AI chat for Flutter
Streaming AI chat with tool calling
A hand-rolled SSE client against any OpenAI-compatible endpoint, no OpenAI SDK, with optional tool calling and conversation persistence.
What this feature covers
- Streaming: Streaming chat completions rendered token by token.
- Any endpoint: A user-supplied base URL, API key and model.
- Tool calling: Optional function calling from the assistant.
- Persistence: Optional conversation history and message storage.
How Flutter Agent Kit implements it
The skill uses no OpenAI SDK. A hand-rolled HTTP client owns the server-sent events stream, and a ChatNotifier drives the streaming loop into the UI. Provider configuration, the endpoint, model and key, is handled by a separate AiProviderNotifier, with the key kept in secure storage and the rest in shared preferences.
Only the http package is required, already part of the scaffold. Markdown rendering and message ids use optional packages, and the whole flow runs on Android, iOS, macOS and web, where the streaming client switches to a Fetch-based transport.
Frequently asked questions
Does the kit use the OpenAI SDK?
No. The kit calls the endpoint with raw http and a hand-rolled server-sent events client. That keeps the integration compatible with any OpenAI-compatible endpoint and avoids an extra SDK dependency, matching the kit's http-only networking rule.
Where is the API key stored?
In secure storage, never in plain preferences. The endpoint and model are non-sensitive and live in shared preferences, but the key goes through the kit's secure-storage provider, the same one used for other secrets.
Get Flutter Agent Kit
Flutter Agent Kit ships this as a ready feature skill, together with the AGENTS.md context layer and the rest of the skills. €99.99 one-time, lifetime updates, unlimited personal and commercial projects.
Browse all features included in Flutter Agent Kit.