@gatanot/orrery 0.1.0

package/dist/extensions/index.js

@@ -0,0 +1,3 @@
+import llamaExtension from "./llama/index.js";
+export const builtInExtensions = [{ name: "llama.cpp", factory: llamaExtension, hidden: true }];
+//# sourceMappingURL=index.js.map

package/dist/extensions/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/extensions/index.ts"],"names":[],"mappings":"AACA,OAAO,cAAc,MAAM,kBAAkB,CAAC;AAE9C,MAAM,CAAC,MAAM,iBAAiB,GAAsB,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,OAAO,EAAE,cAAc,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAC","sourcesContent":["import type { InlineExtension } from \"../core/extensions/types.ts\";\nimport llamaExtension from \"./llama/index.ts\";\n\nexport const builtInExtensions: InlineExtension[] = [{ name: \"llama.cpp\", factory: llamaExtension, hidden: true }];\n"]}

package/dist/extensions/llama/client.d.ts

@@ -0,0 +1,67 @@
+export type LlamaModelStatus = "unloaded" | "loading" | "loaded" | "downloading" | "sleeping";
+export interface LlamaModelInfo {
+    id: string;
+    aliases?: string[];
+    status: {
+        value: LlamaModelStatus;
+        args?: string[];
+        failed?: boolean;
+        exit_code?: number;
+        progress?: Record<string, {
+            done: number;
+            total: number;
+        }>;
+    };
+    architecture?: {
+        input_modalities?: string[];
+        output_modalities?: string[];
+    };
+    source?: string;
+    meta?: {
+        n_ctx?: number;
+        n_ctx_train?: number;
+        size?: number;
+        ftype?: string;
+    };
+}
+export interface LlamaModelsResponse {
+    data: LlamaModelInfo[];
+    object?: string;
+}
+export interface LlamaServerProps {
+    models_autoload?: boolean;
+}
+export interface LlamaModelEvent {
+    model: string;
+    event: string;
+    data?: unknown;
+}
+export interface LlamaProgress {
+    message: string;
+    ratio?: number;
+    detail?: string;
+}
+export declare function formatBytes(bytes: number): string;
+export declare function normalizeLlamaServerUrl(value: string): string;
+export declare function llamaInferenceUrl(serverUrl: string): string;
+export declare class LlamaClient {
+    readonly serverUrl: string;
+    private readonly apiKey;
+    constructor(serverUrl: string, apiKey?: string);
+    private request;
+    list(options?: {
+        reload?: boolean;
+        signal?: AbortSignal;
+    }): Promise<LlamaModelInfo[]>;
+    props(options?: {
+        signal?: AbortSignal;
+    }): Promise<LlamaServerProps>;
+    load(model: string, signal?: AbortSignal): Promise<void>;
+    unload(model: string, signal?: AbortSignal): Promise<void>;
+    unloadAndWait(model: string, signal?: AbortSignal): Promise<void>;
+    download(model: string, signal?: AbortSignal): Promise<void>;
+    watch(onEvent: (event: LlamaModelEvent) => void, signal?: AbortSignal): Promise<void>;
+    loadAndWait(model: string, onProgress: (progress: LlamaProgress) => void, signal?: AbortSignal): Promise<LlamaModelInfo>;
+    downloadAndWait(model: string, onProgress: (progress: LlamaProgress) => void, signal?: AbortSignal): Promise<LlamaModelInfo[]>;
+}
+//# sourceMappingURL=client.d.ts.map

package/dist/extensions/llama/client.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.d.ts","sourceRoot":"","sources":["../../../src/extensions/llama/client.ts"],"names":[],"mappings":"AAAA,MAAM,MAAM,gBAAgB,GAAG,UAAU,GAAG,SAAS,GAAG,QAAQ,GAAG,aAAa,GAAG,UAAU,CAAC;AAE9F,MAAM,WAAW,cAAc;IAC9B,EAAE,EAAE,MAAM,CAAC;IACX,OAAO,CAAC,EAAE,MAAM,EAAE,CAAC;IACnB,MAAM,EAAE;QACP,KAAK,EAAE,gBAAgB,CAAC;QACxB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC;QAChB,MAAM,CAAC,EAAE,OAAO,CAAC;QACjB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE;YAAE,IAAI,EAAE,MAAM,CAAC;YAAC,KAAK,EAAE,MAAM,CAAA;SAAE,CAAC,CAAC;KAC3D,CAAC;IACF,YAAY,CAAC,EAAE;QACd,gBAAgB,CAAC,EAAE,MAAM,EAAE,CAAC;QAC5B,iBAAiB,CAAC,EAAE,MAAM,EAAE,CAAC;KAC7B,CAAC;IACF,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE;QACN,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,WAAW,CAAC,EAAE,MAAM,CAAC;QACrB,IAAI,CAAC,EAAE,MAAM,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;KACf,CAAC;CACF;AAED,MAAM,WAAW,mBAAmB;IACnC,IAAI,EAAE,cAAc,EAAE,CAAC;IACvB,MAAM,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,gBAAgB;IAChC,eAAe,CAAC,EAAE,OAAO,CAAC;CAC1B;AAED,MAAM,WAAW,eAAe;IAC/B,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,OAAO,CAAC;CACf;AAED,MAAM,WAAW,aAAa;IAC7B,OAAO,EAAE,MAAM,CAAC;IAChB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,MAAM,CAAC;CAChB;AAwFD,wBAAgB,WAAW,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAUjD;AAED,wBAAgB,uBAAuB,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAS7D;AAED,wBAAgB,iBAAiB,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAE3D;AAED,qBAAa,WAAW;IACvB,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAE5C,YAAY,SAAS,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,EAG7C;YAEa,OAAO;IAiBf,IAAI,CAAC,OAAO,GAAE;QAAE,MAAM,CAAC,EAAE,OAAO,CAAC;QAAC,MAAM,CAAC,EAAE,WAAW,CAAA;KAAO,GAAG,OAAO,CAAC,cAAc,EAAE,CAAC,CAQ9F;IAEK,KAAK,CAAC,OAAO,GAAE;QAAE,MAAM,CAAC,EAAE,WAAW,CAAA;KAAO,GAAG,OAAO,CAAC,gBAAgB,CAAC,CAK7E;IAEK,IAAI,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAE7D;IAEK,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAE/D;IAEK,aAAa,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAOtE;IAEK,QAAQ,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAEjE;IAEK,KAAK,CAAC,OAAO,EAAE,CAAC,KAAK,EAAE,eAAe,KAAK,IAAI,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC,CAgC1F;IAEK,WAAW,CAChB,KAAK,EAAE,MAAM,EACb,UAAU,EAAE,CAAC,QAAQ,EAAE,aAAa,KAAK,IAAI,EAC7C,MAAM,CAAC,EAAE,WAAW,GAClB,OAAO,CAAC,cAAc,CAAC,CAmCzB;IAEK,eAAe,CACpB,KAAK,EAAE,MAAM,EACb,UAAU,EAAE,CAAC,QAAQ,EAAE,aAAa,KAAK,IAAI,EAC7C,MAAM,CAAC,EAAE,WAAW,GAClB,OAAO,CAAC,cAAc,EAAE,CAAC,CAuC3B;CACD","sourcesContent":["export type LlamaModelStatus = \"unloaded\" | \"loading\" | \"loaded\" | \"downloading\" | \"sleeping\";\n\nexport interface LlamaModelInfo {\n\tid: string;\n\taliases?: string[];\n\tstatus: {\n\t\tvalue: LlamaModelStatus;\n\t\targs?: string[];\n\t\tfailed?: boolean;\n\t\texit_code?: number;\n\t\tprogress?: Record<string, { done: number; total: number }>;\n\t};\n\tarchitecture?: {\n\t\tinput_modalities?: string[];\n\t\toutput_modalities?: string[];\n\t};\n\tsource?: string;\n\tmeta?: {\n\t\tn_ctx?: number;\n\t\tn_ctx_train?: number;\n\t\tsize?: number;\n\t\tftype?: string;\n\t};\n}\n\nexport interface LlamaModelsResponse {\n\tdata: LlamaModelInfo[];\n\tobject?: string;\n}\n\nexport interface LlamaServerProps {\n\tmodels_autoload?: boolean;\n}\n\nexport interface LlamaModelEvent {\n\tmodel: string;\n\tevent: string;\n\tdata?: unknown;\n}\n\nexport interface LlamaProgress {\n\tmessage: string;\n\tratio?: number;\n\tdetail?: string;\n}\n\nfunction errorMessage(payload: unknown, fallback: string): string {\n\tif (typeof payload !== \"object\" || payload === null) return fallback;\n\tconst error = (payload as { error?: unknown }).error;\n\tif (typeof error !== \"object\" || error === null) return fallback;\n\tconst message = (error as { message?: unknown }).message;\n\treturn typeof message === \"string\" && message ? message : fallback;\n}\n\nfunction isModelInfo(value: unknown): value is LlamaModelInfo {\n\tif (typeof value !== \"object\" || value === null) return false;\n\tconst candidate = value as { id?: unknown; status?: { value?: unknown } };\n\treturn typeof candidate.id === \"string\" && typeof candidate.status?.value === \"string\";\n}\n\nfunction linkSignal(source: AbortSignal | undefined, target: AbortController): () => void {\n\tif (!source) return () => {};\n\tif (source.aborted) {\n\t\ttarget.abort(source.reason);\n\t\treturn () => {};\n\t}\n\tconst abort = () => target.abort(source.reason);\n\tsource.addEventListener(\"abort\", abort, { once: true });\n\treturn () => source.removeEventListener(\"abort\", abort);\n}\n\nfunction sleep(ms: number, signal?: