Model Configurations — BYOK vs Shopbell-Managed

Every workflow uses three models: an LLM (brain), a TTS (voice), and an STT (ears). For each, you have two choices. <strong>Shopbell-managed</strong>: use our hosted inference layer — your $99/mo covers all usage, no API keys to manage. <strong>BYOK (Bring Your Own Key)</strong>: paste your OpenAI / Anthropic / Deepgram / ElevenLabs API key — you pay those vendors directly for token usage. This guide covers when to use each and how to switch.

When BYOK makes sense

BYOK is right when: you already pay for OpenAI/Anthropic for other apps and want unified billing; you need a specific model Shopbell doesn't host (e.g., a fine-tuned model, an entreprise Antropic tier); you need rock-bottom per-token costs by Cherry-picking the cheapest model per call.

When Shopbell-managed makes sense

Managed is right when: you want zero setup, one bill, predictable $99/mo with no per-minute surprises; you don't want to manage API keys, daily rate limits, or vendor outages; you want our team to monitor and upgrade the underlying models over time at no extra cost to you.

Most Self-Serve DIY customers start with managed, then switch specific workflows to BYOK if they need a specific model. Switch any workflow from managed to BYOK anytime — change the model configuration in the workflow's Settings, save, test with a Web Call.

Setting up BYOK

Click Model Configurations in the sidebar. Click New Configuration. Name it ('My OpenAI GPT-4o'). Pick the provider (OpenAI, Anthropic, etc.). Paste the API key. Save. In any workflow, click the workflow's Settings, change the LLM/TTS/STT model configuration dropdown to your new BYOK configuration. Test the workflow with a Web Call.

Frequently Asked Questions