feat(minor): reference common classes on the Llama instance (#360)

Author: giladgd

docs/guide/downloading-models.md

@@ -87,7 +87,7 @@ When using a URI to reference a model,
 it's recommended [to add it to your `package.json` file](#cli) to ensure it's downloaded when running `npm install`,
 and also resolve it using the [`resolveModelFile`](../api/functions/resolveModelFile.md) method to get the full path of the resolved model file.
 
-Here's and example usage of the [`resolveModelFile`](../api/functions/resolveModelFile.md) method:
+Here's an example usage of the [`resolveModelFile`](../api/functions/resolveModelFile.md) method:
 ```typescript
 import {fileURLToPath} from "url";
 import path from "path";

src/bindings/Llama.ts

@@ -7,6 +7,7 @@ import {GbnfJsonSchema} from "../utils/gbnfJson/types.js";
 import {LlamaJsonSchemaGrammar} from "../evaluator/LlamaJsonSchemaGrammar.js";
 import {LlamaGrammar, LlamaGrammarOptions} from "../evaluator/LlamaGrammar.js";
 import {ThreadsSplitter} from "../utils/ThreadsSplitter.js";
+import {getLlamaClasses, LlamaClasses} from "../utils/getLlamaClasses.js";
 import {BindingModule} from "./AddonTypes.js";
 import {BuildGpu, BuildMetadataFile, LlamaGpuType, LlamaLocks, LlamaLogLevel} from "./types.js";
 import {MemoryOrchestrator, MemoryReservation} from "./utils/MemoryOrchestrator.js";
@@ -56,6 +57,7 @@ export class Llama {
     /** @internal */ private _nextLogNeedNewLine: boolean = false;
     /** @internal */ private _disposed: boolean = false;
 
+    private _classes?: LlamaClasses;
     public readonly onDispose = new EventRelay<void>();
 
     private constructor({
@@ -137,6 +139,13 @@ export class Llama {
         return this._disposed;
     }
 
+    public get classes() {
+        if (this._classes == null)
+            this._classes = getLlamaClasses();
+
+        return this._classes;
+    }
+
     public get gpu() {
         return this._gpu;
     }

src/index.ts

@@ -100,6 +100,7 @@ import {
 import {GgmlType, type GgufTensorInfo} from "./gguf/types/GgufTensorInfoTypes.js";
 import {type ModelFileAccessTokens} from "./utils/modelFileAccesTokens.js";
 import {type OverridesObject} from "./utils/OverridesObject.js";
+import type {LlamaClasses} from "./utils/getLlamaClasses.js";
 
 
 export {
@@ -108,6 +109,7 @@ export {
     type LlamaOptions,
     type LastBuildOptions,
     type LlamaGpuType,
+    type LlamaClasses,
     LlamaLogLevel,
     NoBinaryFoundError,
     resolveModelFile,

src/utils/getLlamaClasses.ts

@@ -0,0 +1,22 @@
+import {LlamaChatSession} from "../evaluator/LlamaChatSession/LlamaChatSession.js";
+import {LlamaChat} from "../evaluator/LlamaChat/LlamaChat.js";
+import {LlamaCompletion} from "../evaluator/LlamaCompletion.js";
+
+export type LlamaClasses = {
+    readonly LlamaChatSession: typeof LlamaChatSession,
+    readonly LlamaChat: typeof LlamaChat,
+    readonly LlamaCompletion: typeof LlamaCompletion
+};
+
+let cachedClasses: LlamaClasses | undefined = undefined;
+
+export function getLlamaClasses(): LlamaClasses {
+    if (cachedClasses == null)
+        cachedClasses = Object.seal({
+            LlamaChatSession,
+            LlamaChat,
+            LlamaCompletion
+        });
+
+    return cachedClasses;
+}