MiniiChat

A small, open-source Android chat client for any OpenAI-compatible LLM API. Native Kotlin + Jetpack Compose, no backend, bring your own keys.

Features

• Multi-provider — point at any OpenAI-compatible endpoint (OpenAI, OpenRouter, DeepSeek, Groq, Mistral, Together, Gemini OpenAI shim, SiliconFlow, Ollama, LM Studio, custom). 11 built-in presets.
• Auto-fetch models — GET /models pulls the model list per provider; manual add and remove also supported. Lists collapse after 5 entries.
• Bottom-sheet model picker — tap the chip in the top bar, search across all providers.
• Custom assistants — name, emoji avatar, system prompt, optional temperature override. 4 built-in presets (Default / Coder / Translator / Writer).
• Prompt variables — {model} {provider} {assistant} {date} {time} {datetime} {weekday} {locale} rendered into the system prompt.
• Per-provider customization — custom HTTP headers and extra body params (data layer; visual editor coming).
• Media attachments — pick image or file from + menu. Images are sent as base64 data URLs (works with vision-capable models).
• Streaming — Server-Sent-Events with a stop button.
• Markdown — headings, fenced code blocks, lists, quotes, bold / italic / inline code / strikethrough.
• Material You + theme mode — dynamic color on Android 12+, plus follow-system / light / dark.
• Bilingual — English and 简体中文 (Settings → Language).
• Pure local storage — DataStore preferences, no analytics, no backend.

Download

• Latest release: https://github.com/Minis233/miniichat/releases/latest
• Each release ships a signed release APK (for general use) and a debug APK (for development).

Configuration

1. Open the drawer (top-left) → Settings → Providers
2. Add provider, pick a preset (or "Custom"), paste your API key
3. Tap Fetch models, or add a model id manually
4. Back to chat, tap the model chip in the top bar to switch model
5. (Optional) Settings → Assistants to set up role-specific system prompts

Other compatible endpoints:

Provider	Base URL	Example model
OpenAI	https://api.openai.com/v1	gpt-4o-mini
OpenRouter	https://openrouter.ai/api/v1	openrouter/auto
DeepSeek	https://api.deepseek.com/v1	deepseek-chat
Groq	https://api.groq.com/openai/v1	llama-3.1-70b-versatile
Mistral	https://api.mistral.ai/v1	mistral-small-latest
Together	https://api.together.xyz/v1	meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo
Gemini (OpenAI shim)	https://generativelanguage.googleapis.com/v1beta/openai	gemini-1.5-flash
SiliconFlow	https://api.siliconflow.cn/v1	Qwen/Qwen2.5-7B-Instruct
Ollama	http://<host>:11434/v1	llama3.2
LM Studio	http://<host>:1234/v1	(your loaded model)

The base URL field auto-appends /v1 if you forget it. For Ollama / LM Studio on a real device, replace 10.0.2.2 (emulator-only) with your computer's LAN IP.

Build

The repo includes a GitHub Actions workflow that produces signed release and debug APKs on every push.

# Local build (needs JDK 17 + Android SDK)
./gradlew :app:assembleDebug
./gradlew :app:assembleRelease   # falls back to debug-signed if no release.keystore present
# APKs at app/build/outputs/apk/{debug,release}/

To make a release locally, drop a release.keystore next to app/build.gradle.kts and set RELEASE_STORE_PASSWORD, RELEASE_KEY_ALIAS, RELEASE_KEY_PASSWORD env vars (or gradle properties).

CI uses an ephemeral keystore generated at build time, so artifacts won't match a previous install signature — uninstall the debug APK before installing the release APK on the same device.

Project layout

app/src/main/kotlin/com/miniichat
├── MainActivity.kt                  # Compose entry, locale + theme
├── ChatViewModel.kt                 # State + send/stream/persist
├── api/LlmClient.kt                 # ktor + SSE, multipart content
├── data/                            # DataStore (settings + conversations + providers + assistants)
├── ui/AppRoot.kt                    # Routing, drawer, back-stack
├── ui/ChatScreen.kt                 # Top bar, message stream, bubbles
├── ui/InputBar.kt                   # Text input + attachment picker
├── ui/ModelPicker.kt                # Bottom-sheet model selector
├── ui/Drawer.kt                     # Date-grouped chat list, search
├── ui/SettingsScreen.kt             # Settings hub
├── ui/ProvidersScreen.kt            # Provider editor + fetch models
├── ui/AssistantsScreen.kt           # Assistant editor
├── ui/MarkdownText.kt               # Inline markdown renderer
└── ui/theme/Theme.kt                # Material 3 + dynamic color

License

MIT

Acknowledgements

This project's feature scope is inspired by two excellent open-source LLM clients. No source code from either project is copied — MiniiChat is original Kotlin/Compose, and the references below are to their public README feature lists (functionality is not protected by copyright). Both deserve a star.

• rikkahub/rikkahub — A native Android LLM chat client, Apache-2.0. Inspired the Material You theming, multi-provider switching, prompt-variables format, and custom-headers idea.
• Chevey339/kelivo — A Flutter LLM chat client, AGPL-3.0. Inspired the assistant-presets concept, multi-language support, and the per-provider extra-body customization angle.

If you build something on top of MiniiChat, please go give those projects a ⭐ as well.