LVGPT

LVGPT explores generating LabVIEW block diagrams from LLM-produced code. The early pipeline transpiled Go, but the repository now focuses on a Python-only graph builder that turns AST XML into LabVIEW VI XML. The Go AST tooling has been removed to keep the project focused on a single implementation.

Repository layout

• python/lvgpt: package entry points for AST-to-graph conversion
• python/lvgpt/lvgraph: LabVIEW graph data structures and nodes
• LabVIEW/: example assets, XML fixtures, and helper data

Quick start

1. Install the package from the repo root:
   • pip install -e .
2. Run the CLI on an AST XML file:
   • lvgpt-build path/to/ast.xml -o output.vi.xml

Examples

• examples/simple_ast.xml includes a tiny AST that prints a + b.
• examples/simple_ast.vi.xml is the expected LabVIEW XML output for that AST.

Graph IR (LLM output)

The preferred LLM interface is a small JSON/YAML graph IR that maps directly to nodes and wires.

Example:

{
  "nodes": [
    {"id": "a", "kind": "NumericConstant", "params": {"name": "a", "numType": "Long", "value": "1"}},
    {"id": "b", "kind": "NumericConstant", "params": {"name": "b", "numType": "Long", "value": "2"}},
    {"id": "add", "kind": "Add"},
    {"id": "out", "kind": "NumericIndicator", "params": {"name": "output", "numType": "Long"}}
  ],
  "edges": [
    {"from": "a.a", "to": "add.x"},
    {"from": "b.b", "to": "add.y"},
    {"from": "add.x+y", "to": "out.output"}
  ]
}

You can build a graph from the IR with:

• from lvgpt.graph_ir import parse_graph_ir, build_graph_from_spec Edges can also be written with explicit fields:

{"from_node": "add", "from_terminal": "x+y", "to_node": "out", "to_terminal": "output"}

YAML support requires PyYAML to be installed, and schema validation uses jsonschema when available.

CLI:

• lvgpt-build-ir path/to/graph_ir.json -o output.vi.xml

Node catalog

The node catalog is generated at runtime from available LVNode subclasses:

• from lvgpt.catalog import build_node_catalog

Testing

1. Install test dependencies:
   • pip install -e .[dev]
2. Run pytest (with the repo Python package on the path):
   • PYTHONPATH=python pytest

Status

This is an experimental toolchain. Expect the AST coverage and node vocabulary to evolve as more test cases are added.