# Future Development Roadmap for text-to-cad: 7 Phases from Format Support to ROS2 Integration

> Explore the 7-phase text-to-cad development roadmap: from format support to ROS2 integration, AI design, and CAM workflows. See the future of CAD generation.

- Repository: [earthtojake/text-to-cad](https://github.com/earthtojake/text-to-cad)
- Tags: roadmap
- Published: 2026-09-13

---

**The earthtojake/text-to-cad repository follows a seven-phase future development roadmap spanning robust CAD format expansion, real-time 3D viewer editing, AI-driven generative design, advanced CAM workflows, and native ROS2 robotics integration.**

The `text-to-cad` project is a modular library of agent skills that transforms natural-language or image prompts into manufacturing artifacts like STEP files, URDF models, and G-code. Its architecture separates independent skills under `skills/` from a shared runtime in `packages/cadgen/`, making the **future development roadmap for text-to-cad** a structured progression toward comprehensive design-to-fabrication automation. The planned evolution prioritizes data ingestion, interactive visualization, generative AI, and robotics simulation without breaking existing skill boundaries.

## Phase 1: Core Enhancements for Robust Format Support

The immediate focus targets expanding import and export capabilities beyond current STEP, STL, and GLB support. Planned additions include native handlers for **OBJ, FBX, and Collada (DAE)** formats, alongside enhanced STEP parsing to cover **AP242 features** for better semantic product data.

These changes primarily affect the core CAD engine described in [`packages/cadgen/README.md`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/README.md) and the underlying geometry writers like [`packages/cadgen/src/cadgen/step.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/step.py). The modular design allows new format handlers to plug into the existing conversion pipeline without modifying skill-specific logic.

## Phase 2: Interactive 3D Viewer and UI Improvements

The **CAD Viewer** back-end in [`packages/cadgen/src/cadgen/viewer/main.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/viewer/main.py) currently serves a React-based front-end for model inspection. The roadmap calls for real-time editing capabilities—including direct manipulation of geometry for move, scale, and rotate operations—plus material and texture previews.

Future iterations will introduce cloud-based collaborative sessions, enabling multiple users to review and annotate models simultaneously. The viewer’s entry point in [`main.py`](https://github.com/earthtojake/text-to-cad/blob/main/main.py) will be extended to support WebSocket connections for these real-time features.

## Phase 3: AI-Driven Design and Generative CAD

Building on the command-generation patterns in [`skills/cad/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/cad/SKILL.md), the project plans to integrate large-language models that suggest design variations based on functional requirements. The system will auto-optimize part geometry for specific constraints like **weight reduction** or **FDM printability**.

This phase leverages the existing skill architecture to add an "AI designer" skill that calls the core generation APIs with iterative refinement loops, keeping the deterministic CAD generation separate from the probabilistic AI layer.

## Phase 4: Extended Fabrication and Advanced CAM

Current G-code generation in [`skills/gcode/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/gcode/SKILL.md) supports Cura-based slicing workflows. The roadmap expands this to **multi-material G-code generation** and integration with additional slicers like PrusaSlicer and Simplify3D.

Direct upload capabilities to cloud-printing services such as AstroPrint are also planned. The format conversion pipeline in [`packages/cadgen/src/cadgen/viewer/cadgen_ops.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/viewer/cadgen_ops.py) will handle mesh preprocessing for these advanced CAM outputs.

## Phase 5: Robotics Integration with Dynamic URDF and ROS2

The existing URDF and SRDF skills defined in [`skills/urdf/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/urdf/SKILL.md) and [`skills/srdf/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/srdf/SKILL.md) will evolve to support real-time kinematics updates driven by CAD geometry changes. Automatic generation of collision meshes and inertia tensors will eliminate manual robot configuration.

**ROS2-native bindings** will enable live simulation feedback, allowing the CAD Viewer to act as a visualizer for MoveIt! planning or Gazebo simulations directly from the generated description files.

## Phase 6: DevOps, CI Coverage, and Release Automation

Infrastructure improvements target the `.github/workflows/` directory to expand test matrices covering new file formats and viewer WebSocket features. The pipeline will automate release note generation from GitHub Issues and enable continuous delivery of the CAD Viewer as a standalone web application.

