An agent runtime for geometry — a native Rust B-Rep kernel with a semantic surface AI can query, reason about, and act on.
Native B-Rep kernel: datum system, topology, tessellated viewport — running through the same REST/WebSocket bridge an external agent would use.
Roshera is an agent runtime for geometry. The product is the kernel and the bridge it exposes: a native Rust B-Rep engine whose primitives, topology, and operations carry enough semantic structure for an LLM to query, reason about, and drive directly. Humans orchestrate; agents execute. The React/Three.js frontend that ships in this repo is one client of that runtime — it talks to the kernel over REST + WebSocket the same way an external agent would.
The differentiator is the readable surface: geometry is not just triangles, it is a queryable model with named features, intent, and history. Many things work, many things don't — see Status for what's actually usable today.
| Dark Mode | Light Mode |
|---|---|
 |
 |
roshera-backend/
geometry-engine/ B-Rep kernel: NURBS math, primitives, topology, operations, tessellation
ai-integration/ LLM providers (Claude, OpenAI) + vision pipeline + smart routing
timeline-engine/ Event-sourced design history with branching
session-manager/ Multi-user collaboration with RBAC
export-engine/ STL, OBJ, encrypted .ros (AES-256-GCM), STEP (in progress)
rag-engine/ Vamana-indexed retrieval for design knowledge
api-server/ Axum REST + WebSocket API
shared-types/ Common type definitions
roshera-app/ React + Three.js + TypeScript browser client
The kernel is the product. The viewport you see in the screenshots is one client of it, written to prove the bridge works at human latency. The same REST + WebSocket surface drives agents.
Where this is going: agents that pick faces, fillet edges, run sections, and walk a part the way a designer would — by reasoning over the readable surface (named features, persistent IDs, the timeline of events that produced the current state), not by parsing triangle soup. A CLI client and an MCP server are next on the list, then embeddings and retrieval over the timeline so an agent can answer questions like "how did this corner get to be 4mm" without re-deriving the model from scratch.
What follows is honest about what's tested versus what's implemented but rough. Measured perf numbers are in Performance.
| Layer | Component | Status |
|---|---|---|
| Math | Vector3, Matrix4, Quaternion | Tested |
| B-spline, NURBS evaluation | Tested; perf above budget (see Performance) | |
| Primitives | Box, Sphere, Cylinder, Cone, Torus | B-Rep topology with Euler validation |
| Topology | Manifold detection, adjacency | Tested |
| Tessellation | Per-surface dispatch, adaptive subdivision | Works for analytic surfaces; extruded curved profiles still have watertightness issues |
| Operations | Extrude (draft, taper, twist) | Implemented; side-face seam bugs being chased |
| Boolean (union, intersect, difference) | Implemented; edge cases (e.g. drill-through-cube) still fail in places | |
| Fillet (constant-radius) | Implemented, lightly tested | |
| Chamfer, Offset, Sewing | Implemented, lightly tested | |
| Revolve (full/partial) | Implemented, lightly tested | |
| Sweep (single path) | Implemented; multi-guide not done | |
| Loft (ruled surfaces) | Implemented; smooth NURBS loft not done | |
| Sketch 2D | Newton-Raphson constraint solver | Implemented |
| Assembly | Data model + mates | Defined; constraint solver not done |
| Export | STL, OBJ, encrypted .ros | Works |
| STEP | Skeleton in place; output not validated | |
| AI | Claude + OpenAI providers | Works |
| Vision pipeline + smart routing | Implemented | |
| Natural language command parsing | Works for common commands | |
| Infrastructure | Timeline (event-sourced history) | Works |
| RAG (Vamana vector index) | Works | |
| Session manager (multi-user, RBAC) | Works | |
| Frontend | React + R3F viewport, toolbar, chat | Works; rough around the edges |
Measured numbers from the geometry kernel. See individual sections below for detail and methodology. Full table and reproduction commands in BENCHMARKS.md.
