fix: ZLUDA compatibility issues (#582)

fix: include cudnn in zluda

fix!: require CUDA compute cap 8.0

perf: FLUX.2 improvements

feat: koharu-ai crate

refactor: move keyring to koharu-runtime

fix: download stuck at 100% sometimes

refactor(runtime): Tauri shell + downloads queue for the v2 backend

feat: portable mode (#415)

fix: resolve incorrect ZLUDA DLLs usage (#442) (#458)

fix: load correct ZLUDA DLLs

fix: upgrade zluda to v6-preview.64

chore: enable Sentry only for production builds (#397)

refactor(llm): deel & google translate (#361)

feat: add AMD GPU support via ZLUDA (#362)

feat(mt): DeepL/Google as koharu-llm providers (#357)

feat: update llama.cpp to b8665

fix: http timeouts

feat: pluggable engine system with DAG pipeline

refactor(ui): centralize error handling with XState

refactor: simplify Tauri setup and dev server

refactor: rewrite backend storage, config, and runtime

chore: remove macos warnings

feat: bootstrap window

refactor: merge crates

fix(linux): load native libraryes by using explicit path to resolve missing libggml errors on Unix systems (#295)

chore: make clippy happy

feat: llama.cpp-based paddleocr-vl-1.5

refactor: koharu-runtime

feat: add open folder import, CUDA 13.2 build fixes, local LLM settin… (#271)

ci: add cargo fmt

remove unnecessary change

feat: implement API provider support with OpenAI, Gemini, and Claude integration

refactor: reduce duplicates

chore: remove unused fonts

refactor: split modules

feat: CUDA 13.1

feat: download progress indicator

chore: add more context info

fix: pin cudnn version

chore: reorganize code

tests: ignore dylib tests

feat: google fonts

tests: fix tests

chore: make clippy happy

chore: cleanup tracing logs

refactor: re-introduce koharu-core

fix: portable mode libs and models location

fix: refine url

fix: pypi mirror