AbortSignal): Promise<void> {\n\treturn new Promise((resolve, reject) => {\n\t\tif (signal?.aborted) {\n\t\t\treject(signal.reason ?? new Error(\"Cancelled\"));\n\t\t\treturn;\n\t\t}\n\t\tconst abort = () => {\n\t\t\tclearTimeout(timeout);\n\t\t\treject(signal?.reason ?? new Error(\"Cancelled\"));\n\t\t};\n\t\tconst timeout = setTimeout(() => {\n\t\t\tsignal?.removeEventListener(\"abort\", abort);\n\t\t\tresolve();\n\t\t}, ms);\n\t\tsignal?.addEventListener(\"abort\", abort, { once: true });\n\t});\n}\n\nfunction parseLoadProgress(data: unknown): LlamaProgress | undefined {\n\tif (typeof data !== \"object\" || data === null) return undefined;\n\tconst progress = (data as { progress?: unknown }).progress;\n\tif (typeof progress !== \"object\" || progress === null) return undefined;\n\tconst value = progress as { stages?: unknown; current?: unknown; stage?: unknown; value?: unknown };\n\tconst stage =\n\t\ttypeof value.current === \"string\" ? value.current : typeof value.stage === \"string\" ? value.stage : undefined;\n\tconst stages = Array.isArray(value.stages)\n\t\t? value.stages.filter((entry): entry is string => typeof entry === \"string\")\n\t\t: [];\n\tconst stageRatio = typeof value.value === \"number\" ? Math.max(0, Math.min(1, value.value)) : undefined;\n\tlet ratio = stageRatio;\n\tif (stage && stages.length > 0) {\n\t\tconst index = stages.indexOf(stage);\n\t\tif (index >= 0) ratio = (index + (stageRatio ?? 0)) / stages.length;\n\t}\n\treturn {\n\t\tmessage: stage ? `Loading ${stage.replaceAll(\"_\", \" \")}` : \"Loading model\",\n\t\tratio,\n\t};\n}\n\nfunction parseDownloadProgress(data: unknown): LlamaProgress | undefined {\n\tif (typeof data !== \"object\" || data === null) return undefined;\n\tconst nested = (data as { progress?: unknown }).progress;\n\tconst files = typeof nested === \"object\" && nested !== null ? nested : data;\n\tlet done = 0;\n\tlet total = 0;\n\tfor (const value of Object.values(files as Record<string, unknown>)) {\n\t\tif (typeof value !== \"object\" || value === null) continue;\n\t\tconst entry = value as { done?: unknown; total?: unknown };\n\t\tif (typeof entry.done !== \"number\" || typeof entry.total !== \"number\") continue;\n\t\tdone += entry.done;\n\t\ttotal += entry.total;\n\t}\n\tif (total <= 0) return undefined;\n\treturn {\n\t\tmessage: \"Downloading model\",\n\t\tratio: done / total,\n\t\tdetail: `${formatBytes(done)} / ${formatBytes(total)}`,\n\t};\n}\n\nexport function formatBytes(bytes: number): string {\n\tif (bytes < 1024) return `${bytes} B`;\n\tconst units = [\"KiB\", \"MiB\", \"GiB\", \"TiB\"];\n\tlet value = bytes / 1024;\n\tlet unit = units[0]!;\n\tfor (let index = 1; index < units.length && value >= 1024; index++) {\n\t\tvalue /= 1024;\n\t\tunit = units[index]!;\n\t}\n\treturn `${value >= 10 ? value.toFixed(1) : value.toFixed(2)} ${unit}`;\n}\n\nexport function normalizeLlamaServerUrl(value: string): string {\n\tconst url = new URL(value.trim());\n\tif (url.protocol !== \"http:\" && url.protocol !== \"https:\") {\n\t\tthrow new Error(\"Server URL must use http or https\");\n\t}\n\turl.hash = \"\";\n\turl.search = \"\";\n\turl.pathname = url.pathname.replace(/\\/+$/u, \"\").replace(/\\/v1$/u, \"\") || \"/\";\n\treturn url.toString().replace(/\\/$/u, \"\");\n}\n\nexport function llamaInferenceUrl(serverUrl: string): string {\n\treturn `${normalizeLlamaServerUrl(serverUrl)}/v1`;\n}\n\nexport class LlamaClient {\n\treadonly serverUrl: string;\n\tprivate readonly apiKey: string | undefined;\n\n\tconstructor(serverUrl: string, apiKey?: string) {\n\t\tthis.serverUrl = normalizeLlamaServerUrl(serverUrl);\n\t\tthis.apiKey = apiKey;\n\t}\n\n\tprivate async request(path: string, init: RequestInit = {}): Promise<unknown> {\n\t\tconst headers = new Headers(init.headers);\n\t\tif (init.body !