Criterion, release build, median of 100 samples.
| Operation | Time |
|---|---|
| Vector3 dot | 500 ps |
| Vector3 cross | 884 ps |
| Vector3 normalize | 1.68 ns |
| Vector3 add | 984 ps |
| Matrix4 multiply | 5.14 ns |
| Matrix4 inverse | 29.4 ns |
| Matrix4 transpose | 5.50 ns |
| Point3 distance | 505 ps |
| Point3 translate | 868 ps |
Full B-Rep topology construction with a fresh BRepModel per iteration. Criterion, release build.
| Primitive | Time |
|---|---|
| Box | 65 µs |
| Sphere | 49 µs |
| Cylinder | 50 µs |
Internal benchmark suite on 1k-face inputs, release build.
| Operation | Measured | Internal target |
|---|---|---|
| Boolean union | 50.5 ms | <100 ms |
| Boolean intersection | 75.4 ms | <150 ms |
| Face–face intersection | 25.3 ms | <50 ms |
All three are within their internal regression budgets on this host.
1M-point sweep, release build.
| Operation | Measured | Internal target |
|---|---|---|
| NURBS surface eval | 158 ms | <25 ms |
| B-spline curve eval | 36.8 ms | <10 ms |
Both are above their internal regression budgets — NURBS eval by ~6× and B-spline eval by ~3.7×. Flagged for profiling and optimization.
| Metric | Measured | Internal target |
|---|---|---|
| Tessellation (1M triangles, estimated) | 5,350 ms | <250 ms |
| Memory per 1M vertices (SoA layout) | 34.3 MB | <192 MB |
Memory is under budget on this host. Tessellation is the largest performance gap (~21× over target) and the next thing to profile.
- Delete primitives — only correctness tests (
delete_solid,delete_face, cascade, orphan cleanup). No Criterion target yet. - 2D sketch creation (sketch2d) — ~5k LoC subsystem with 69 passing correctness tests, but no timing benchmark target.
- Cone / Torus — primitive creation benchmarks only cover Box / Sphere / Cylinder.
These are tracked as blind spots to add to benches/geometry_bench.rs.
- Host: Windows 11, x86_64, release build.
- Profile overrides:
CARGO_PROFILE_BENCH_LTO=off,CARGO_PROFILE_BENCH_CODEGEN_UNITS=16. Full-LTO would shave an additional 15–30% off these numbers but currently hits a rustc-LLVM OOM on this host — fix tracked separately. - Criterion numbers reported as the median of 100 samples after a 3-second warmup.
- Internal-suite numbers (Boolean / NURBS / tessellation) are single-sample wall-clock with per-operation iteration counts in the 10–100 range. Treat Criterion numbers as the statistical baseline.
- "Internal target" = internal regression budget. Not a comparison against any third-party kernel.
Reproduce:
cd roshera-backend
CARGO_PROFILE_BENCH_LTO=off CARGO_PROFILE_BENCH_CODEGEN_UNITS=16 \
cargo bench -p geometry-engine --bench geometry_bench
CARGO_PROFILE_RELEASE_LTO=off CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16 \
cargo test --release -p geometry-engine --lib test_performance_benchmark_suite -- --nocapture# Backend
cd roshera-backend
cargo run --bin api-server
# API on http://localhost:3000, WebSocket on ws://localhost:3000/ws
# Frontend (separate terminal)
cd roshera-app
npm install
npm run dev
# UI on http://localhost:5173cd roshera-backend
docker compose up- Rust 1.75+
- Node.js 20+
# Create a box
curl -X POST http://localhost:3000/api/geometry \
-H "Content-Type: application/json" \
-d '{"operation": "create_primitive", "parameters": {"type": "box", "width": 10, "height": 10, "depth": 10}}'// WebSocket
const ws = new WebSocket("ws://localhost:3000/ws");
ws.send(JSON.stringify({
type: "GeometryCommand",
data: { command: "CreatePrimitive", parameters: { type: "sphere", radius: 5.0 } }
}));
The Roshera mark is a Boolean union of a rectangle (2x × 3.236x) and a circle (radius x).
Dual licensed. Free for non-commercial use (research, education, personal projects). Commercial use requires a paid license.
See LICENSE for details. Contact 29.varuns@gmail.com for commercial licensing.