5.5 KiB
Fresh Workspace Handoff
Immediate Re-entry Instruction
Read state/map.yaml, this file, notes/vibegeometry-current-system-map.md,
and notes/vibegeometry-implementation-plan.md, then run:
npm run state:status
Restate the persisted next action before editing.
Do not continue implementation automatically from a rehydrate-only request.
Do not trust this file for the exact live HEAD. Use git for volatile truth:
git status --short --branch
git log --oneline -5
Current State
VibeGeometry is an initial scaffold. It has persistence machinery, project orientation docs, helper tools, and the first Geometry Nodes translation corpus entry, but no Blender builder library or artifact schemas yet.
The first completed translation is
docs/research/translations/shriinivas-cartesian-helper.md, backed by
examples/geometry_script/shriinivas_cartesian_helper.py.
The completed Cartesian function translation is
docs/research/translations/shriinivas-cartesian-function-graphs.md. It covers
Geometry Nodes Line, Geometry Nodes Parabola, and Geometry Nodes Circle
from cartesian.blend.
The completed pie chart translation is
docs/research/translations/shriinivas-piechart.md. It covers NodeGroup and
Pie Chart from piechart.blend.
The completed field-value translation is
docs/research/translations/shriinivas-fieldvalue-digit-at.md. It covers
Create Decimal, Create Segment, Seven Segments, Delete Segments,
Next Digit, Digit At, and the full Field Value display graph from
fieldvalue.blend. The full graph is verified on a default-route float-value
case with matching evaluated geometry.
The completed parametric/polar translation is
docs/research/translations/shriinivas-paramnpolareq.md. It covers scalar
equation helpers, Archimedes Spiral, Epicycloid, and a mirrored root spiral
from paramnpolareq.blend.
The CurveToMeshUV utility sample has started at
docs/research/translations/quellenform-curve-to-mesh-uv.md. So far it covers
the small Auto Smooth helper from the quellenform file. The main
Curve to Mesh UV group is a 126-node utility quarry and should be staged into
subcontracts rather than translated as one heroic wall.
docs/research/procedural-doctrine.md now stores the current spatial reasoning
playbook for mapping geometric ideas to Geometry Script. Doctrine entries should
explain when a tool becomes salient for a visual form, the mental move that
turns the form into graph structure, a small code sketch, and the verification
cue.
Current authoring prior: use the full Python stack. Python structures intent,
tables, loops, variants, cleanup, validation, docs, and scene/render
orchestration; Geometry Script emits inspectable Geometry Nodes groups; bpy
handles Blender scene machinery and evaluated evidence.
The repo-local Geometry Script clone has a Blender 5.1 patch for nested @tree
group reuse, a ported nodes_to_script converter, and a menu-socket default
tolerance patch needed by Blender 5 menu inputs. See
docs/research/geometry-script-fork-notes.md.
Run .\tools\setup_geometry_script_clone.ps1 if external/geometry-script is
missing in a fresh workspace.
The local toolchain branch is vibegeometry/blender-5-nodes-to-script.
The upstream PR branch is vibegeometry/blender-5-nested-tree-groups, and PR
#69 is intentionally limited to the nested-group fix:
https://github.com/carson-katri/geometry-script/pull/69.
The GameCult fork is intentionally pruned to main plus VibeGeometry-owned
tooling branches. Upstream branch clutter should not be copied forward unless a
future patch actually needs it.
The current next action is to continue the translation corpus with either the
main CurveToMeshUV utility staged into subcontracts or a scattering/instancing
sample. Keep docs/research/procedural-doctrine.md compact; put
sample-specific graph maps and verification details in
docs/research/translations/.
Important Invariants
- Target Blender 5.0+ unless the task says otherwise.
- Geometry nodes are the primary procedural geometry surface.
- Python is the authoring/orchestration layer, not a fallback after Geometry Script runs out of rope.
- Python constructs and orchestrates graphs; it should not bury the design in opaque imperative mesh generation.
- Visual target evidence must remain attached to scene briefs and render review.
- A graph is not accepted just because it runs. Render evidence decides.
- Node groups, exposed sockets, and attributes need names clear enough for future agents to inspect and modify.
- Add evidence only when it changes future belief.
Open Questions
- What exact JSON shape should the scene brief use for references, instructions, target traits, and acceptance criteria?
- What should the first node graph manifest schema require versus merely allow?
- What node-group inspection JSON shape is most useful for translating public
.blendexamples into Geometry Script? - Should the Geometry Script nested-group patch be upstreamed, or should VibeGeometry maintain a fork as its toolchain?
- Which Blender 5.0+ APIs and node construction helpers should become the first stable builder layer?
- What is the smallest useful render-review artifact?
Compaction Notes
When preparing for imminent compaction, write hot memory first to
state/scratch-compaction-<guid>.md, then run:
npm run state:prepare-compaction
Fold only durable lessons into canonical state and delete the temporary scratch-compaction file after it has served its purpose.