Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/thePrnvBot/dispel-web-stylist/llms.txt

Use this file to discover all available pages before exploring further.

Dispel supports five AI providers out of the box and can connect to any additional service that exposes an OpenAI-compatible REST API. Every API key you enter is written exclusively to chrome.storage.local — the browser’s own sandboxed, extension-scoped storage — and is never transmitted to any Dispel server. Keys leave your machine only when Dispel makes a streaming request directly to the provider endpoint you configured, carried in the Authorization request header over HTTPS.

​
Built-in providers

Each provider has a preset base URL that Dispel fills in automatically when you select it from the provider dropdown. You can override the base URL at any time for custom deployments.
ProviderBase URLNotes
OpenAIhttps://api.openai.com/v1Requires an OpenAI API key
Anthropichttps://api.anthropic.com/v1Requires an Anthropic API key
Googlehttps://generativelanguage.googleapis.com/v1betaRequires a Google API key
OpenRouterhttps://openrouter.ai/api/v1Requires an OpenRouter API key
Localhttp://localhost:11434/v1 (Ollama default)No API key required

​
Adding an API key

1

Open the Dispel side panel

Click the Dispel icon in your browser toolbar. The side panel opens on the right side of the active tab.
2

Open the Options page

Click the Settings (gear) icon in the side panel header, or right-click the Dispel toolbar icon and choose Options.
3

Navigate to the Models tab

In the Options page, click the Models tab at the top of the page.
4

Click Add Model

The model configuration form is displayed at the top of the Models page. Select your provider from the Provider dropdown — Dispel automatically fills in the preset base URL and loads any previously saved API key for that provider.
5

Enter your API key

Paste your API key into the API Key field. Use the eye icon to reveal the value and confirm it was pasted correctly.
6

Fill in the Model ID and save

Enter the model ID (for example gpt-4o for OpenAI) and click Save. The model appears in the Your models table below the form.

​
Custom OpenAI-compatible endpoints

Any inference server that exposes an OpenAI-compatible chat completions API can be connected to Dispel by selecting the Local \ OpenAI-Compatible provider type and entering a custom base URL. The local provider type uses the @ai-sdk/openai-compatible adapter under the hood, which means it works with any endpoint that follows the OpenAI streaming protocol.

Ollama

Run open-weight models locally. Set the base URL to http://localhost:11434/v1. No API key is needed — leave the API Key field blank.

vLLM

High-throughput inference server. Set the base URL to your vLLM instance, e.g. http://localhost:8000/v1. Use any non-empty string as the API key if your instance requires one.

LiteLLM

Proxy that exposes hundreds of models behind a single OpenAI-compatible endpoint. Point Dispel at your LiteLLM proxy URL and provide a LiteLLM virtual key.

Other endpoints

Any server implementing POST /v1/chat/completions with streaming (text/event-stream) works. Select the Local provider, enter the full base URL, and supply an API key if required.

​
Security

API keys are stored using the browser’s built-in chrome.storage.local API. Storage is:
  • Sandboxed — only the Dispel extension can read or write its own storage area.
  • Extension-scoped — no website or other extension can access Dispel’s storage.
  • Local-only — keys are never synced to chrome.storage.sync and never sent to any Dispel backend.
When Dispel makes a request to a provider, it places your API key in the Authorization: Bearer <key> header of an HTTPS request sent directly from your browser to the provider’s endpoint. Dispel’s own servers are not involved in that request path.
Never share your API keys with anyone. Before uninstalling Dispel or lending your browser profile to someone else, remove your keys by opening Options → Settings and clicking Clear all data, or by deleting each model entry individually from the Models tab.

​
Provider-specific notes

Dispel uses the @ai-sdk/google adapter with the v1beta base URL (https://generativelanguage.googleapis.com/v1beta). Google model IDs follow a different naming convention from OpenAI — use identifiers like gemini-2.0-flash or gemini-2.5-pro rather than a slash-separated path. The v1beta endpoint exposes features (such as system instructions and streaming tool calls) that are not yet available on v1.
OpenRouter provides access to models from many providers through a single API key, using a unified endpoint at https://openrouter.ai/api/v1. Model IDs on OpenRouter follow the format provider/model-name — for example openai/gpt-4o or anthropic/claude-opus-4-5. When Dispel detects that the configured base URL’s origin matches https://openrouter.ai, it normalizes the URL by omitting the base URL override and lets the @openrouter/ai-sdk-provider handle routing, so the preset URL is used automatically.
No API key is required for local models. Set the base URL to http://localhost:11434/v1 (the Ollama default) and enter the Ollama model name as the Model ID (for example llama3.2 or mistral). Ensure Ollama is running before submitting a prompt. For other local runtimes, adjust the base URL to match that server’s address and port.
Each provider’s own privacy policy governs how they handle the data you send in prompts and page context. Review the relevant policy before sending sensitive content: OpenAI Privacy Policy · Anthropic Privacy Policy · Google Privacy Policy · OpenRouter Privacy Policy

Build docs developers (and LLMs) love

Get started for freeTalk to us