dna-sdk 0.4.0

DNA — Domain Notation of Anything: microkernel + extensions for declarative agent notation

dna-sdk (Python)

Python SDK for DNA — Domain Notation of Anything: a microkernel + extensions runtime for declarative agent notation. See the repository README for the thesis, the architecture and the full Kind catalog.

Install

pip install dna-sdk        # or: uv add dna-sdk

Pre-release / exact-pin alternative — from the repo:

cd packages/sdk-py
uv sync            # or: pip install -e .

Optional extras: postgres, sqlite, sql (SqlAlchemySource — one adapter for both SQL dialects, same tables; see docs/PORT-CONTRACT.md § "Using the SQLAlchemy adapter"), redis, s3, tools (langchain @dna_tool decorator), safety-ml (PII/toxicity models), all, dev.

Minimal example

from dna import Kernel

# Scan a scope (directory of YAML/Markdown manifests under .dna/)
mi = Kernel.quick("hello-genome", base_dir="examples/hello-genome/.dna")

# Every document is identified by (apiVersion, kind, name)
for d in mi.documents:
    print(d.api_version, d.kind, d.name)

# Compose agent + soul + skills + guardrails into one system prompt
print(mi.build_prompt(agent="greeter"))

Runnable version: examples/hello-genome/run.py.

Layout

dna/
├── kernel/       # Kernel (mediator over 5 ports), Document, ManifestInstance
├── adapters/     # filesystem (core); sqlite/postgres/sqlalchemy_/redis/s3 via extras
├── extensions/   # helix (core Kinds) + market formats + governance
├── sync/         # lockfile + document hashing
└── safety/       # safety pipeline (optional ML extras)

Tests

uv run pytest tests/ -v

The suite includes the market-fidelity conformance tests (tests/test_market_conformance.py) and the Py↔TS parity fixtures shared with the TypeScript twin.