How It Works for Users
1
Use the app as normal
Open the AI-enabled app in your browser (in the live preview or as an exported HTML file). Everything works as expected until the first feature that calls AI.
2
The Connect dialog appears automatically
The first time the app needs to make an AI call, a Connect your AI provider dialog appears. You do not need to do anything special to trigger it — the app handles this for you.
3
Enter your provider details
Fill in the three required fields:
- API endpoint — the base URL for your OpenAI-compatible provider (defaults to
https://api.openai.com/v1) - Model — the model name to use, for example
gpt-4o-mini - API key — your secret key for that provider (for example, an
sk-…key from OpenAI)
4
Choose whether to remember your credentials
Check Remember on this device if you want your credentials saved across sessions. Without this, the key is cleared when the tab closes and you will be prompted again next time.
5
Click Connect AI
Your credentials are saved in browser storage and the original AI request completes immediately. Future calls in the same session skip the dialog entirely.
The Connect Dialog
The dialog asks for three fields and one option:
The dialog can also be opened at any time by calling
blip.ai.configure() — for example, if the app includes a Settings button that lets users update their provider.
Security Notes
A few things to keep in mind:- The key is never sent to AppBlips or written into the generated HTML file.
- The key is accessible to scripts running inside the app (the same origin), which is unavoidable for client-side browser storage.
- If you use Remember on this device, the key persists in
localStorageuntil you callblip.ai.clearConfiguration()or clear your browser data manually. - Without Remember on this device, the key lives in
sessionStorageand is cleared automatically when the tab closes.
Compatible Providers
BYOK mode works with any provider that implements the OpenAI chat completions API format and allows browser requests via CORS. Examples include:- OpenAI —
https://api.openai.com/v1, models likegpt-4o,gpt-4o-mini - Azure OpenAI — your Azure deployment endpoint
- Groq —
https://api.groq.com/openai/v1, fast open-source models - Ollama —
http://localhost:11434/v1for locally running models (must have CORS enabled) - LM Studio —
http://localhost:1234/v1for locally running models (must have CORS enabled)
For Self-Hosted Operators
BYOK is the default mode and requires no additional server-side configuration beyond what is already in your.env file. The relevant variable is:
APPBLIPS_APP_LLM_* variables are needed — those are only used in relay mode.
If you want to fund AI access for your users so they never need to enter a key, see Relay Mode.
