wllama-service 1.0.2 → 2.0.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.js +10 -5
- package/dist/index.js.map +1 -1
- package/dist/index.mjs +10 -5
- package/dist/index.mjs.map +1 -1
- package/package.json +7 -4
- package/scripts/postinstall.js +26 -0
package/dist/index.js
CHANGED
|
@@ -74,7 +74,8 @@ var WllamaService = class {
|
|
|
74
74
|
return { success: true, usedWebGPU: hasWebGPU };
|
|
75
75
|
} catch (e) {
|
|
76
76
|
await this.unload();
|
|
77
|
-
|
|
77
|
+
const message = e instanceof Error ? e.message : "Failed to load model";
|
|
78
|
+
return { success: false, error: message };
|
|
78
79
|
}
|
|
79
80
|
}
|
|
80
81
|
/** Generate a response from the loaded model */
|
|
@@ -101,7 +102,8 @@ var WllamaService = class {
|
|
|
101
102
|
timeMs: Math.round(performance.now() - t0)
|
|
102
103
|
};
|
|
103
104
|
} catch (e) {
|
|
104
|
-
|
|
105
|
+
const message = e instanceof Error ? e.message : "Generation failed";
|
|
106
|
+
return { success: false, error: message };
|
|
105
107
|
}
|
|
106
108
|
}
|
|
107
109
|
/** Unload the current model and free memory */
|
|
@@ -132,9 +134,12 @@ var WllamaService = class {
|
|
|
132
134
|
async storeInCache(file) {
|
|
133
135
|
const url = `${window.location.origin}/wllama-local-models/${file.name}`;
|
|
134
136
|
const cache = await caches.open("wllama-local-models");
|
|
135
|
-
await cache.put(
|
|
136
|
-
|
|
137
|
-
|
|
137
|
+
await cache.put(
|
|
138
|
+
url,
|
|
139
|
+
new Response(file, {
|
|
140
|
+
headers: { "Content-Type": "application/octet-stream" }
|
|
141
|
+
})
|
|
142
|
+
);
|
|
138
143
|
}
|
|
139
144
|
async deleteFromCache(fileName) {
|
|
140
145
|
const url = `${window.location.origin}/wllama-local-models/${fileName}`;
|
package/dist/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/index.ts","../src/WllamaService.ts"],"sourcesContent":["export { WllamaService } from './WllamaService';\nexport type {\n WllamaServiceConfig,\n GenerateRequest,\n GenerateResult,\n LoadModelResult,\n ConnectionResult,\n} from './types';","import { Wllama } from '@wllama/wllama/esm/index.js';\nimport type {\n WllamaServiceConfig,\n GenerateRequest,\n GenerateResult,\n LoadModelResult,\n ConnectionResult,\n} from './types';\n\nexport class WllamaService {\n private wllama: Wllama | null = null;\n private modelName: string = '';\n private config: Required<WllamaServiceConfig>;\n\n constructor(config: WllamaServiceConfig = {}) {\n this.config = {\n wasmPath: config.wasmPath ?? '/wllama/wllama.wasm',\n nGpuLayers: config.nGpuLayers ?? 999,\n nCtx: config.nCtx ?? 2048,\n };\n }\n\n /** Check browser compatibility and capabilities */\n checkEnvironment(): ConnectionResult {\n if (typeof WebAssembly === 'undefined') {\n return { success: false, error: 'WebAssembly is not supported in this browser.' };\n }\n if (!('caches' in window)) {\n return { success: false, error: 'Cache API is not supported in this browser.' };\n }\n const hasWebGPU = !!