== undefined) headers.set(\"Content-Type\", \"application/json\");\n\t\tif (this.apiKey) headers.set(\"Authorization\", `Bearer ${this.apiKey}`);\n\t\tconst timeout = AbortSignal.timeout(15_000);\n\t\tconst signal = init.signal ? AbortSignal.any([init.signal, timeout]) : timeout;\n\t\tconst response = await fetch(`${this.serverUrl}${path}`, { ...init, headers, signal });\n\t\tlet payload: unknown;\n\t\ttry {\n\t\t\tpayload = await response.json();\n\t\t} catch {\n\t\t\tpayload = undefined;\n\t\t}\n\t\tif (!response.ok) throw new Error(errorMessage(payload, `llama.cpp returned HTTP ${response.status}`));\n\t\treturn payload;\n\t}\n\n\tasync list(options: { reload?: boolean; signal?: AbortSignal } = {}): Promise<LlamaModelInfo[]> {\n\t\tconst payload = await this.request(`/models${options.reload ? \"?reload=1\" : \"\"}`, { signal: options.signal });\n\t\tif (typeof payload !== \"object\" || payload === null || !Array.isArray((payload as { data?: unknown }).data)) {\n\t\t\tthrow new Error(\"llama.cpp returned an invalid model catalog\");\n\t\t}\n\t\tconst data = (payload as { data: unknown[] }).data;\n\t\tif (!data.every(isModelInfo)) throw new Error(\"Server is not running in llama.cpp router mode\");\n\t\treturn data;\n\t}\n\n\tasync props(options: { signal?: AbortSignal } = {}): Promise<LlamaServerProps> {\n\t\tconst payload = await this.request(\"/props\", { signal: options.signal });\n\t\tif (typeof payload !== \"object\" || payload === null) return {};\n\t\tconst { models_autoload: modelsAutoload } = payload as Record<string, unknown>;\n\t\treturn typeof modelsAutoload === \"boolean\" ? { models_autoload: modelsAutoload } : {};\n\t}\n\n\tasync load(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.request(\"/models/load\", { method: \"POST\", body: JSON.stringify({ model }), signal });\n\t}\n\n\tasync unload(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.request(\"/models/unload\", { method: \"POST\", body: JSON.stringify({ model }), signal });\n\t}\n\n\tasync unloadAndWait(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.unload(model, signal);\n\t\twhile (true) {\n\t\t\tconst entry = (await this.list({ signal })).find((candidate) => candidate.id === model);\n\t\t\tif (!entry || entry.status.value === \"unloaded\") return;\n\t\t\tawait sleep(100, signal);\n\t\t}\n\t}\n\n\tasync download(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.request(\"/models\", { method: \"POST\", body: JSON.stringify({ model }), signal });\n\t}\n\n\tasync watch(onEvent: (event: LlamaModelEvent) => void, signal?: AbortSignal): Promise<void> {\n\t\tconst headers = new Headers();\n\t\tif (this.apiKey) headers.set(\"Authorization\", `Bearer ${this.apiKey}`);\n\t\tconst response = await fetch(`${this.serverUrl}/models/sse`, { headers, signal });\n\t\tif (!response.ok || !response.body) throw new Error(`llama.cpp SSE returned HTTP ${response.status}`);\n\t\tconst reader = response.body.getReader();\n\t\tconst decoder = new TextDecoder();\n\t\tlet buffer = \"\";\n\t\twhile (true) {\n\t\t\tconst chunk = await reader.read();\n\t\t\tif (chunk.done) break;\n\t\t\tbuffer += decoder.decode(chunk.value, { stream: true }).replaceAll(\"\\r\\n\", \"\\n\");\n\t\t\tlet boundary = buffer.indexOf(\"\\n\\n\");\n\t\t\twhile (boundary >= 0) {\n\t\t\t\tconst frame = buffer.slice(0, boundary);\n\t\t\t\tbuffer = buffer.slice(boundary + 2);\n\t\t\t\tconst data = frame\n\t\t\t\t\t.split(\"\\n\")\n\t\t\t\t\t.filter((line) => line.startsWith(\"data:\"))\n\t\t\t\t\t.map((line) => line.slice(5).trimStart())\n\t\t\t\t\t.join(\"\\n\");\n\t\t\t\tif (data) {\n\t\t\t\t\ttry {\n\t\t\t\t\t\tconst event = JSON.parse(data) as LlamaModelEvent;\n\t\t\t\t\t\tif (event && typeof event.model === \"string\" && typeof event.event === \"string\") onEvent(event);\n\t\t\t\t\t} catch {\n\t\t\t\t\t\t// Ignore malformed events; catalog polling remains authoritative.