This phase ensures that the growing complexity of format support and AI integration remains stable through automated regression testing of the `cadgen` package and all skill definitions.

## Phase 7: Documentation, Tutorials, and Community Ecosystem

The final phase focuses on community growth through step-by-step Jupyter notebooks for common use cases and a community-contributed skill marketplace. Roadmap voting will move to Discord or GitHub Discussions, guided by [`CONTRIBUTING.md`](https://github.com/earthtojake/text-to-cad/blob/main/CONTRIBUTING.md) and the skills table in the main [`README.md`](https://github.com/earthtojake/text-to-cad/blob/main/README.md).

Documentation site updates in [`apps/docs/README.md`](https://github.com/earthtojake/text-to-cad/blob/main/apps/docs/README.md) will centralize tutorials for extending the `cadgen` runtime and authoring new skills.

## Practical Code Examples for Current Usage

While the roadmap advances, the current toolkit already supports end-to-end CAD generation and visualization.

Generate a STEP model from plain text using the Skills CLI:

```bash

# Using the Skills CLI (installed via `npx skills add earthtojake/text-to-cad`)

skills run cad "Create a 30 mm diameter gear with 20 teeth and a 5 mm hub"

```

*Result*: A STEP file saved to `output/gear.step` generated via the logic in [`skills/cad/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/cad/SKILL.md) and written by [`packages/cadgen/src/cadgen/step.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/step.py).

Launch the CAD Viewer to render the generated geometry:

```bash

# Launch the viewer (Python back-end) and open the browser

python -m cadgen.viewer

```

Open `http://localhost:8000` to explore the model. The server implementation lives in [`packages/cadgen/src/cadgen/viewer/main.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/viewer/main.py).

Export models to multiple formats simultaneously:

```bash
skills run cad "Create a simple box 10 mm × 20 mm × 30 mm" \
  --export stl glb

```

The `--export` flag triggers the **format conversion pipeline** defined in [`packages/cadgen/src/cadgen/viewer/cadgen_ops.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/viewer/cadgen_ops.py).

## Summary

- The **text-to-cad** roadmap follows seven distinct phases from core format support (Phase 1) through community ecosystem growth (Phase 7).
- Key technical targets include AP242 STEP support, real-time viewer editing in [`main.py`](https://github.com/earthtojake/text-to-cad/blob/main/main.py), AI-driven optimization layers, and ROS2-native robotics bindings.
- The architecture separates concerns between `packages/cadgen/` (runtime) and `skills/` (capabilities), enabling parallel development across roadmap phases.
- Current functionality already supports CLI-driven CAD generation, multi-format export, and local 3D visualization.

## Frequently Asked Questions

### What new file formats will text-to-cad support in the future?

The roadmap prioritizes **OBJ, FBX, and Collada (DAE)** for import and export, alongside enhanced **AP242 STEP** features for richer semantic data. These additions extend the existing format handlers referenced in [`packages/cadgen/README.md`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/README.md) without breaking current STEP or STL workflows.

### Will text-to-cad include a web-based collaborative viewer?

Yes, Phase 2 of the roadmap specifically targets cloud-based collaborative sessions within the CAD Viewer. The implementation extends the HTTP server in [`packages/cadgen/src/cadgen/viewer/main.py`](https://github.com/earthtojake/text-to-cad/blob/main/packages/cadgen/src/cadgen/viewer/main.py) with WebSocket support to enable real-time multi-user editing and annotation.

### How does the roadmap address AI integration?

Phase 3 introduces **generative CAD** capabilities by plugging large-language models into the skill system defined in [`skills/cad/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/cad/SKILL.md). This will enable automated design variation suggestions and geometry optimization for printability while maintaining the existing deterministic generation pipeline.

### Can text-to-cad generate ROS-compatible robot descriptions?

Current URDF and SRDF generation in [`skills/urdf/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/urdf/SKILL.md) and [`skills/srdf/SKILL.md`](https://github.com/earthtojake/text-to-cad/blob/main/skills/srdf/SKILL.md) will expand in Phase 5 to include automatic collision mesh generation, inertia tensor calculation, and **ROS2-native bindings** for direct integration with simulation environments like Gazebo and MoveIt!.