(navigator as any).gpu;\n const isolated = window.crossOriginIsolated;\n return {\n success: true,\n hasWebGPU,\n crossOriginIsolated: isolated,\n };\n }\n\n /** Load a GGUF model file from the user's filesystem */\n async loadModel(\n file: File,\n onProgress?: (progress: number) => void,\n ): Promise<LoadModelResult> {\n const env = this.checkEnvironment();\n if (!env.success) return { success: false, error: env.error };\n\n await this.unload();\n\n try {\n onProgress?.(10);\n\n const configPaths = { default: this.config.wasmPath };\n this.wllama = new Wllama(configPaths
|
|
1
|
+
{"version":3,"sources":["../src/index.ts","../src/WllamaService.ts"],"sourcesContent":["export { WllamaService } from './WllamaService';\nexport type {\n WllamaServiceConfig,\n GenerateRequest,\n GenerateResult,\n LoadModelResult,\n ConnectionResult,\n} from './types';","import { Wllama } from '@wllama/wllama/esm/index.js';\nimport type {\n WllamaServiceConfig,\n GenerateRequest,\n GenerateResult,\n LoadModelResult,\n ConnectionResult,\n} from './types';\n\nexport class WllamaService {\n private wllama: Wllama | null = null;\n private modelName: string = '';\n private config: Required<WllamaServiceConfig>;\n\n constructor(config: WllamaServiceConfig = {}) {\n this.config = {\n wasmPath: config.wasmPath ?? '/wllama/wllama.wasm',\n nGpuLayers: config.nGpuLayers ?? 999,\n nCtx: config.nCtx ?? 2048,\n };\n }\n\n /** Check browser compatibility and capabilities */\n checkEnvironment(): ConnectionResult {\n if (typeof WebAssembly === 'undefined') {\n return { success: false, error: 'WebAssembly is not supported in this browser.' };\n }\n if (!('caches' in window)) {\n return { success: false, error: 'Cache API is not supported in this browser.' };\n }\n const hasWebGPU = !!(navigator as any).gpu;\n const isolated = window.crossOriginIsolated;\n return {\n success: true,\n hasWebGPU,\n crossOriginIsolated: isolated,\n };\n }\n\n /** Load a GGUF model file from the user's filesystem */\n async loadModel(\n file: File,\n onProgress?: (progress: number) => void,\n ): Promise<LoadModelResult> {\n const env = this.checkEnvironment();\n if (!env.success) return { success: false, error: env.error };\n\n await this.unload();\n\n try {\n onProgress?.(10);\n\n // AssetsPathConfig requires `default` — this is the key the\n // Wllama runtime actually reads (confirmed against the real\n // @wllama/wllama@3.5.1 implementation, not just its docs).\n // No `as any` needed: this shape is type-correct as written.\n const configPaths = { default: this.config.wasmPath };\n this.wllama = new Wllama(configPaths);\n onProgress?.(30);\n\n await this.storeInCache(file);\n onProgress?.(50);\n\n const hasWebGPU = !!(navigator as any).gpu;\n await (this.wllama as any).loadModel([file], {\n n_ctx: this.config.nCtx,\n ...(hasWebGPU && this.config.nGpuLayers > 0\n ? { n_gpu_layers: this.config.nGpuLayers }\n : {}),\n });\n\n this.modelName = file.name;\n onProgress?.