Skip to main content

An opinionated, backend-agnostic runtime for agents that work in production — deterministic, auditable, injection-resistant. Import as `zu`.

Project description

Zu

An opinionated, backend-agnostic runtime for agents that work in production — deterministic, auditable, and injection-resistant by construction.

New here? One line:

pip install 'zu-runtime[all]'       # everything: web tools, both model SDKs, server, Docker, MCP

Prefer a lean install? pip install zu-runtime gives you import zu, the zu command, the web tools (http_fetch/html_parse/render_dom), the model-provider adapters, detectors, validators, and a SQLite event sink. Add the heavy/situational bits as extras:

pip install 'zu-runtime[anthropic]' # + the Anthropic SDK to call a real model (also: [openai])
pip install 'zu-runtime[serve]'     # + the HTTP server (zu serve)
pip install 'zu-runtime[docker]'    # + the Docker sandbox (tier-2 browser)
pip install 'zu-runtime[mcp]'       # + the MCP server (zu mcp)

The zu-* packages are also standalone on PyPI, but you rarely install them individually — that's for plugin authors depending on just zu-core.

Embed it

import zu

result = zu.run(
    {"query": "Extract the product name and price.",
     "target": "https://example.com/product/123",
     "output_schema": {"type": "object",
                       "properties": {"name": {"type": "string"}, "price": {"type": "string"}},
                       "required": ["name", "price"]}},
    config={"provider": {"name": "anthropic", "model": "claude-sonnet-4-6",
                         "api_key_env": "ANTHROPIC_API_KEY"},
            "plugins": {"tools": ["http_fetch", "html_parse", "render_dom"],
                        "detectors": ["empty", "error", "js-shell", "bot-wall"],
                        "validators": ["schema", "grounding"]}},
)
print(result.status, result.value)

Swapping the model is a one-line edit to the provider block — Anthropic, OpenRouter, OpenAI, or a local model (Ollama / vLLM) — because the runtime only ever speaks to a ModelProvider port. Credentials are named by environment variable (api_key_env), never passed in code or config.

Run it from the command line, or as a service

zu run agent.yaml             # one-shot
zu run agent.yaml --every 5m  # scheduled worker
zu serve -c agent.yaml                     # HTTP: POST /run  (needs the [serve] extra)

What it is

A small, stable core (the loop, registry, contracts, event bus) surrounded by six swappable ports. Every capability that can vary is a plugin behind a port, so the production system is reached by adding adapters — never by reopening the core. Full source, architecture, and examples: https://github.com/k3-mt/zu

Apache-2.0.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

zu_runtime-0.3.1.tar.gz (9.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

zu_runtime-0.3.1-py3-none-any.whl (7.3 kB view details)

Uploaded Python 3

File details

Details for the file zu_runtime-0.3.1.tar.gz.

File metadata

  • Download URL: zu_runtime-0.3.1.tar.gz
  • Upload date:
  • Size: 9.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.13

File hashes

Hashes for zu_runtime-0.3.1.tar.gz
Algorithm Hash digest
SHA256 dfd6b571175ac5e92d979538120fd0abf26c02d3afd17ab1e92be4553233cd6a
MD5 48497ab4af45f727a4692a8661c09d25
BLAKE2b-256 69f3c5280232f7c5a70ce389e99bfd0f5aa394c07ee9bde3dce3910675b09ec0

See more details on using hashes here.

File details

Details for the file zu_runtime-0.3.1-py3-none-any.whl.

File metadata

  • Download URL: zu_runtime-0.3.1-py3-none-any.whl
  • Upload date:
  • Size: 7.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.13

File hashes

Hashes for zu_runtime-0.3.1-py3-none-any.whl
Algorithm Hash digest
SHA256 bef1aa7eba107b6db01de3d6e502dc8fa4faf16541c363d6b739eebf8df26c39
MD5 3efe738e0cbace3aa74e987d66be540a
BLAKE2b-256 f76de6873bfc0eeb71e5bff71b9a6de44886b7bf3944739182ec40f6c513562f

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page