Wisp Open Source Notices

Last updated: September 23, 2026

Wisp uses open source software, model tooling, and design assets with gratitude.

App dependencies and assets

Textual: Markdown editing and rendering foundation. MIT. https://github.com/gonzalezreal/textual
DSWaveformImage 14.5.0: live recording waveform rendering. MIT. https://github.com/dmrschmidt/DSWaveformImage
RevenueCat Purchases iOS 5.82.0: Apple purchase verification, entitlement management, and purchase restoration. MIT. https://github.com/RevenueCat/purchases-ios
swiftui-math: Textual's mathematical text rendering dependency. MIT. https://github.com/gonzalezreal/swiftui-math
swift-concurrency-extras: Textual's concurrency helper dependency. MIT. https://github.com/pointfreeco/swift-concurrency-extras
MLX Swift: Apple-silicon array and model runtime foundation. MIT. https://github.com/ml-explore/mlx-swift
MLX Swift LM: optional on-device language-model loader and generation adapter. MIT. https://github.com/ml-explore/mlx-swift-lm
Swift Transformers: tokenizer, model and generation support used by MLX Swift LM. Apache-2.0. https://github.com/huggingface/swift-transformers
Swift Hugging Face: Hub support used transitively by Swift Transformers. Apache-2.0. https://github.com/huggingface/swift-huggingface
Swift Jinja: chat-template rendering used transitively by Swift Transformers. Apache-2.0. https://github.com/huggingface/swift-jinja
EventSource: server-sent-events support used transitively by Swift Hugging Face. MIT. https://github.com/mattt/EventSource
Swift Collections, Swift Crypto and Swift ASN.1: Apple packages used transitively by the tokenizer/Hub stack. Apache-2.0. https://github.com/apple/swift-collections, https://github.com/apple/swift-crypto, https://github.com/apple/swift-asn1
yyjson: JSON parsing support used transitively by Swift Transformers. MIT. https://github.com/ibireme/yyjson
Swift Numerics: numeric types used by MLX Swift. Apache-2.0. https://github.com/apple/swift-numerics
Swift Argument Parser: MLX Swift build-tool support. Apache-2.0. https://github.com/apple/swift-argument-parser
Iconoir: interface icons. MIT. https://iconoir.com/
Tabler Icons: complementary interface icons, normalized to Wisp's 1.5-point line style. MIT. https://github.com/tabler/tabler-icons
Lobe Icons: AI provider and model logos. MIT; individual brand rules still apply. https://github.com/lobehub/lobe-icons
Notion and Obsidian marks: official product marks shown only to identify user-initiated export destinations. The marks remain owned by their respective owners; inclusion does not imply sponsorship or affiliation. https://www.notion.so/, https://obsidian.md/
whisper.cpp: optional on-device Whisper transcription runtime. MIT. https://github.com/ggml-org/whisper.cpp
llama.cpp: optional on-device GGUF local LLM runtime. MIT. https://github.com/ggml-org/llama.cpp
ONNX Runtime 1.24.2: bundled runtime for on-device inference with the optional multilingual MiniLM INT8 association model. MIT; the original license and third-party notices are linked below. https://github.com/microsoft/onnxruntime

Optional downloaded models

Downloaded model files are not bundled with Wisp. They are user-selected assets stored in Application Support. The downloader may automatically choose between pinned official endpoints and verified mirrors, but every installation uses the same expected SHA256 verification.

Whisper models: OpenAI Whisper / ggerganov whisper.cpp distribution, generally MIT.
Qwen GGUF models: Qwen official or GGUF distribution, commonly Apache-2.0.
DeepSeek Distill GGUF models: DeepSeek or GGUF distributors; verify model page license.
Llama GGUF models: Meta Llama community license.
Gemma GGUF models: Google Gemma terms.
Hy-MT2 Q4 GGUF: Tencent Hunyuan distribution, Apache-2.0. Optional public download for direct Chinese, English, and Japanese translation.
Qwen3.5 0.8B GGUF: Qwen / ggml-org distribution, Apache-2.0.
Qwen3 0.6B MLX 4-bit: Qwen / MLX Community distribution, Apache-2.0. Its config, tokenizer and weight files are verified as one atomic model directory.
LFM2.5 1.2B GGUF: Liquid AI distribution, LFM Open License 1.0.
multilingual MiniLM INT8: optional association model, listed as Apache-2.0 in the model manifest; use is subject to the linked distribution terms. https://huggingface.co/sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
EmbeddingGemma 300M: optional association model governed by the Gemma Terms of Use, not MIT or Apache-2.0. Explicit consent is still required before the first download; the original notices below include the EmbeddingGemma notice. https://ai.google.dev/gemma/terms

Model manifest requirements

Every production model manifest entry should include name, version, format, download URL, file size, SHA256, license, capability, minimum device recommendation, recommendation level, and update notes. Models without verified checksums should not be marked production-ready.

Fonts

Share cards and the wordmark retain LXGW WenKai, Fusion Pixel, Newsreader, Playfair Display, Courier Prime, and IBM Plex Mono. These six families use SIL Open Font License 1.1; original licenses, copyrights and Reserved Font Name terms are retained. The Fusion Pixel file is byte-identical to the proportional zh_hans WOFF2 in the official 2024.05.12 release, whose upstream notices are also included. The fixed Wisp pixel wordmark is vector artwork generated from the official Fusion Pixel version described above. It retains a pixel style; its source font uses SIL Open Font License 1.1, which permits logo and graphic design uses. Mathematical typesetting also uses the math fonts included with swiftui-math, with their original license and copyright notices. The Playfair Display and IBM Plex Mono subsets are renamed Wisp Editorial Serif and Wisp Mono, respectively; only font names change, with glyphs and layout data preserved. Original project names identify their sources.

Original license texts

The original licenses and copyright notices are retained below in their original languages. Separate model licenses and trademark terms still apply.