Skip to main content
Cloacina Documentation
Toggle Dark/Light/Auto mode Toggle Dark/Light/Auto mode Toggle Dark/Light/Auto mode Back to homepage
  1. Cloacina
  2. /
  3. Engine & Primitives
  4. /
  5. Computation Graphs
  6. /
  7. How-to Guides
  8. /
  9. Package a Python Computation Graph

Package a Python Computation Graph

Package a Python Computation Graph

A Python-authored computation graph packages and deploys exactly like a Python workflow: a package.toml plus a module tree under workflow/. The server loads it through the same path as a Rust-authored graph. This how-to packages the market_maker graph end to end. (For the workflow analog, see Packaging Python Workflows.)

Prerequisites

Step 1: Lay out the package 

market-maker/
├── package.toml
└── workflow/
    └── market_maker/
        ├── __init__.py
        └── graph.py        # @cloaca.reactor + ComputationGraphBuilder + @cloaca.node

The module tree must live under workflow/. graph.py holds the reactor, the accumulators, the ComputationGraphBuilder, and the @cloaca.node functions (see the Topology Dict Schema).

Step 2: Write package.toml

A computation-graph package sets graph_name (this is what marks it a CG rather than a workflow) and entry_module (the dotted path, relative to workflow/, the loader imports):

[package]
name = "market-maker"
version = "0.1.0"
interface = "cloacina-workflow-plugin"
interface_version = 1
extension = "cloacina"

[metadata]
language = "python"
graph_name = "market_maker"
entry_module = "market_maker.graph"
reaction_mode = "when_any"     # or "when_all"
input_strategy = "latest"      # or "sequential"
requires_python = ">=3.10"

Both are optional (defaults: reaction_mode = "when_any", input_strategy = "latest"):

  • reaction_mode should match the mode= on your @cloaca.reactor (when_any = fire when any accumulator has new data; when_all = wait for all).
  • input_strategy controls how queued boundaries are consumed: latest collapses to the newest boundary (one execution on the most recent data); sequential runs the graph once per boundary in order.

Importing entry_module runs your decorators, which register the graph.

Step 3: Validate and pack 

cloacinactl package validate market-maker
cloacinactl package pack market-maker --out market-maker-0.1.0.cloacina

validate checks the layout and the closed [metadata] schema; pack writes the .cloacina archive. (No maturin/compile step — a Python package ships source.)

Step 4: Upload and verify 

cloacinactl --tenant my_tenant package upload market-maker-0.1.0.cloacina
cloacinactl --tenant my_tenant graph list

Once the server loads it, the graph appears in graph list and its health is queryable via cloacinactl graph status market_maker.

See also

gdoc_arrow_left_alt How-to Guides Choosing and Using Accumulator Types gdoc_arrow_right_alt