juli/tsjetpiti
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity

docs: README reflects Q8 model + persona

Browse Source 
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
juli
2026-08-09 20:35:46 +02:00
co-authored by Claude Opus 4.8
parent 60cadd0adb
commit 24991e7483
1 changed files with 8 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+8 -4
View File
@@ -4,7 +4,9 @@ A dead-simple Android chat app that runs an **uncensored Qwen3.5-2B** model
fully **on-device** via an embedded [llama.cpp](https://github.com/ggml-org/llama.cpp), fully **on-device** via an embedded [llama.cpp](https://github.com/ggml-org/llama.cpp),
wrapped in an **extremely barebones WebView UI**. wrapped in an **extremely barebones WebView UI**.
Talk to the tsjet. Hit **new tsjet** to wipe the conversation and start fresh. tsjet has a deliberate personality: it **always answers**, sounds confident and
plausible, and is **hilariously, purposely wrong**. Hit **new tsjet** to wipe the
conversation and start fresh.
--- ---
@@ -28,9 +30,11 @@ Talk to the tsjet. Hit **new tsjet** to wipe the conversation and start fresh.
- **Inference:** llama.cpp compiled from source (pinned tag `b10333`) through the - **Inference:** llama.cpp compiled from source (pinned tag `b10333`) through the
NDK. Pulled automatically at build time via CMake `FetchContent` — no submodule NDK. Pulled automatically at build time via CMake `FetchContent` — no submodule
to init. to init.
- **Model:** `Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-Q4_K_M.gguf` (~1.2 GB) is - **Model:** `Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-Q8_0.gguf` (~1.9 GB) is
**downloaded on first launch** from Hugging Face into the app's private storage. **downloaded on first launch** from Hugging Face into the app's private storage.
It is *not* bundled in the APK. It is *not* bundled in the APK. (Swap the quant in `ModelDownloader.kt`.)
- **Persona:** set via `SYSTEM_PROMPT` in `MainActivity.kt`; sampling is a little
hot (temp 0.9) for playful answers.
- **Conversation:** the web layer holds the full history and sends it each turn; - **Conversation:** the web layer holds the full history and sends it each turn;
native rebuilds the ChatML prompt and clears the KV cache before every reply, so native rebuilds the ChatML prompt and clears the KV cache before every reply, so
"new tsjet" is just: clear JS state + reset cache. "new tsjet" is just: clear JS state + reset cache.
@@ -41,7 +45,7 @@ Talk to the tsjet. Hit **new tsjet** to wipe the conversation and start fresh.
in **Android Studio** (it will offer to install the matching NDK `27.2.12479018` in **Android Studio** (it will offer to install the matching NDK `27.2.12479018`
and CMake `3.22.1`). and CMake `3.22.1`).
- **A phone:** 64-bit ARM (`arm64-v8a`), Android 8.0+ (minSdk 26), and enough free - **A phone:** 64-bit ARM (`arm64-v8a`), Android 8.0+ (minSdk 26), and enough free
RAM to hold a ~1.2 GB Q4 model (~3–4 GB RAM device recommended). RAM to hold a ~1.9 GB Q8 model (~4+ GB RAM device recommended).
- Network on first launch to download the model. - Network on first launch to download the model.
## Build ## Build