(100);\n\n return { success: true, usedWebGPU: hasWebGPU };\n } catch (e: unknown) {\n await this.unload();\n const message = e instanceof Error ? e.message : 'Failed to load model';\n return { success: false, error: message };\n }\n }\n\n /** Generate a response from the loaded model */\n async generate(req: GenerateRequest): Promise<GenerateResult> {\n if (!this.wllama) {\n return { success: false, error: 'No model loaded. Call loadModel() first.' };\n }\n\n try {\n const t0 = performance.now();\n\n const messages: { role: string; content: string }[] = [];\n if (req.system) messages.push({ role: 'system', content: req.system });\n messages.push({ role: 'user', content: req.prompt });\n\n const response = await (this.wllama as any).createChatCompletion({\n messages,\n max_tokens: req.maxTokens ?? 512,\n temperature: req.temperature ?? 0.7,\n top_k: req.topK ?? 40,\n top_p: req.topP ?? 0.95,\n });\n\n const text: string = response?.choices?.[0]?.message?.content ?? '';\n return {\n success: true,\n text,\n timeMs: Math.round(performance.now() - t0),\n };\n } catch (e: unknown) {\n const message = e instanceof Error ? e.message : 'Generation failed';\n return { success: false, error: message };\n }\n }\n\n /** Unload the current model and free memory */\n async unload(): Promise<void> {\n if (this.wllama) {\n try {\n await (this.wllama as any).exit();\n } catch (_) {\n /* ignore */\n }\n this.wllama = null;\n }\n if (this.modelName) {\n try {\n await this.deleteFromCache(this.modelName);\n } catch (_) {\n /* ignore */\n }\n this.modelName = '';\n }\n }\n\n /** Whether a model is currently loaded */\n get isLoaded(): boolean {\n return this.wllama !== null;\n }\n\n /** Name of the currently loaded model */\n get currentModel(): string {\n return this.modelName;\n }\n\n private async storeInCache(file: File): Promise<void> {\n const url = `${window.location.origin}/wllama-local-models/${file.name}`;\n const cache = await caches.open('wllama-local-models');\n await cache.put(\n url,\n new Response(file, {\n headers: { 'Content-Type': 'application/octet-stream' },\n }),\n );\n }\n\n private async deleteFromCache(fileName: string): Promise<void> {\n const url = `${window.location.origin}/wllama-local-models/${fileName}`;\n const cache = await caches.open('wllama-local-models');\n await cache.delete(url);\n }\n}"],"mappings":";;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,iBAAuB;AAShB,IAAM,gBAAN,MAAoB;AAAA,EAKzB,YAAY,SAA8B,CAAC,GAAG;AAJ9C,SAAQ,SAAwB;AAChC,SAAQ,YAAoB;AAI1B,SAAK,SAAS;AAAA,MACZ,UAAU,OAAO,YAAY;AAAA,MAC7B,YAAY,OAAO,cAAc;AAAA,MACjC,MAAM,OAAO,QAAQ;AAAA,IACvB;AAAA,EACF;AAAA;AAAA,EAGA,mBAAqC;AACnC,QAAI,OAAO,gBAAgB,aAAa;AACtC,aAAO,EAAE,SAAS,OAAO,OAAO,gDAAgD;AAAA,IAClF;AACA,QAAI,EAAE,YAAY,SAAS;AACzB,aAAO,EAAE,SAAS,OAAO,OAAO,8CAA8C;AAAA,IAChF;AACA,UAAM,YAAY,CAAC,CAAE,UAAkB;AACvC,UAAM,WAAW,OAAO;AACxB,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,MACA,qBAAqB;AAAA,IACvB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,UACJ,MACA,YAC0B;AAC1B,UAAM,MAAM,KAAK,iBAAiB;AAClC,QAAI,CAAC,IAAI,QAAS,QAAO,EAAE,SAAS,OAAO,OAAO,IAAI,MAAM;AAE5D,UAAM,KAAK,OAAO;AAElB,QAAI;AACF,mBAAa,EAAE;AAMf,YAAM,cAAc,EAAE,SAAS,KAAK,OAAO,SAAS;AACpD,WAAK,SAAS,IAAI,kBAAO,WAAW;AACpC,mBAAa,EAAE;AAEf,YAAM,KAAK,aAAa,IAAI;AAC5B,mBAAa,EAAE;AAEf,YAAM,YAAY,CAAC,CAAE,UAAkB;AACvC,YAAO,KAAK,OAAe,UAAU,CAAC,IAAI,GAAG;AAAA,QAC3C,OAAO,KAAK,OAAO;AAAA,QACnB,GAAI,aAAa,KAAK,OAAO,aAAa,IACtC,EAAE,cAAc,KAAK,OAAO,WAAW,IACvC,CAAC;AAAA,MACP,CAAC;AAED,WAAK,YAAY,KAAK;AACtB,mBAAa,GAAG;AAEhB,aAAO,EAAE,SAAS,MAAM,YAAY,UAAU;AAAA,IAChD,SAAS,GAAY;AACnB,YAAM,KAAK,OAAO;AAClB,YAAM,UAAU,aAAa,QAAQ,EAAE,UAAU;AACjD,aAAO,EAAE,SAAS,OAAO,OAAO,QAAQ;AAAA,IAC1C;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,SAAS,KAA+C;AAC5D,QAAI,CAAC,KAAK,QAAQ;AAChB,aAAO,EAAE,SAAS,OAAO,OAAO,2CAA2C;AAAA,IAC7E;AAEA,QAAI;AACF,YAAM,KAAK,YAAY,IAAI;AAE3B,YAAM,WAAgD,CAAC;AACvD,UAAI,IAAI,OAAQ,UAAS,KAAK,EAAE,MAAM,UAAU,SAAS,IAAI,OAAO,CAAC;AACrE,eAAS,KAAK,EAAE,MAAM,QAAQ,SAAS,IAAI,OAAO,CAAC;AAEnD,YAAM,WAAW,MAAO,KAAK,OAAe,qBAAqB;AAAA,QAC/D;AAAA,QACA,YAAY,IAAI,aAAa;AAAA,QAC7B,aAAa,IAAI,eAAe;AAAA,QAChC,OAAO,IAAI,QAAQ;AAAA,QACnB,OAAO,IAAI,QAAQ;AAAA,MACrB,CAAC;AAED,YAAM,OAAe,UAAU,UAAU,CAAC,GAAG,SAAS,WAAW;AACjE,aAAO;AAAA,QACL,SAAS;AAAA,QACT;AAAA,QACA,QAAQ,KAAK,MAAM,YAAY,IAAI,IAAI,EAAE;AAAA,MAC3C;AAAA,IACF,SAAS,GAAY;AACnB,YAAM,UAAU,aAAa,QAAQ,EAAE,UAAU;AACjD,aAAO,EAAE,SAAS,OAAO,OAAO,QAAQ;AAAA,IAC1C;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,SAAwB;AAC5B,QAAI,KAAK,QAAQ;AACf,UAAI;AACF,cAAO,KAAK,OAAe,KAAK;AAAA,MAClC,SAAS,GAAG;AAAA,MAEZ;AACA,WAAK,SAAS;AAAA,IAChB;AACA,QAAI,KAAK,WAAW;AAClB,UAAI;AACF,cAAM,KAAK,gBAAgB,KAAK,SAAS;AAAA,MAC3C,SAAS,GAAG;AAAA,MAEZ;AACA,WAAK,YAAY;AAAA,IACnB;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,WAAoB;AACtB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA;AAAA,EAGA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAc,aAAa,MAA2B;AACpD,UAAM,MAAM,GAAG,OAAO,SAAS,MAAM,wBAAwB,KAAK,IAAI;AACtE,UAAM,QAAQ,MAAM,OAAO,KAAK,qBAAqB;AACrD,UAAM,MAAM;AAAA,MACV;AAAA,MACA,IAAI,SAAS,MAAM;AAAA,QACjB,SAAS,EAAE,gBAAgB,2BAA2B;AAAA,MACxD,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,MAAc,gBAAgB,UAAiC;AAC7D,UAAM,MAAM,GAAG,OAAO,SAAS,MAAM,wBAAwB,QAAQ;AACrE,UAAM,QAAQ,MAAM,OAAO,KAAK,qBAAqB;AACrD,UAAM,MAAM,OAAO,GAAG;AAAA,EACxB;AACF;","names":[]}
|
package/dist/index.mjs
CHANGED
|
@@ -48,7 +48,8 @@ var WllamaService = class {
|
|
|
48
48
|
return { success: true, usedWebGPU: hasWebGPU };
|
|
49
49
|
} catch (e) {
|
|
50
50
|
await this.unload();
|
|
51
|
-
|
|
51
|
+
const message = e instanceof Error ? e.message : "Failed to load model";
|
|
52
|
+
return { success: false, error: message };
|
|
52
53
|
}
|
|
53
54
|
}
|
|
54
55
|
/** Generate a response from the loaded model */
|
|
@@ -75,7 +76,8 @@ var WllamaService = class {
|
|
|
75
76
|
timeMs: Math.round(performance.now() - t0)
|
|
76
77
|
};
|
|
77
78
|
} catch (e) {
|
|
78
|
-
|
|
79
|
+
const message = e instanceof Error ? e.message : "Generation failed";
|
|
80
|
+
return { success: false, error: message };
|
|
79
81
|
}
|
|
80
82
|
}
|
|
81
83
|
