RikkaHub is an Android app that lets users chat with multiple LLM providers and manage conversations, with a history of releases improving UI, performance, and feature support.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
RikkaHub is an Android application written in Kotlin that supports switching between multiple AI providers for conversations. It includes features like Markdown rendering, memory, and prompts customization, and targets Android devices with Jetpack Compose-based UI.
How it works
The project describes a multi-provider chat client with a workspace concept and support for custom HTTP request headers and bodies, plus a structured UI for selecting models/providers. It includes a proot-based Linux agent workspace and a modular UI for model selection and chat interaction. See the Releases for improvements across rendering, conversation context handling, and model support.
Getting started
Install instructions are not fully shown in the README excerpt. The repository notes that a google-services.json file is required to build the app (in the Tech stack section):
"You need a google-services.json file at app folder to build the app."
Recent releases
-
2.4.5 (2026-07-31):
- Code highlighting now uses native implementation for faster, smoother blocks
- Flow diagrams display speed improved
- Context handling for long conversations optimized to reduce duplication and costs
- Speech playback default speed setting added
- Improved support for Moon Dark Side model with default enabling and parameter compatibility
- Fixed image generation errors from Grok
- Workspace now reads HEIC/HEIF/AVIF/ICO image formats
- Fixed issue where switching assistants caused new conversations to belong to the previous
-
2.4.3 (2026-07-25):
- Added minimalist White and Claude themes
- System file manager copy/move to workspace support
- Removed assistant context message limit
- Local and cloud backups become independently managed
- Fixed crashes when selecting voice services and issues with skill recognition
- Fixed overlapping between built-in and custom tools
- Various fixes to workspace file reads and message time display, image cropping prompts
-
2.4.2 (2026-07-16):
- Exa web search now retrieves webpage content for fuller data
- One-click chat file cleanup to free storage
- File management supports text editing and media previews
- Web search switches per-assistant
- Speech playback can skip content in parentheses
- Added Kimi K3 model
- Optimized folder sync and auto-repair missing/invalid files
- Improved default prompts for suggested replies
- Fixed web service startup issues on some devices
-
2.4.1 (2026-07-08):
- Added MiniMax, MIMO providers and Fish Audio speech support
- Some TTS support prompting AI for tone control (currently only MIMO)
- Support uploading and cropping more image formats (HEIF/HEIC)
- Image generation supports custom sizes
- Web version adds folder management
- Fixed occasional character loss in replies and formula rendering issues
- UI polish: fixed button clipping
-
Nightly Build (2026-07-08): auto daily build for testing (commit 4b6449e37854809302cdf5f1e72231b420f57382)
Traction
The repository shows 6629 stars and 526 forks as of the current data. Open issues: 242.
Behind the repo
Not applicable as no linked startup/company information is provided in the facts.
Caveats
License: AGPL-3.0. Created 2025-03-11; last push 2026-08-04. The README mentions forks may exist with privacy or permission concerns; caution advised when using forks. The PR policy within the README indicates not accepting translation changes or new features, and mentions nightly builds may be unstable.