\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\tboundary = buffer.indexOf(\"\\n\\n\");\n\t\t\t}\n\t\t}\n\t}\n\n\tasync loadAndWait(\n\t\tmodel: string,\n\t\tonProgress: (progress: LlamaProgress) => void,\n\t\tsignal?: AbortSignal,\n\t): Promise<LlamaModelInfo> {\n\t\tconst watcher = new AbortController();\n\t\tconst unlink = linkSignal(signal, watcher);\n\t\tlet eventLoaded = false;\n\t\tlet eventError: string | undefined;\n\t\tvoid this.watch((event) => {\n\t\t\tif (event.model !== model) return;\n\t\t\tif (event.event !== \"model_status\" && event.event !== \"status_change\") return;\n\t\t\tconst data = event.data as { status?: unknown } | undefined;\n\t\t\tif (data?.status === \"loaded\") eventLoaded = true;\n\t\t\tif (data?.status === \"unloaded\") eventError = \"Model failed to load\";\n\t\t\tconst progress = parseLoadProgress(event.data);\n\t\t\tif (progress) onProgress(progress);\n\t\t}, watcher.signal).catch(() => {});\n\t\ttry {\n\t\t\tawait this.load(model, signal);\n\t\t\tonProgress({ message: \"Loading model\" });\n\t\t\twhile (true) {\n\t\t\t\tif (signal?.aborted) throw signal.reason ?? new Error(\"Cancelled\");\n\t\t\t\tconst entry = (await this.list({ signal })).find((candidate) => candidate.id === model);\n\t\t\t\tif (entry?.status.value === \"loaded\") return entry;\n\t\t\t\tif (eventLoaded && !entry) return { id: model, status: { value: \"loaded\" } };\n\t\t\t\tif (entry?.status.failed || eventError) {\n\t\t\t\t\tthrow new Error(\n\t\t\t\t\t\tentry?.status.exit_code === undefined\n\t\t\t\t\t\t\t? (eventError ?? \"Model failed to load\")\n\t\t\t\t\t\t\t: `Model exited with code ${entry.status.exit_code}`,\n\t\t\t\t\t);\n\t\t\t\t}\n\t\t\t\tawait sleep(250, signal);\n\t\t\t}\n\t\t} finally {\n\t\t\tunlink();\n\t\t\twatcher.abort();\n\t\t}\n\t}\n\n\tasync downloadAndWait(\n\t\tmodel: string,\n\t\tonProgress: (progress: LlamaProgress) => void,\n\t\tsignal?: AbortSignal,\n\t): Promise<LlamaModelInfo[]> {\n\t\tconst watcher = new AbortController();\n\t\tconst unlink = linkSignal(signal, watcher);\n\t\tlet finished = false;\n\t\tlet failure: string | undefined;\n\t\tlet sawDownloading = false;\n\t\tlet polls = 0;\n\t\tvoid this.watch((event) => {\n\t\t\tif (event.model !== model) return;\n\t\t\tif (event.event === \"download_finished\") finished = true;\n\t\t\tif (event.event === \"download_failed\") failure = errorMessage(event.data, \"Download failed\");\n\t\t\tif (event.event === \"download_progress\") {\n\t\t\t\tsawDownloading = true;\n\t\t\t\tconst progress = parseDownloadProgress(event.data);\n\t\t\t\tif (progress) onProgress(progress);\n\t\t\t}\n\t\t}, watcher.signal).catch(() => {});\n\t\ttry {\n\t\t\tawait this.download(model, signal);\n\t\t\tonProgress({ message: \"Downloading model\" });\n\t\t\twhile (true) {\n\t\t\t\tif (signal?.aborted) throw signal.reason ?? new Error(\"Cancelled\");\n\t\t\t\tif (failure) throw new Error(failure);\n\t\t\t\tconst models = await this.list({ signal });\n\t\t\t\tpolls++;\n\t\t\t\tconst entry = models.find((candidate) => candidate.id === model);\n\t\t\t\tif (entry?.status.value === \"downloading\") {\n\t\t\t\t\tsawDownloading = true;\n\t\t\t\t\tconst progress = parseDownloadProgress(entry.status.progress);\n\t\t\t\t\tif (progress) onProgress(progress);\n\t\t\t\t} else if (finished || (entry && (sawDownloading || polls >= 2))) {\n\t\t\t\t\treturn this.list({ reload: true, signal });\n\t\t\t\t}\n\t\t\t\tawait sleep(500, signal);\n\t\t\t}\n\t\t} finally {\n\t\t\tunlink();\n\t\t\twatcher.abort();\n\t\t}\n\t}\n}\n"]}

package/dist/extensions/llama/client.js

@@ -0,0 +1,309 @@
+function errorMessage(payload, fallback) {
+    if (typeof payload !== "object" || payload === null)
+        return fallback;
+    const error = payload.error;
+    if (typeof error !== "object" || error === null)
+        return fallback;
+    const message = error.message;
+    return typeof message === "string" && message ? message : fallback;
+}
+function isModelInfo(value) {
+    if (typeof value !== "object" || value === null)
+        return false;
+    const candidate = value;
+    return typeof candidate.id === "string" && typeof candidate.status?.value === "string";
+}
