10 Commits

Author SHA1 Message Date
8d5269a420 Add model and project credits to README 2026-08-09 23:57:08 +02:00
bd9f311b4b Polish README branding 2026-08-09 23:51:25 +02:00
f417fb7331 Fix WebView system bar and keyboard insets v0.1.1 2026-08-09 22:44:13 +02:00
164bd2130a Switch model to Q6_K (~1.5 GB), lighter/faster than Q8
- ModelDownloader now pulls the Q6_K quant
- README: reflect Q6_K sizes + note prebuilt APK on Releases

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
v0.1.0
2026-08-09 22:12:48 +02:00
b06a0349a9 On-device fixes: optimized native build, 16 KB align, disable thinking
Validated on a Pixel 7 (adb). Key fixes:
- Force Release/-O3 for native even in the debug variant (AGP defaulted to -O0,
  making llama.cpp ~10x too slow / effectively unusable).
- 16 KB-align all native LOAD segments (Play requirement; 16 KB-page devices).
- Disable Qwen3 thinking (empty <think></think> prefill + /no_think) so replies
  are fast and punchy instead of spending the whole budget on hidden reasoning.
- Cap replies at 220 tokens.

Verified: model downloads, loads (n_ctx=4096, 6 threads), streams an on-persona
reply with emoji intact, and "new tsjet" clears the conversation.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-09 22:06:30 +02:00
20a16c9faf Build verified; add icon-generator tool and install/README notes
- ./gradlew assembleDebug produces a 13 MB arm64-v8a APK (libtsjet/libllama/
  libggml/libc++_shared bundled; JNI compiles cleanly against llama.cpp b10333)
- Add tools/gen_launcher_icons.py so the launcher icon is reproducible from art
- README: mark build verified, add sideload/install section

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-09 21:04:49 +02:00
dd69cb6d65 Add app icon (adaptive + legacy) and header logo
- Launcher icon generated from piti-icon.png: adaptive icon (black bg +
  brain foreground) for API 26+, plus legacy square/round PNGs at all densities
- Web header now uses the tsjetpiti wordmark (piti-logo.png -> assets/web/logo.png)
- App label -> "tsjetpiti"; keep source art (piti-icon.png, piti-logo.png) in repo

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-09 21:01:09 +02:00
24991e7483 docs: README reflects Q8 model + persona
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-09 20:35:46 +02:00
60cadd0adb Q8 model, confidently-wrong persona, sleek white-on-black UI
- Switch model to Q8_0 (~1.9 GB)
- Add tsjet persona: always replies, plausible-sounding but hilariously wrong
- Bump sampling (temp 0.9 / top-k 60) for playful output; cap replies at 384 tokens
- Redesign UI: monochrome white-on-black, generic IT sans font stack

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-09 20:35:10 +02:00
ac8490b452 Scaffold tsjetpiti: on-device Qwen3.5-2B chat app
Android app that runs an uncensored Qwen3.5-2B GGUF fully on-device via an
embedded llama.cpp (pinned b10333, built through the NDK) behind a barebones
WebView chat UI. Model is downloaded on first launch. "new tsjet" wipes the
conversation and resets the KV cache.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-09 20:24:14 +02:00