Use capacity info to decide on concurrency [chapter] deterministic

capacity = get_capacity("llama-3-70b") if capacity["gpu_available"] > 0: # Proceed with parallel processing ...

`

This pattern ensures that workflows scale gracefully, avoiding resource contention and reducing costs.

Document-Driven Development Pipelines

Document-driven development (DDD) is an emerging paradigm where structured documentation serves as the primary artifact driving code generation, testing, and deployment. Unlike traditional workflows that treat documentation as an afterthought, DDD places it at the center of the development lifecycle.

The Core Idea

In DDD, a document (e.g., a markdown file, JSON schema, or YAML configuration) defines the expected behavior, constraints, and interfaces of a system. Automated tools then:

1. Parse the document into a machine-readable format. 2. Generate code stubs, tests, or configurations. 3. Validate outputs against the document's constraints.

This approach reduces boilerplate, enforces consistency, and accelerates iteration cycles.

Implementing a Document-Driven Pipeline

Let's walk through a concrete example of a document-driven pipeline that processes a markdown file describing a service API.

#### Step 1: Define the Document

Assume we have a file api_spec.md that describes a REST API:

```markdown