Updated swift tools, refactor

Author: srgtuszy

Package.swift

@@ -1,4 +1,4 @@
-// swift-tools-version: 5.10.1
+// swift-tools-version: 6.0
 import PackageDescription
 
 let package = Package(
@@ -17,7 +17,7 @@ let package = Package(
     ],
     dependencies: [
         .package(url: "https://github.com/ggerganov/llama.cpp", branch: "master"),
-        .package(url: "https://github.com/apple/swift-log.git", from: "1.0.0"),
+        .package(url: "https://github.com/apple/swift-log.git", from: "1.6.1"),
     ],
     targets: [
         .target(

Sources/llama-cpp-swift/InferError.swift

@@ -1,8 +1,10 @@
-public struct InferError: Error {
+public struct InferError: Error, Sendable {
     public let message: String
     public let code: Code
 
-    public enum Code: Int {
+    public enum Code: Int, Sendable {
         case cancelled = 1
+        case kvCacheFailure
+        case decodingFailure
     }
 }

Sources/llama-cpp-swift/InitializationError.swift

@@ -1,8 +1,8 @@
-struct InitializationError: Error {
+struct InitializationError: Error, Sendable {
     let message: String
     let code: Code
 
-    enum Code: Int {
+    enum Code: Int, Sendable {
         case failedToLoadModel = 1
         case failedToInitializeContext
     }

Sources/llama-cpp-swift/LLama.swift

@@ -2,7 +2,7 @@ import Foundation
 import Logging
 import llama
 
-public actor LLama {
+public class LLama {
   private let logger = Logger.llama
   private let model: OpaquePointer
   private let context: OpaquePointer
@@ -74,7 +74,7 @@ public actor LLama {
   // MARK: - Inference
 
   public func infer(prompt: String, maxTokens: Int32 = 128) async throws -> String {
-    completionInit(text: prompt)
+    try completionInit(text: prompt)
     var generatedText = ""
 
     while !isDone && nCur < nLen && nCur - batch.n_tokens < maxTokens {
@@ -109,7 +109,7 @@ public actor LLama {
     batch.n_tokens += 1
   }
 
-  private func completionInit(text: String) {
+  private func completionInit(text: String) throws {
     logger.debug("Attempting to complete \"\(text)\"")
 
     tokensList = tokenize(text: text, add_bos: true)
@@ -121,7 +121,8 @@ public actor LLama {
     logger.debug("\nn_len = \(self.nLen), n_ctx = \(nCtx), n_kv_req = \(nKvReq)")
 
     if nKvReq > nCtx {
-      print("Error: n_kv_req > n_ctx, the required KV cache size is not big enough")
+      logger.error("Error: n_kv_req > n_ctx, the required KV cache size is not big enough")
+      throw InferError(message: "KV cache too small", code: .kvCacheFailure)
     }
 
     batch.clear()
@@ -134,7 +135,7 @@ public actor LLama {
     }
 
     if llama_decode(context, batch) != 0 {
-      print("llama_decode() failed")
+      throw InferError(message: "llama_decode failed", code: .decodingFailure)
     }
 
     nCur = batch.n_tokens
@@ -230,8 +231,8 @@ public actor LLama {
   }
 }
 
-private extension llama_batch {
-  mutating func clear() {
+extension llama_batch {
+  fileprivate mutating func clear() {
     n_tokens = 0
   }
 }

Sources/llama-cpp-swift/Logger+LLama.swift

@@ -1,5 +1,7 @@
 import Logging
 
 extension Logger {
-  static var llama = Logger(label: "llama-cpp-swift")
+    static var llama: Logger {
+        Logger(label: "llama-cpp-swift")
+    }
 }

example/Package.swift

@@ -1,4 +1,4 @@
-// swift-tools-version: 5.10
+// swift-tools-version: 6.0
 import PackageDescription
 
 let package = Package(