System requirements (Desktop scanner)
RAM, disk, estimated times and when you need to download the AI model.
Summary
| Mode | Recommended RAM | Free disk | CPU | GPU |
|---|---|---|---|---|
| Structural scan only (default) | 4 GB | ~500 MB (app + artifacts) | Any recent x64 | No |
| + Infer schemas with AI | 16 GB or more | ~5–6 GB additional (model) | 4+ cores recommended | No (CPU only) |
Capydox Desktop
- Operating system: Windows 10/11, macOS or Linux (depending on published builds).
- Architecture: 64-bit.
- Network: required to download the AI model (optional) and to upload OpenAPI to Capydox web.
Included components
| Component | Usage |
|---|---|
| llama.cpp (bundled binary) | Only if you enable AI schema inference |
GGUF model (qwen2.5-coder-7b-instruct-q4_k_m, ~4–5 GB) | Only with optional AI; saved in the app user data folder |
Structural route and DTO scanning does not load the model.
Project size
- Large projects (thousands of files) take longer on the initial scan.
- Common folders are skipped:
node_modules,target,build,.git,vendor, Python virtual environments, etc. - If the analysis fails due to path size, analyze a submodule (just the API folder).
Estimated times
Approximate values on a modern CPU:
| Scenario | Time |
|---|---|
| Small API (< 20 route files) | 10–30 s |
| Medium API (50 files) | 30 s – 2 min |
| Large backend (100+ controllers) | 2–5 min |
| + AI schema inference | +1–3 min (hardware-dependent) |
Avoid closing Capydox Desktop during analysis.
Upload to Capydox web
In addition to the desktop app:
- Logged into Capydox.
- Capydox backend accessible (same configuration as the rest of the app).
- Permissions to create or update specs in the chosen workspace.
Don't meet AI requirements?
You can use structural scan only: you get routes and, in many stacks, schemas read from code without downloading the model.