Editor Setup

pyconveyor ships a JSONSchema for the pipeline YAML format. Pointing your editor at it gives you inline autocomplete, field validation, and hover documentation while writing pipelines.

VS Code

Automatic setup via pyconveyor init

pyconveyor init generates .vscode/settings.json and pyconveyor-schema.json automatically. If you used init, you're already set up.

Manual setup

Step 1 — Install the YAML extension by Red Hat.

Step 2 — Export the schema:

pyconveyor schema > pyconveyor-schema.json

Step 3 — Add to .vscode/settings.json:

{
  "yaml.schemas": {
    "./pyconveyor-schema.json": "pipeline.yaml"
  }
}

To apply the schema to all YAML files matching a pattern:

{
  "yaml.schemas": {
    "./pyconveyor-schema.json": ["pipeline.yaml", "pipelines/*.yaml"]
  }
}

---

JetBrains IDEs (PyCharm, IntelliJ)

Step 1 — Export the schema:

pyconveyor schema > pyconveyor-schema.json

Step 2 — Open Settings → Languages & Frameworks → Schemas and DTDs → JSON Schema Mappings.

Step 3 — Add a new mapping: - Schema file: pyconveyor-schema.json - File pattern: pipeline.yaml (or *.pipeline.yaml)

---

Neovim (with nvim-lspconfig)

Install yaml-language-server and configure it to use the schema:

require('lspconfig').yamlls.setup({
  settings = {
    yaml = {
      schemas = {
        ["./pyconveyor-schema.json"] = "pipeline.yaml",
      },
    },
  },
})

---

Keeping the schema up to date

Re-run pyconveyor schema > pyconveyor-schema.json after upgrading pyconveyor to pick up new fields added in the new version. The schema reflects the exact version of pyconveyor installed in your environment.

---

What the schema covers

The JSONSchema validates:

• Top-level keys (models, parsers, steps)
• All model block fields with their types and defaults
• All step fields with their types, enum values, and defaults
• Step type (llm, transform, io, validate, parallel, condition)
• on_error enum values (raise, continue, skip_remaining)
• retry_on array items (schema, parse, timeout, http_error, rate_limit)

It does not validate cross-references (e.g., whether a step's model: exists in models:). Those checks happen at PipelineRunner load time and are reported with line numbers via pyconveyor validate.