+function linkSignal(source, target) {
+    if (!source)
+        return () => { };
+    if (source.aborted) {
+        target.abort(source.reason);
+        return () => { };
+    }
+    const abort = () => target.abort(source.reason);
+    source.addEventListener("abort", abort, { once: true });
+    return () => source.removeEventListener("abort", abort);
+}
+function sleep(ms, signal) {
+    return new Promise((resolve, reject) => {
+        if (signal?.aborted) {
+            reject(signal.reason ?? new Error("Cancelled"));
+            return;
+        }
+        const abort = () => {
+            clearTimeout(timeout);
+            reject(signal?.reason ?? new Error("Cancelled"));
+        };
+        const timeout = setTimeout(() => {
+            signal?.removeEventListener("abort", abort);
+            resolve();
+        }, ms);
+        signal?.addEventListener("abort", abort, { once: true });
+    });
+}
+function parseLoadProgress(data) {
+    if (typeof data !== "object" || data === null)
+        return undefined;
+    const progress = data.progress;
+    if (typeof progress !== "object" || progress === null)
+        return undefined;
+    const value = progress;
+    const stage = typeof value.current === "string" ? value.current : typeof value.stage === "string" ? value.stage : undefined;
+    const stages = Array.isArray(value.stages)
+        ? value.stages.filter((entry) => typeof entry === "string")
+        : [];
+    const stageRatio = typeof value.value === "number" ? Math.max(0, Math.min(1, value.value)) : undefined;
+    let ratio = stageRatio;
+    if (stage && stages.length > 0) {
+        const index = stages.indexOf(stage);
+        if (index >= 0)
+            ratio = (index + (stageRatio ?? 0)) / stages.length;
+    }
+    return {
+        message: stage ? `Loading ${stage.replaceAll("_", " ")}` : "Loading model",
+        ratio,
+    };
+}
+function parseDownloadProgress(data) {
+    if (typeof data !== "object" || data === null)
+        return undefined;
+    const nested = data.progress;
+    const files = typeof nested === "object" && nested !== null ? nested : data;
+    let done = 0;
+    let total = 0;
+    for (const value of Object.values(files)) {
+        if (typeof value !== "object" || value === null)
+            continue;
+        const entry = value;
+        if (typeof entry.done !== "number" || typeof entry.total !== "number")
+            continue;
+        done += entry.done;
+        total += entry.total;
+    }
+    if (total <= 0)
+        return undefined;
+    return {
+        message: "Downloading model",
+        ratio: done / total,
+        detail: `${formatBytes(done)} / ${formatBytes(total)}`,
+    };
+}
+export function formatBytes(bytes) {
+    if (bytes < 1024)
+        return `${bytes} B`;
+    const units = ["KiB", "MiB", "GiB", "TiB"];
+    let value = bytes / 1024;
+    let unit = units[0];
+    for (let index = 1; index < units.length && value >= 1024; index++) {
+        value /= 1024;
+        unit = units[index];
+    }
+    return `${value >= 10 ? value.toFixed(1) : value.toFixed(2)} ${unit}`;
+}
+export function normalizeLlamaServerUrl(value) {
+    const url = new URL(value.trim());
+    if (url.protocol !== "http:" && url.protocol !== "https:") {
+        throw new Error("Server URL must use http or https");
+    }
+    url.hash = "";
+    url.search = "";
+    url.pathname = url.pathname.replace(/\/+$/u, "").replace(/\/v1$/u, "") || "/";
+    return url.toString().replace(/\/$/u, "");
+}
+export function llamaInferenceUrl(serverUrl) {
+    return `${normalizeLlamaServerUrl(serverUrl)}/v1`;
+}
+export class LlamaClient {
+    serverUrl;
+    apiKey;
+    constructor(serverUrl, apiKey) {
+        this.serverUrl = normalizeLlamaServerUrl(serverUrl);
+        this.apiKey = apiKey;
+    }
+    async request(path, init = {}) {
+        const headers = new Headers(init.headers);
+        if (init.body !== undefined)
