Running against the localhost API. No WebGPU required. CV text is sent only to 127.0.0.1 on this machine.

Draft thematic CV sections on your computer

Turn a conventional CV into evidence for a Canadian Tri-Agency narrative CV. This desktop build keeps extract, verify, ranking, and draft in the browser UI while a Python process on localhost runs the language model (Ollama or llama-cpp-python).

Why desktop + local API?

No WebGPU requirement, easier model management, and stronger offline control. CV text is sent only to 127.0.0.1—never to a cloud model API. However, all data stays on your local machine and is not sent to the cloud.

What stays in the browser

CV parsing, evidence cards, gap audit, discipline ranking UI, and draft editing remain the same constrained workflow as the web prototype.

What Python owns

Model load, chat completions, and streaming drafts via Ollama (preferred) or GGUF files through llama-cpp-python.

How would you like to begin?

All three routes lead to the same evidence review. Choose the source that gives you the best factual starting point.

Requires the desktop app (or python -m localtcv_desktop) listening on localhost. WebGPU is not required.

Made in Canada 🇨🇦

Local preferences

Settings

Lists Ollama tags and GGUF files known to the desktop API.

Optional browser save

Opt-in only. CV text, verified evidence, selections, and drafts stay on this device in this browser’s site storage. This storage is not described as encrypted.

Browser saving is off.

Changing models takes effect the next time you start local AI.

Local model setup

Choose a model

The first use may download model weights through Ollama. The download and inference stay on this computer.

About this project

Local TCV Desktop

Local TCV helps researchers draft Canada’s Tri-Agency thematic CV sections using a focused, evidence-based workflow and a small local language model.

Privacy firstCV text is processed in this browser and sent only to the localhost app and your local model backend.
Three sectionsPersonal statement, significant contributions and experiences, and supervisory or mentorship activities.
Built in CanadaIndependent project by Sidney Shapiro at the University of Lethbridge.
VersionLocal TCV Desktop v0.1.2 · macOS and Windows.

Not affiliated with SSHRC, NSERC, or CIHR. Always use the current official application instructions and verify every claim.

Quick help

Using Local TCV Desktop

1. Start local AI

Choose a model and select Download and start. Basic or Lightweight models are best for computers with limited memory.

2. Add your source material

Upload a text-based PDF, DOCX, or TXT file, paste selected notes, or use the guided interview. Scanned image PDFs are not supported.

3. Verify before drafting

Review every extracted card, fill gaps about your role and impact, then rank the most relevant contributions. The model can omit or misunderstand facts.

Troubleshooting

Open the full documentation on localtcv.ca