🐳 CCP Runtime Architecture

This page explains how the CCP platform executes methods inside containerized runtimes using a deploy/execute approach.

🧱 Base Architecture

Each method runs in an isolated container (e.g. Docker) provided by a runtime (ccpimage).

CCP mounts a temporary execution volume at:

/ccp_data/

This folder is used for:

  • Storing downloaded input files

  • Writing outputs and logs

  • Temporary data exchange

πŸ“œ Deploy and Execute Scripts

Each execution involves two sequential scripts:

πŸ›  Deploy Script

Runs first.
Used to prepare the environment:

  • Clones code repositories

  • Installs dependencies

  • Downloads data

  • Sets permissions

πŸš€ Execute Script

Runs after deploy.
Used to run the actual algorithm:

  • Executes your script or binary

  • Uses data in /ccp_data/

  • Writes outputs to /ccp_data/

Both scripts are defined in the method’s method.json.

πŸ§ͺ Runtime Example

Given this configuration:

"ccpimage": "python:3.9",
"deploy-script": [
  "git clone https://github.com/user/tool.git /src",
  "pip install -r /src/requirements.txt",
  "wget {{input_url}} -O /ccp_data/input.csv"
],
"execute-script": [
  "python3 /src/main.py --input /ccp_data/input.csv --output /ccp_data/result.json"
]

CCP will:

  1. Start a python:3.9 container

  2. Run the deploy commands

  3. Run the execution commands

  4. Collect all /ccp_data/* files as output

πŸ” Credentials Handling

Secret values (e.g., API keys) can be passed using secure inputs (format: Secret).
You can use them via placeholders like {{api_key}} inside the scripts.

🌍 Supported Runtimes

CCP currently supports:

  • Docker images (bash, python, jupyter, gdal, r-base, etc.)

  • Custom base images via ccpimage parameter

  • Experimental: pre-bundled consolidated runtimes

🧩 Inputs and Placeholders

Inputs defined in the method form (e.g., text, file, URL) are passed to the scripts using placeholders like:

--input {{input_file}} --option {{threshold}}

At runtime, these placeholders are replaced with the actual values provided by the user.

Files from inputs (workspace, URLs, uploads) are automatically downloaded or decoded into /ccp_data/.

πŸ“‘ Metadata and Provenance

During execution, CCP generates a set of metadata files:

  • metadata_request.json: user inputs and configuration

  • metadata_jobStatus.json: status and timestamps

  • metadata_provo.xml: provenance metadata (agents, entities, activities)

These help track:

  • Who ran what

  • When and with which parameters

  • How outputs were generated

You can inspect these files from the execution results view or download them for auditing and reproducibility.

πŸ”— Related Pages