+            headers.set("Content-Type", "application/json");
+        if (this.apiKey)
+            headers.set("Authorization", `Bearer ${this.apiKey}`);
+        const timeout = AbortSignal.timeout(15_000);
+        const signal = init.signal ? AbortSignal.any([init.signal, timeout]) : timeout;
+        const response = await fetch(`${this.serverUrl}${path}`, { ...init, headers, signal });
+        let payload;
+        try {
+            payload = await response.json();
+        }
+        catch {
+            payload = undefined;
+        }
+        if (!response.ok)
+            throw new Error(errorMessage(payload, `llama.cpp returned HTTP ${response.status}`));
+        return payload;
+    }
+    async list(options = {}) {
+        const payload = await this.request(`/models${options.reload ? "?reload=1" : ""}`, { signal: options.signal });
+        if (typeof payload !== "object" || payload === null || !Array.isArray(payload.data)) {
+            throw new Error("llama.cpp returned an invalid model catalog");
+        }
+        const data = payload.data;
+        if (!data.every(isModelInfo))
+            throw new Error("Server is not running in llama.cpp router mode");
+        return data;
+    }
+    async props(options = {}) {
+        const payload = await this.request("/props", { signal: options.signal });
+        if (typeof payload !== "object" || payload === null)
+            return {};
+        const { models_autoload: modelsAutoload } = payload;
+        return typeof modelsAutoload === "boolean" ? { models_autoload: modelsAutoload } : {};
+    }
+    async load(model, signal) {
+        await this.request("/models/load", { method: "POST", body: JSON.stringify({ model }), signal });
+    }
+    async unload(model, signal) {
+        await this.request("/models/unload", { method: "POST", body: JSON.stringify({ model }), signal });
+    }
+    async unloadAndWait(model, signal) {
+        await this.unload(model, signal);
+        while (true) {
+            const entry = (await this.list({ signal })).find((candidate) => candidate.id === model);
+            if (!entry || entry.status.value === "unloaded")
+                return;
+            await sleep(100, signal);
+        }
+    }
+    async download(model, signal) {
+        await this.request("/models", { method: "POST", body: JSON.stringify({ model }), signal });
+    }
+    async watch(onEvent, signal) {
+        const headers = new Headers();
+        if (this.apiKey)
+            headers.set("Authorization", `Bearer ${this.apiKey}`);
+        const response = await fetch(`${this.serverUrl}/models/sse`, { headers, signal });
+        if (!response.ok || !response.body)
+            throw new Error(`llama.cpp SSE returned HTTP ${response.status}`);
+        const reader = response.body.getReader();
+        const decoder = new TextDecoder();
+        let buffer = "";
+        while (true) {
+            const chunk = await reader.read();
+            if (chunk.done)
+                break;
+            buffer += decoder.decode(chunk.value, { stream: true }).replaceAll("\r\n", "\n");
+            let boundary = buffer.indexOf("\n\n");
+            while (boundary >= 0) {
+                const frame = buffer.slice(0, boundary);
+                buffer = buffer.slice(boundary + 2);
+                const data = frame
+                    .split("\n")
+                    .filter((line) => line.startsWith("data:"))
+                    .map((line) => line.slice(5).trimStart())
+                    .join("\n");
+                if (data) {
+                    try {
+                        const event = JSON.parse(data);
+                        if (event && typeof event.model === "string" && typeof event.event === "string")
+                            onEvent(event);
+                    }
+                    catch {
+                        // Ignore malformed events; catalog polling remains authoritative.
