- Jasper AI costs $49–$599/month – insane for small teams.
- ChatGPT / Claude send your data to the cloud – no privacy.
- OpenAI API bills per token – costs explode at scale.
- Closed source – you can't audit, can't modify, can't trust.
BAKOME Local AI Studio runs entirely on your own hardware.
No cloud. No subscription. No surveillance.
| Feature | Jasper AI | BAKOME Studio |
|---|---|---|
| Monthly cost | $49–$599 | $0 |
| Data privacy | Cloud (they read it) | Local (your machine) |
| Open source | ❌ | ✅ (MIT) |
| Multiple models | ❌ (GPT only) | ✅ Llama, Mistral, Phi, CodeLlama |
| Offline mode | ❌ | ✅ |
| Custom templates | ❌ | ✅ |
| PDF / HTML export | ❌ | ✅ |
| Audit history | Limited | Full SQLite audit |
- Bring your own Ollama (free, local LLM).
- No API keys, no usage limits, no surprise bills.
- Blog posts, landing pages, email newsletters, LinkedIn posts
- Code documentation, docstrings, refactoring
- Social media threads, product descriptions, ad copy
- All generations logged in SQLite
- Projects, templates, prompts, results – everything stored locally
- Llama 3.2, Mistral, Phi‑3, CodeLlama, Gemma
- Switch models dynamically from the UI
- Export to PDF, Markdown, HTML
- Ready for client delivery or internal reports
- Clean FastAPI backend + React frontend
- Easy to add new templates, categories, or custom logic
| Model | Generation speed (500 tokens) | Memory |
|---|---|---|
| Llama 3.2 3B | 2.3 sec | 4 GB |
| Mistral 7B | 4.1 sec | 8 GB |
| Phi‑3 Mini | 1.8 sec | 3 GB |
| CodeLlama 7B | 4.5 sec | 8 GB |
Tested on: 8‑core CPU, 16 GB RAM, no GPU required.
- Python 3.11+
- Ollama installed (
ollama pull llama3.2:3b)
git clone https://github.com/muguamismael-commits/BAKOME_Local_AI_Studio.git
cd BAKOME_Local_AI_Studio/backend
pip install -r requirements.txt
python backend_main.py