Expand description
Domain types shared across the crate: document representations, CLI arguments, search results, and provider enums.
Structs§
- Attached
Document - An externally attached document — parsed text injected into the RAG prompt alongside (or instead of) vector store results. Not added to chat memory.
- Chat
Config - Chat / generation configuration.
- Chunk
Config - Configuration for text chunking — size and overlap in tokens.
- Chunk
Meta - Positional metadata attached to every chunk for source citation.
- Document
Chunk - A single text chunk from a document, tagged with its source file and optional positional metadata.
- Embed
Config - Embedding / retrieval configuration.
- File
Hash Entry - Metadata for a document file: filename + SHA-256 hash.
- Generation
Params - Model generation parameters shared across all providers.
- Index
Manifest - Reproducibility manifest recorded alongside the vector index.
- Memory
Config - Memory / query-rewrite configuration.
- Paper
Result - A paper result from academic search APIs (Semantic Scholar, arXiv).
- Parse
Config - Document parsing and chunking configuration.
- Ragrig
Config - Top-level configuration for a ragrig session.
- Rank
Score - Ranker-agnostic score produced by hybrid search.
- Source
File - A document source filename, used consistently across the pipeline.
Enums§
- Context
Size Mode - Controls how context-size overflow is handled.
- Document
Type - A document file on disk, tagged with its format and path.
- Embedding
Provider - Embedding backend: local Ollama or CPU-only Fastembed.
- Epub
Parser Backend - EPUB parser backend.
- PdfParser
Backend - PDF parser backend.
- Provider
- Chat backend: local Ollama or cloud DeepSeek. Chat generation provider.