Apple-first local search and semantic retrieval for Swift apps.

• Overview
• Quick Start
• Usage
• Package Status
• Release Notes
• License

SwiftlyFetch is now shelved indefinitely, along with other simultaneous projects made sustainable by cheap, automated assistance which I can no longer afford. Should the economy and job market ever improve, I'll pick this back up.

SwiftlyFetch is a Swift package family for apps that need to index local documents, search them, and assemble useful retrieval context without sending the job to a remote service. The current package is early, but useful: it ships semantic retrieval through RAGCore and RAGKit, conventional search through FetchCore and FetchKit, and the first umbrella SwiftlyFetch facade for one-corpus ingestion.

Use SwiftlyFetch when you want:

• local semantic retrieval over plain text and markdown
• deterministic hashing embeddings for tests, previews, and examples
• Apple Natural Language embeddings on supported Apple platforms
• persistent semantic chunks and embeddings through Core Data
• coordinated one-corpus ingestion through SwiftlyFetchLibrary
• conventional document search with title/body evidence, query-aware snippets, and a macOS SearchKit-backed index path
• one package family that keeps retrieval, indexing, and context assembly separate from chat, generation, agents, and remote-provider workflows

The package family is intentionally split by job:

• RAGCore defines semantic retrieval vocabulary.
• RAGKit provides the default semantic retrieval implementation and KnowledgeBase actor.
• FetchCore defines portable conventional-search models.
• FetchKit provides the first conventional-search facade, Core Data document storage, pending index-sync tracking, and a macOS SearchKit backend.
• SwiftlyFetch composes both sibling package families so callers can add a document once, then use conventional search and semantic retrieval over the same corpus.

SwiftlyFetch has tagged releases stable enough to try locally, and the umbrella SwiftlyFetch surface is available in the current codebase. See GitHub Releases for the latest published version details.

SwiftlyFetch is not a chat framework, LLM SDK, agent runtime, or remote-provider abstraction. Its job is local retrieval: document preparation, indexing, search, filtering, and context assembly.

Add SwiftlyFetch to your Package.swift dependencies:

.package(url: "https://github.com/gaelic-ghost/SwiftlyFetch", from: "0.2.0"),

Then add the library product to your target dependencies:

.product(name: "SwiftlyFetch", package: "SwiftlyFetch"),

The package is still a bit early, but the retrieval surface is real enough to try locally. For the coordinated corpus surface, import SwiftlyFetch:

import FetchCore
import RAGCore
import SwiftlyFetch

let library = SwiftlyFetchLibrary.default()

try await library.addDocument(
    FetchDocumentRecord(
        id: "guide",
        title: "Fruit Guide",
        body: "Apples are bright and crisp.",
        contentType: .markdown,
        kind: .guide,
        language: "en"
    )
)

let searchResults = try await library.search(FetchSearchQuery("fruit guide"))
let semanticResults = try await library.retrieve(SearchQuery("bright fruit"))

For lower-level semantic retrieval, import RAGCore and RAGKit directly:

import RAGCore
import RAGKit

let kb = KnowledgeBase.hashingDefault()

try await kb.addDocument(
    Document(
        id: "guide",
        content: .markdown(
            """
            # Fruit Guide

            ## Apples

            Apples are bright and crisp.
            """
        )
    )
)

let results = try await kb.search("bright fruit")
let context = try await kb.makeContext(for: "bright fruit")

The current public surface centers on five library products: RAGCore, RAGKit, FetchCore, FetchKit, and SwiftlyFetch.

For coordinated one-corpus ingestion, use SwiftlyFetchLibrary from SwiftlyFetch:

import FetchCore
import RAGCore
import SwiftlyFetch

let library = SwiftlyFetchLibrary.default()

let mutation = try await library.addDocument(
    FetchDocumentRecord(
        id: "guide",
        title: "Apple Guide",
        body: "Apples are bright and crisp.",
        contentType: .markdown
    )
)

let conventionalResults = try await library.search(FetchSearchQuery("apple guide"))
let semanticResults = try await library.retrieve(SearchQuery("bright crisp"))
let sideBySideResults = try await library.searchAndRetrieve(
    conventional: FetchSearchQuery("apple guide"),
    semantic: SearchQuery("bright crisp")
)

SwiftlyFetchMutationResult reports conventional and semantic outcomes separately. If the corpus write succeeds but semantic indexing fails, the facade queues a semantic retry instead of pretending the whole write failed. retrySemanticIndexing(limit:) respects retry cooldowns through nextRetryAt and reports deferred document IDs separately from completed, missing, and failed retries. searchAndRetrieve(...) returns conventional and semantic results side by side without combining scores; ranked hybrid search remains future work.

