v-modal/vmodal_sdk_swift_iphoneduo

VModal SDK for Apple platforms (Swift)

What it is – An open‑source Swift package that lets iOS and macOS apps talk to the VModal cloud service. The service provides AI‑powered, multimodal video search (semantic, speech‑to‑text, OCR, image‑based) and upload/ indexing capabilities. The SDK handles authentication, request/response models, streaming uploads, progress reporting, cancellation, and typed results, while keeping the UI completely in the developer’s app.

Key capabilities

• Semantic video & image search – Find moments by natural‑language description, spoken words, on‑screen text, or visual objects.
• ASR & OCR indexing – The backend extracts speech and text from uploaded videos, making them searchable.
• Streaming uploads – Upload large video files directly from a URL without loading the whole file into memory; progress is exposed via an AsyncStream and can be cancelled.
• Scoped organization – Projects, collections, and streams are identified by simple strings (projectID, collectionName, streamName). All operations are scoped to a specific collection/stream.
• Runtime API‑key handling – Supply a bearer token at runtime, rotate it, or clear it without rebuilding the app; the SDK never shows a login UI.
• Swift‑native API – Fully typed, async/await‑compatible, works with Swift 6.0 and Xcode 26.6+, targeting iOS 16+ and macOS 13+.

Typical usage flow

1. Add the package in Xcode via the URL https://github.com/v-modal/vmodal_sdk_swift_iphoneduo.
2. Configure a project with an API key:

   let keys = try MutableAPIKeyProvider("runtime-api-key")
   let project = try VModal.configure(projectID: "my_app", apiKeyProvider: keys)
   let scope = try project.scope(collectionName: "user_42", streamName: "uploads")
   

3. Upload a video:

   let source = try UploadSource(fileURL: videoURL)
   let task = scope.upload(source)
   for await p in task.progress { print("\(p.percent)%") }
   let result = try await task.result
   

4. Search with natural language or metadata:

   let results = try await scope.search(
       "cyclist in red jacket crossing the bridge",
       options: ScopedSearchOptions(searchSources: ["image"], limit: 20)
   )
   

5. Handle cancellation – Call task.cancel() or cancel the async search task when the view disappears.

Examples & tooling

• StarterIOS – A minimal app that demonstrates configuration, collection discovery, upload, indexing, and search.
• Framebase example – Shows how to browse a local video archive, prepare it, and play search results.
• SDKSimulation tool – Simulates each integration stage for debugging.
• Docs – Full API reference and getting‑started guide are generated with DocC (Sources/VModalSDK/VModalSDK.docc).

Supported platforms

• iOS (SwiftUI or UIKit) – ✅
• macOS – ✅
• watchOS, tvOS, visionOS – ❌ (not in this release)
• Other platforms – ❌ (use the corresponding VModal SDK for those environments)

Prerequisites

• Swift 6.0, Xcode 26.6 or newer
• iOS 16+ / macOS 13+
• A VModal API key (request via https://v-modal.com/page/contact.ts)

Development workflow

git clone https://github.com/v-modal/vmodal_sdk_swift_iphoneduo.git
cd vmodal_sdk_swift_iphoneduo
bash install.sh check          # verify toolchain & dependencies
bash build.sh analyze          # static analysis
bash test.sh test              # unit tests (offline)
# For live integration tests (requires credentials)
bash test.sh live

The repo also provides helper scripts to list simulators, pick a device, and launch the example app.

License – MIT.

---

Bottom line – If you need to add AI‑driven video search or searchable video uploads to an Apple app, VModal’s Swift SDK gives you a small, type‑safe wrapper around the VModal cloud service, handling all the heavy‑lifting (vector indexes, speech‑to‑text, image recognition) on the backend while you stay in pure Swift/SwiftUI.