/** Unload the current model and free memory */
|
|
@@ -106,9 +108,12 @@ var WllamaService = class {
|
|
|
106
108
|
async storeInCache(file) {
|
|
107
109
|
const url = `${window.location.origin}/wllama-local-models/${file.name}`;
|
|
108
110
|
const cache = await caches.open("wllama-local-models");
|
|
109
|
-
await cache.put(
|
|
110
|
-
|
|
111
|
-
|
|
111
|
+
await cache.put(
|
|
112
|
+
url,
|
|
113
|
+
new Response(file, {
|
|
114
|
+
headers: { "Content-Type": "application/octet-stream" }
|
|
115
|
+
})
|
|
116
|
+
);
|
|
112
117
|
}
|
|
113
118
|
async deleteFromCache(fileName) {
|
|
114
119
|
const url = `${window.location.origin}/wllama-local-models/${fileName}`;
|
package/dist/index.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/WllamaService.ts"],"sourcesContent":["import { Wllama } from '@wllama/wllama/esm/index.js';\nimport type {\n WllamaServiceConfig,\n GenerateRequest,\n GenerateResult,\n LoadModelResult,\n ConnectionResult,\n} from './types';\n\nexport class WllamaService {\n private wllama: Wllama | null = null;\n private modelName: string = '';\n private config: Required<WllamaServiceConfig>;\n\n constructor(config: WllamaServiceConfig = {}) {\n this.config = {\n wasmPath: config.wasmPath ?? '/wllama/wllama.wasm',\n nGpuLayers: config.nGpuLayers ?? 999,\n nCtx: config.nCtx ?? 2048,\n };\n }\n\n /** Check browser compatibility and capabilities */\n checkEnvironment(): ConnectionResult {\n if (typeof WebAssembly === 'undefined') {\n return { success: false, error: 'WebAssembly is not supported in this browser.' };\n }\n if (!('caches' in window)) {\n return { success: false, error: 'Cache API is not supported in this browser.' };\n }\n const hasWebGPU = !!(navigator as any).gpu;\n const isolated = window.crossOriginIsolated;\n return {\n success: true,\n hasWebGPU,\n crossOriginIsolated: isolated,\n };\n }\n\n /** Load a GGUF model file from the user's filesystem */\n async loadModel(\n file: File,\n onProgress?: (progress: number) => void,\n ): Promise<LoadModelResult> {\n const env = this.checkEnvironment();\n if (!env.success) return { success: false, error: env.error };\n\n await this.unload();\n\n try {\n onProgress?.(10);\n\n const configPaths = { default: this.config.wasmPath };\n this.wllama = new Wllama(configPaths
|
|
1
|
+
{"version":3,"sources":["../src/WllamaService.ts"],"sourcesContent":["import { Wllama } from '@wllama/wllama/esm/index.js';\nimport type {\n WllamaServiceConfig,\n GenerateRequest,\n GenerateResult,\n LoadModelResult,\n ConnectionResult,\n} from './types';\n\nexport class WllamaService {\n private wllama: Wllama | null = null;\n private modelName: string = '';\n private config: Required<WllamaServiceConfig>;\n\n constructor(config: WllamaServiceConfig = {}) {\n this.config = {\n wasmPath: config.wasmPath ?? '/wllama/wllama.wasm',\n nGpuLayers: config.nGpuLayers ?? 999,\n nCtx: config.nCtx ?? 