For semantic retrieval, use KnowledgeBase from RAGKit:

import RAGCore
import RAGKit

let localKB = KnowledgeBase.hashingDefault()
let appleKB = try await KnowledgeBase.naturalLanguageDefault(languageHint: "en")
let semanticStore = FileManager.default
    .temporaryDirectory
    .appendingPathComponent("SwiftlyFetchSemantic.sqlite")
let persistentKB = try await KnowledgeBase.persistentHashingDefault(
    configuration: .init(store: .sqlite(semanticStore))
)

For conventional search, use FetchKitLibrary from FetchKit:

import FetchCore
import FetchKit

let library = FetchKitLibrary()

try await library.addDocument(
    FetchDocumentRecord(
        id: "guide",
        title: "Apple Guide",
        body: "Apples are bright and crisp.",
        contentType: .markdown,
        kind: .guide,
        language: "en"
    )
)

let results = try await library.search("apple guide")
let firstResult = results.first
let matchedFields = firstResult?.matchedFields
let snippetField = firstResult?.snippetField
try await library.removeDocuments(withIDs: ["guide"])

matchedFields identifies every indexed field that contributed to a search result. snippetField identifies the field used to build the returned snippet. Simple result lists can show why a result appeared immediately, while richer UIs can render title evidence differently from body evidence. Mutation results report the affected document IDs, including removeAllDocuments(), so app code can update local UI state without re-deriving which records changed.

On macOS, the persistent conventional-search surface is now also shaped around one library storage location instead of separate store and index URLs:

import FetchKit

let persistentLibrary = try await FetchKitLibrary.macOSPersistentLibrary()
let previewLibrary = try await FetchKitLibrary.macOSPersistentLibrary(
    at: URL(fileURLWithPath: "/tmp/SwiftlyFetchPreview", isDirectory: true)
)

If a caller needs raw markdown link destinations for downstream indexing or fetch-oriented work, opt in at the chunker boundary instead of widening default chunk text:

import RAGKit

let chunker = HeadingAwareMarkdownChunker(linkDestinationMetadataMode: .include)

Current defaults:

• plain text uses paragraph chunking
• markdown uses parser-backed heading-aware chunking
• markdown link destinations stay out of chunk text by default, but HeadingAwareMarkdownChunker(linkDestinationMetadataMode: .include) can record raw destinations in chunk metadata when downstream indexing or fetch-oriented work needs them
• hashingDefault() gives a synchronous deterministic local path for tests and examples
• naturalLanguageDefault() uses the Apple Natural Language backend on supported platforms and remains throwing because backend setup can fail
• persistentHashingDefault(configuration:dimension:) and persistentNaturalLanguageDefault(configuration:languageHint:) use the same retrieval defaults with a Core Data-backed semantic vector index
• SwiftlyFetchLibrary.default() is synchronous and non-throwing because it uses deterministic in-memory defaults; persistent constructors remain async and throwing
• metadata filtering supports explicit exclusions, ordered comparisons for int, double, and date, plus case-insensitive startsWith and endsWith string matching
• markdown list items keep heading and immediate lead-in context in chunk text, and also carry structured chunk metadata for list kind, lead-in, ordinal, and heading path
• markdown block quotes stay secondary by default, but are promoted into the primary retrieval stream when they make up more than one third of the document's chunkable block structure
• markdown code blocks stay secondary by default, but promoted code-heavy documents emit code chunks with language metadata, and all chunks from a document can carry document-level code-language metadata
• markdown thematic breaks act as section-boundary hints, carrying a short lead-in into the next chunk instead of becoming their own retrieval chunks
• markdown images keep alt text primary in chunk text while recording image references as chunk metadata, and whitelisted HTML blocks currently cover img plus details / summary
• markdown fallback is selective: ordinary supported prose still chunks normally, but policy-rejected markdown like unsupported raw-HTML-only or reference-definition-only content does not fall back through the plain paragraph chunker
• conventional search now uses modest field-aware ranking, prefers title hits over body-only hits when both are relevant, and builds query-aware snippets with multi-term highlights instead of a single fixed-width first-term window
• in-memory all-term search gives a small boost to compact evidence, so focused passages rank ahead of scattered near-matches when they satisfy the same query terms
• conventional-search results report matchedFields and snippetField, keeping title-only snippets visible while letting consumers distinguish title evidence from body evidence
• makeContext(...) suppresses redundant same-document chunk text, groups annotated output by document, and skips annotated sections that only have room for labels

SwiftlyFetch is usable today as a local Apple-first package family, but it is still early in the broader product arc.

Good current fits:

• app-level semantic retrieval over local plain-text and markdown corpora
• conventional-search experimentation through FetchCore and FetchKit
• Apple-first local search prototypes where Core Data, SearchKit, and on-device retrieval matter
• downstream UI or model features that need ranked search results or assembled context, but do not need SwiftlyFetch to own generation

Current constraints:

• the SearchKit backend is macOS-first
• Natural Language asset-backed verification runs in local maintainer validation by default, but stays out of the default GitHub-hosted CI lane because hosted macOS still stalls in the asset-backed step
• the package family direction is broader than the currently shipped polished surface, especially on the FetchKit side
• hybrid search still waits on follow-up result-shape work; the umbrella facade currently exposes conventional search and semantic retrieve separately
• conventional-search quality coverage uses a small checked-in Project Gutenberg fixture corpus plus synthetic near-miss and longer-body records; larger app-like corpora are still future validation work

For contributor setup, branch workflow, verification commands, and review expectations, use CONTRIBUTING.md. Maintainer planning and architecture notes live under docs/maintainers/.

See the repository's GitHub releases for published package notes. Release workflow details belong in CONTRIBUTING.md and the maintainer docs, not in this user-facing README.

Licensed under the Apache License 2.0. See LICENSE.