How ibz-04/quaynor works
Quaynor is a Rust-based local inference engine with language and mobile bindings. The principal workflow loads a GGUF model, builds a chat session, and streams generated output; optional capabilities include embeddings, reranking, grammar-constrained tools, and model downloads. The graph includes the documented binding surfaces and sampled core layers. Binding-to-engine links reflect README documentation; unsampled binding internals and some runtime wiring are not inferred.
Main entry points: chat.rs Chat session · llm.rs Model loading · lib.rs Flutter API · chat.ts React Native chat
Components 22
Inference engine
- Model loading
quaynor/core/src/llm.rs - Chat session
quaynor/core/src/chat.rs - Context planning
quaynor/core/src/memory.rs - Chat templates
quaynor/core/src/template.rs - Tokenization
quaynor/core/src/tokenizer.rs
Generation features
- Sampling config
quaynor/core/src/sampler_config.rs - Embeddings
quaynor/core/src/encoder.rs - Cross-encoder
quaynor/core/src/crossencoder.rs
Grammar and tools
- Tool format handlers
quaynor/core/src/tool_calling - Grammar builder
quaynor/grammar/gbnf/src/builder.rs - Schema conversion
quaynor/grammar/gbnf/src/json.rs
Python binding
- Python API
quaynor/python/src/lib.rs
Mobile bindings
- Flutter API
quaynor/flutter/rust/src/lib.rs - Flutter bridge
quaynor/flutter/rust/src/frb_generated.rs - React Native chat
quaynor/react-native/src/chat.ts - React Native model
quaynor/react-native/src/model.ts
Apple and JVM bindings
- Swift API
quaynor/swift/Sources/Quaynor/Chat.swift - Swift FFI
quaynor/swift/Sources/QuaynorFFI/QuaynorFFI.swift - Kotlin API
quaynor/kotlin/common/src/Chat.kt - UniFFI interface
quaynor/uniffi/src/lib.rs
Other components
- App developer
- Model URL or Hugging Face
Connections 28
- App developer → Model loading: chooses model
README.md - Model loading → Chat session: builds session
README.md - Chat session → App developer: streams response
README.md - Model loading → Model URL or Hugging Face: downloads model
README.md - Chat session → Context planning: plans context
quaynor/core/src/chat.rs - Chat session → Chat templates: renders prompts
quaynor/core/src/chat.rs - Chat session → Tokenization: tokenizes prompt
quaynor/core/src/chat.rs - Model loading → Context planning: plans context
quaynor/core/src/llm.rs - Chat session → Sampling config: uses configuration
quaynor/core/src/chat.rs - App developer → Embeddings: requests embeddings
README.md - App developer → Cross-encoder: requests reranking
README.md - Chat session → Tool format handlers: dispatches tool calls
quaynor/core/src/chat.rs - Chat session → Grammar builder: constrains generation
README.md - Grammar builder → Schema conversion: converts schemas
quaynor/grammar/gbnf/src/json.rs - App developer → Python API: uses binding
README.md - Python API → Chat session: invokes engine
README.md - App developer → Flutter API: uses binding
README.md - Flutter API → Chat session: invokes engine
README.md - Flutter API → Flutter bridge: exposes bridge calls
quaynor/flutter/rust/src/lib.rs - App developer → React Native chat: uses binding
README.md - React Native chat → React Native model: loads model
quaynor/react-native/src/chat.ts - React Native chat → Chat session: invokes engine
README.md - App developer → Swift API: uses binding
README.md - Swift API → Swift FFI: calls FFI
quaynor/swift/Sources/QuaynorFFI/QuaynorFFI.swift - Swift FFI → Chat session: invokes engine
README.md - App developer → Kotlin API: uses binding
README.md - Kotlin API → UniFFI interface: uses UniFFI
README.md - UniFFI interface → Chat session: invokes engine
README.md