2048,\n };\n }\n\n /** Check browser compatibility and capabilities */\n checkEnvironment(): ConnectionResult {\n if (typeof WebAssembly === 'undefined') {\n return { success: false, error: 'WebAssembly is not supported in this browser.' };\n }\n if (!('caches' in window)) {\n return { success: false, error: 'Cache API is not supported in this browser.' };\n }\n const hasWebGPU = !!(navigator as any).gpu;\n const isolated = window.crossOriginIsolated;\n return {\n success: true,\n hasWebGPU,\n crossOriginIsolated: isolated,\n };\n }\n\n /** Load a GGUF model file from the user's filesystem */\n async loadModel(\n file: File,\n onProgress?: (progress: number) => void,\n ): Promise<LoadModelResult> {\n const env = this.checkEnvironment();\n if (!env.success) return { success: false, error: env.error };\n\n await this.unload();\n\n try {\n onProgress?.(10);\n\n // AssetsPathConfig requires `default` — this is the key the\n // Wllama runtime actually reads (confirmed against the real\n // @wllama/wllama@3.5.1 implementation, not just its docs).\n // No `as any` needed: this shape is type-correct as written.\n const configPaths = { default: this.config.wasmPath };\n this.wllama = new Wllama(configPaths);\n onProgress?.(30);\n\n await this.storeInCache(file);\n onProgress?.(50);\n\n const hasWebGPU = !!(navigator as any).gpu;\n await (this.wllama as any).loadModel([file], {\n n_ctx: this.config.nCtx,\n ...(hasWebGPU && this.config.nGpuLayers > 0\n ? { n_gpu_layers: this.config.nGpuLayers }\n : {}),\n });\n\n this.modelName = file.name;\n onProgress?.(100);\n\n return { success: true, usedWebGPU: hasWebGPU };\n } catch (e: unknown) {\n await this.unload();\n const message = e instanceof Error ? e.message : 'Failed to load model';\n return { success: false, error: message };\n }\n }\n\n /** Generate a response from the loaded model */\n async generate(req: GenerateRequest): Promise<GenerateResult> {\n if (!this.wllama) {\n return { success: false, error: 'No model loaded. Call loadModel() first.' };\n }\n\n try {\n const t0 = performance.now();\n\n const messages: { role: string; content: string }[] = [];\n if (req.system) messages.push({ role: 'system', content: req.system });\n messages.push({ role: 'user', content: req.prompt });\n\n const response = await (this.wllama as any).createChatCompletion({\n messages,\n max_tokens: req.maxTokens ?? 512,\n temperature: req.temperature ?? 0.7,\n top_k: req.topK ?? 40,\n top_p: req.topP ?? 0.95,\n });\n\n const text: string = response?.choices?.[0]?.message?.content ?? '';\n return {\n success: true,\n text,\n timeMs: Math.round(performance.now() - t0),\n };\n } catch (e: unknown) {\n const message = e instanceof Error ? e.message : 'Generation failed';\n return { success: false, error: message };\n }\n }\n\n /** Unload the current model and free memory */\n async unload(): Promise<void> {\n if (this.wllama) {\n try {\n await (this.wllama as any).exit();\n } catch (_) {\n /* ignore */\n }\n this.wllama = null;\n }\n if (this.modelName) {\n try {\n await this.deleteFromCache(this.modelName);\n } catch (_) {\n /* ignore */\n }\n this.modelName = '';\n }\n }\n\n /** Whether a model is currently loaded */\n get isLoaded(): boolean {\n return this.wllama !