API Reference

Read a GDS file and return a Library.

Rosette emits a UserWarning when the source contains records that its format-neutral model cannot preserve, such as element properties, BOX, or NODE. Re-exporting such a library omits those records.

lib = read_gds("input.gds")
print("roots:", [cell.name for cell in lib.roots()])
for cell in lib.cells():
    print(cell.name)

Write a Cell or Library to a GDS file.

When writing a Cell built using Instance references (via cell.at()), child cells are automatically collected. A build summary is printed to stderr by default.

Output uses a 1 nm database grid. Export raises ValueError instead of silently collapsing geometry that is not representable on that grid. Library names and text must be ASCII without embedded NUL characters; structure names use the GDS Release 6 alphabet and maximum length of 32 characters.

# Auto-collection (recommended):
top = Cell("top")
top.add_ref(gc_cell.at(0, 0))
write_gds("output.gds", top)

# Suppress output for batch processing:
write_gds("output.gds", top, quiet=True)

Calculate the transform that aligns a component port with a target port, with their directions facing each other.

child = Cell("child")
child_port = Port("in", Point(0, 0), Vector2(-1, 0), width=0.5)
target_port = Port("out", Point(10, 0), Vector2(1, 0), width=0.5)
transform = connect_transform(child_port, target_port)
instance = Instance(child, transform)

Load layer definitions from rosette.toml.

Reads the [layers] section which maps semantic names to GDS layer numbers with optional display properties (color, fill, opacity).

layers = load_layer_map()
layers.silicon.layer   # Layer(1, 0)
layers.silicon.color   # "#ff69b4"

Load DRC rules from rosette.toml.

Searches for rosette.toml in the current directory and parent directories. Rules are defined per-layer in the [drc.layers] section (supporting min_width, min_spacing, min_area, angles, min_edge_length, max_width, acute_angle, snap_to_grid, density, no_self_intersection, and no_overlap), and inter-layer rules in the [[drc.rules]] array. See the project configuration reference for the complete schema.

An optional top-level [drc].warning_margin knob downgrades near-threshold numeric violations to warnings (see DrcRules.warning_margin).

Layer keys can use semantic names from the [layers] section (e.g. [drc.layers.silicon]) or the traditional "number/datatype" format (e.g. [drc.layers."1/0"]). Inter-layer rules also accept semantic names for layer1, layer2, inner, and outer fields.

rules = load_drc_rules()
result = run_drc(cell, rules)

Run DRC on a cell.

from rosette import BBox, Point
from rosette.drc import DrcPolicy, load_drc_rules, run_drc

rules = load_drc_rules()
policy = DrcPolicy()
policy.skip_cell("trusted_pdk_cell")
policy.waive_region("taper", BBox(Point(9, -1), Point(11, 1)))

result = run_drc(cell, rules, policy=policy)
if result.passed:
    print("DRC passed!")
else:
    for v in result.violations:
        print(f"  {v.message}")

Load design checks config from rosette.toml.

Reads the [checks] section. If absent, returns a ChecksConfig with sensible defaults.

config = load_checks_config()
result = run_checks(cell, config)

Run design checks on a cell.

Runs all design checks: connectivity (unconnected ports, width/angle mismatch) and bend radius (below minimum, auto-reduced). Ports on the top-level cell are treated as external I/O and are not flagged as unconnected.

When called with a Cell built using Instance references, child cells are automatically collected into a Library for hierarchy resolution. Bend-radius checks read route diagnostics retained privately alongside Cells created by Route.to_cell(); they do not inspect generic Cell metadata.

config = ChecksConfig(min_bend_radius=5.0)
result = run_checks(cell, config)
if not result.passed:
    for v in result.violations:
        print(f"  {v.message}")

Load DFM configuration from rosette.toml.

Reads the [dfm] section which configures the virtual nanofabrication prediction tool. Layer references in the layers list and [dfm.layer.*] overrides accept semantic names from the [layers] section (e.g. "silicon") or the traditional "number/datatype" format (e.g. "1/0").

Returns None when rosette.toml has no [dfm] section, so callers can treat "DFM not configured" as a graceful skip (matching how [drc] and [checks] degrade) rather than an error. Only genuinely invalid DFM settings raise.

loaded = load_dfm_config()
if loaded is not None:
    config, model, layers = loaded
    result = run_dfm(cell, layers=layers, model=model, config=config)

Run DFM prediction on a cell.

Rasterizes each specified layer, applies the fabrication prediction model, and extracts contour polygons representing the predicted fabricated geometry.

result = run_dfm(cell, layers=[Layer(1, 0)])
for lp in result.layers:
    print(f"  Layer {lp.layer}: {lp.input_polygon_count} -> {lp.predicted_polygon_count}")

Generate points along a circular arc.

Render a Cell or Library to a PNG image, returning a RenderResult with the image bytes and world↔pixel transform metadata.