Bring Your Own Key (BYOK) & Air-Gapped Local Inference via Ollama / vLLM
Route Unchained Code through your own OpenRouter, DeepSeek-V4 Pro, or 100% private local GPUs.
Developers working with sensitive IP, proprietary codebases, or strict compliance requirements cannot send their source trees through closed third-party SaaS wrappers. Unchained Code natively supports BYOK (Bring Your Own Key) for direct wholesale inference pricing and local air-gapped execution via Ollama or vLLM.
To configure BYOK: Open Settings > Model Providers. Paste your DeepSeek, OpenRouter, Groq, Mistral, or Gemini API key. For 100% private offline development, set your base URL to http://localhost:11434/v1 (Ollama) or http://localhost:8000/v1 (vLLM). Your tokens never touch intermediate servers and bypass all platform rate limits.
Step-by-Step Walkthrough & User Guide
Access Provider Configuration in Unchained Code
Open the top-right Settings modal and click on 'API Providers'.
Add Your Custom API Key or Local Endpoint
Enter your OpenRouter key (sk-or-v1-...) or local Ollama URL (http://localhost:11434/v1).
Select Local Model Target
Choose 'qwen3.8-max' or 'deepseek-v4-pro' for offline air-gapped generation.
Start Building with Zero Rate Limits
Run full-stack generations with 100% data sovereignty and zero SaaS markups.
💡Pro Tips & Engineering Best Practices
- •Run 'ollama run qwen2.5-coder:32b' on an RTX 3090/4090 for instant local code generation.
- •Use vLLM with '--tensor-parallel-size 2' on dual-GPU setups to achieve over 120 tokens/sec.
- •BYOK users automatically bypass all throttling tiers and enjoy permanent green-lane routing.
Code & Technical Integration
# Start local vLLM OpenAI-compatible server
python3 -m vllm.entrypoints.openai.api_server \
--model Qwen/Qwen2.5-Coder-32B-Instruct \
--port 8000 \
--host 0.0.0.0 \
--max-model-len 32768