0abe6765f2531879e02
|
File details
Details for the file nps_lib-1.0.0a2-py3-none-any.whl.
File metadata
- Download URL: nps_lib-1.0.0a2-py3-none-any.whl
- Upload date:
- Size: 43.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.14.0
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
b0042d31582a13fcfabe82610eff55e739f9cf4c99c5bb3ef3d9fa619e33f246
|
|
| MD5 |
d591a5268bd726155006291eb8718852
|
|
| BLAKE2b-256 |
10127acea535a9dc7376d1dfa8ebbacaddc8d74e44c2d23a0dd8ef64b20b18e8
|