== null;\n }\n\n /** Name of the currently loaded model */\n get currentModel(): string {\n return this.modelName;\n }\n\n private async storeInCache(file: File): Promise<void> {\n const url = `${window.location.origin}/wllama-local-models/${file.name}`;\n const cache = await caches.open('wllama-local-models');\n await cache.put(\n url,\n new Response(file, {\n headers: { 'Content-Type': 'application/octet-stream' },\n }),\n );\n }\n\n private async deleteFromCache(fileName: string): Promise<void> {\n const url = `${window.location.origin}/wllama-local-models/${fileName}`;\n const cache = await caches.open('wllama-local-models');\n await cache.delete(url);\n }\n}"],"mappings":";AAAA,SAAS,cAAc;AAShB,IAAM,gBAAN,MAAoB;AAAA,EAKzB,YAAY,SAA8B,CAAC,GAAG;AAJ9C,SAAQ,SAAwB;AAChC,SAAQ,YAAoB;AAI1B,SAAK,SAAS;AAAA,MACZ,UAAU,OAAO,YAAY;AAAA,MAC7B,YAAY,OAAO,cAAc;AAAA,MACjC,MAAM,OAAO,QAAQ;AAAA,IACvB;AAAA,EACF;AAAA;AAAA,EAGA,mBAAqC;AACnC,QAAI,OAAO,gBAAgB,aAAa;AACtC,aAAO,EAAE,SAAS,OAAO,OAAO,gDAAgD;AAAA,IAClF;AACA,QAAI,EAAE,YAAY,SAAS;AACzB,aAAO,EAAE,SAAS,OAAO,OAAO,8CAA8C;AAAA,IAChF;AACA,UAAM,YAAY,CAAC,CAAE,UAAkB;AACvC,UAAM,WAAW,OAAO;AACxB,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,MACA,qBAAqB;AAAA,IACvB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,UACJ,MACA,YAC0B;AAC1B,UAAM,MAAM,KAAK,iBAAiB;AAClC,QAAI,CAAC,IAAI,QAAS,QAAO,EAAE,SAAS,OAAO,OAAO,IAAI,MAAM;AAE5D,UAAM,KAAK,OAAO;AAElB,QAAI;AACF,mBAAa,EAAE;AAMf,YAAM,cAAc,EAAE,SAAS,KAAK,OAAO,SAAS;AACpD,WAAK,SAAS,IAAI,OAAO,WAAW;AACpC,mBAAa,EAAE;AAEf,YAAM,KAAK,aAAa,IAAI;AAC5B,mBAAa,EAAE;AAEf,YAAM,YAAY,CAAC,CAAE,UAAkB;AACvC,YAAO,KAAK,OAAe,UAAU,CAAC,IAAI,GAAG;AAAA,QAC3C,OAAO,KAAK,OAAO;AAAA,QACnB,GAAI,aAAa,KAAK,OAAO,aAAa,IACtC,EAAE,cAAc,KAAK,OAAO,WAAW,IACvC,CAAC;AAAA,MACP,CAAC;AAED,WAAK,YAAY,KAAK;AACtB,mBAAa,GAAG;AAEhB,aAAO,EAAE,SAAS,MAAM,YAAY,UAAU;AAAA,IAChD,SAAS,GAAY;AACnB,YAAM,KAAK,OAAO;AAClB,YAAM,UAAU,aAAa,QAAQ,EAAE,UAAU;AACjD,aAAO,EAAE,SAAS,OAAO,OAAO,QAAQ;AAAA,IAC1C;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,SAAS,KAA+C;AAC5D,QAAI,CAAC,KAAK,QAAQ;AAChB,aAAO,EAAE,SAAS,OAAO,OAAO,2CAA2C;AAAA,IAC7E;AAEA,QAAI;AACF,YAAM,KAAK,YAAY,IAAI;AAE3B,YAAM,WAAgD,CAAC;AACvD,UAAI,IAAI,OAAQ,UAAS,KAAK,EAAE,MAAM,UAAU,SAAS,IAAI,OAAO,CAAC;AACrE,eAAS,KAAK,EAAE,MAAM,QAAQ,SAAS,IAAI,OAAO,CAAC;AAEnD,YAAM,WAAW,MAAO,KAAK,OAAe,qBAAqB;AAAA,QAC/D;AAAA,QACA,YAAY,IAAI,aAAa;AAAA,QAC7B,aAAa,IAAI,eAAe;AAAA,QAChC,OAAO,IAAI,QAAQ;AAAA,QACnB,OAAO,IAAI,QAAQ;AAAA,MACrB,CAAC;AAED,YAAM,OAAe,UAAU,UAAU,CAAC,GAAG,SAAS,WAAW;AACjE,aAAO;AAAA,QACL,SAAS;AAAA,QACT;AAAA,QACA,QAAQ,KAAK,MAAM,YAAY,IAAI,IAAI,EAAE;AAAA,MAC3C;AAAA,IACF,SAAS,GAAY;AACnB,YAAM,UAAU,aAAa,QAAQ,EAAE,UAAU;AACjD,aAAO,EAAE,SAAS,OAAO,OAAO,QAAQ;AAAA,IAC1C;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,SAAwB;AAC5B,QAAI,KAAK,QAAQ;AACf,UAAI;AACF,cAAO,KAAK,OAAe,KAAK;AAAA,MAClC,SAAS,GAAG;AAAA,MAEZ;AACA,WAAK,SAAS;AAAA,IAChB;AACA,QAAI,KAAK,WAAW;AAClB,UAAI;AACF,cAAM,KAAK,gBAAgB,KAAK,SAAS;AAAA,MAC3C,SAAS,GAAG;AAAA,MAEZ;AACA,WAAK,YAAY;AAAA,IACnB;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,WAAoB;AACtB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA;AAAA,EAGA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAc,aAAa,MAA2B;AACpD,UAAM,MAAM,GAAG,OAAO,SAAS,MAAM,wBAAwB,KAAK,IAAI;AACtE,UAAM,QAAQ,MAAM,OAAO,KAAK,qBAAqB;AACrD,UAAM,MAAM;AAAA,MACV;AAAA,MACA,IAAI,SAAS,MAAM;AAAA,QACjB,SAAS,EAAE,gBAAgB,2BAA2B;AAAA,MACxD,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,MAAc,gBAAgB,UAAiC;AAC7D,UAAM,MAAM,GAAG,OAAO,SAAS,MAAM,wBAAwB,QAAQ;AACrE,UAAM,QAAQ,MAAM,OAAO,KAAK,qBAAqB;AACrD,UAAM,MAAM,OAAO,GAAG;AAAA,EACxB;AACF;","names":[]}
|
package/package.json
CHANGED
|
@@ -1,8 +1,11 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "wllama-service",
|
|
3
|
-
"version": "
|
|
3
|
+
"version": "2.0.0",
|
|
4
4
|
"description": "Framework-agnostic browser LLM service wrapper with WebGPU and WebAssembly support",
|
|
5
5
|
"main": "./dist/index.js",
|
|
6
|
+
"bin": {
|
|
7
|
+
"copy-wllama-wasm": "./scripts/postinstall.js"
|
|
8
|
+
},
|
|
6
9
|
"module": "./dist/index.mjs",
|
|
7
10
|
"types": "./dist/index.d.ts",
|
|
8
11
|
"exports": {
|
|
@@ -19,6 +22,7 @@
|
|
|
19
22
|
"scripts": {
|
|
20
23
|
"build": "tsup",
|
|
21
24
|
"dev": "tsup --watch",
|
|
25
|
+
"typecheck": "tsc --noEmit",
|
|
22
26
|
"prepublishOnly": "npm run build"
|
|
23
27
|
},
|
|
24
28
|
"keywords": [
|
|
@@ -31,11 +35,10 @@
|
|
|
31
35
|
"gguf"
|
|
32
36
|
],
|
|
33
37
|
"license": "MIT",
|
|
34
|
-
"
|
|
35
|
-
"@wllama/wllama": "
|
|
38
|
+
"dependencies": {
|
|
39
|
+
"@wllama/wllama": "^3.5.1"
|
|
36
40
|
},
|
|
37
41
|
"devDependencies": {
|
|
38
|
-
"@wllama/wllama": "^3.0.0",
|
|
39
42
|
"tsup": "^8.0.0",
|
|
40
43
|
"typescript": "^5.0.0"
|
|
41
44
|
}
|
|
@@ -0,0 +1,26 @@
|
|
|
1
|
+
#!/usr/bin/env node
|
|
2
|
+
'use strict';
|
|
3
|
+
|
|
4
|
+
const path = require('path');
|
|
5
|
+
const { copyWasm } = require('./copyWllamaWasm');
|
|
6
|
+
|
|
7
|
+
// This file is wired up as the "postinstall" script in the CONSUMING
|
|
8
|
+
// APP's package.json (see setup below) — npm runs it automatically
|
|
9
|
+
// right after `npm install` finishes. No command to type, no step
|
|
10
|
+
// to remember.
|
|
11
|
+
//
|
|
12
|
+
// Destination defaults to <app>/public/wllama, but can be overridden
|
|
13
|
+
// with an env var if the app's public folder lives somewhere else:
|
|
14
|
+
// WLLAMA_WASM_DEST=./static/wllama npm install
|
|
15
|
+
|
|
16
|
+
const destArg = process.env.WLLAMA_WASM_DEST || 'public/wllama';
|
|
17
|
+
const destDir = path.resolve(process.cwd(), destArg);
|
|
18
|
+
|
|
19
|
+
const result = copyWasm(destDir);
|
|
20
|
+
|
|
21
|
+
if (!result) {
|
|
22
|
+
// Non-fatal: don't break `npm install` for the app if something's
|
|
23
|
+
// off. Worst case, loadModel() fails clearly at runtime with a 404,
|
|
24
|
+
// which is easier to debug than a broken install.
|
|
25
|
+
process.exit(0);
|
|
26
|
+
}
|