+                    }
+                }
+                boundary = buffer.indexOf("\n\n");
+            }
+        }
+    }
+    async loadAndWait(model, onProgress, signal) {
+        const watcher = new AbortController();
+        const unlink = linkSignal(signal, watcher);
+        let eventLoaded = false;
+        let eventError;
+        void this.watch((event) => {
+            if (event.model !== model)
+                return;
+            if (event.event !== "model_status" && event.event !== "status_change")
+                return;
+            const data = event.data;
+            if (data?.status === "loaded")
+                eventLoaded = true;
+            if (data?.status === "unloaded")
+                eventError = "Model failed to load";
+            const progress = parseLoadProgress(event.data);
+            if (progress)
+                onProgress(progress);
+        }, watcher.signal).catch(() => { });
+        try {
+            await this.load(model, signal);
+            onProgress({ message: "Loading model" });
+            while (true) {
+                if (signal?.aborted)
+                    throw signal.reason ?? new Error("Cancelled");
+                const entry = (await this.list({ signal })).find((candidate) => candidate.id === model);
+                if (entry?.status.value === "loaded")
+                    return entry;
+                if (eventLoaded && !entry)
+                    return { id: model, status: { value: "loaded" } };
+                if (entry?.status.failed || eventError) {
+                    throw new Error(entry?.status.exit_code === undefined
+                        ? (eventError ?? "Model failed to load")
+                        : `Model exited with code ${entry.status.exit_code}`);
+                }
+                await sleep(250, signal);
+            }
+        }
+        finally {
+            unlink();
+            watcher.abort();
+        }
+    }
+    async downloadAndWait(model, onProgress, signal) {
+        const watcher = new AbortController();
+        const unlink = linkSignal(signal, watcher);
+        let finished = false;
+        let failure;
+        let sawDownloading = false;
+        let polls = 0;
+        void this.watch((event) => {
+            if (event.model !== model)
+                return;
+            if (event.event === "download_finished")
+                finished = true;
+            if (event.event === "download_failed")
+                failure = errorMessage(event.data, "Download failed");
+            if (event.event === "download_progress") {
+                sawDownloading = true;
+                const progress = parseDownloadProgress(event.data);
+                if (progress)
+                    onProgress(progress);
+            }
+        }, watcher.signal).catch(() => { });
+        try {
+            await this.download(model, signal);
+            onProgress({ message: "Downloading model" });
+            while (true) {
+                if (signal?.aborted)
+                    throw signal.reason ?? new Error("Cancelled");
+                if (failure)
+                    throw new Error(failure);
+                const models = await this.list({ signal });
+                polls++;
+                const entry = models.find((candidate) => candidate.id === model);
+                if (entry?.status.value === "downloading") {
+                    sawDownloading = true;
+                    const progress = parseDownloadProgress(entry.status.progress);
+                    if (progress)
+                        onProgress(progress);
+                }
+                else if (finished || (entry && (sawDownloading || polls >= 2))) {
+                    return this.list({ reload: true, signal });
+                }
+                await sleep(500, signal);
+            }
+        }
+        finally {
+            unlink();
+            watcher.abort();
+        }
+    }
+}
+//# sourceMappingURL=client.js.map

package/dist/extensions/llama/client.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.js","sourceRoot":"","sources":["../../../src/extensions/llama/client.ts"],"names":[],"mappings":"AA8CA,SAAS,YAAY,CAAC,OAAgB,EAAE,QAAgB,EAAU;IACjE,IAAI,OAAO,OAAO,KAAK,QAAQ,IAAI,OAAO,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAC;IACrE,MAAM,KAAK,GAAI,OAA+B,CAAC,KAAK,CAAC;IACrD,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAC;IACjE,MAAM,OAAO,GAAI,KAA+B,CAAC,OAAO,CAAC;IACzD,OAAO,OAAO,OAAO,KAAK,QAAQ,IAAI,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,QAAQ,CAAC;AAAA,CACnE;AAED,SAAS,WAAW,CAAC,KAAc,EAA2B;IAC7D,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,KAAK,CAAC;IAC9D,MAAM,SAAS,GAAG,KAAuD,CAAC;IAC1E,OAAO,OAAO,SAAS,CAAC,EAAE,KAAK,QAAQ,IAAI,OAAO,SAAS,CAAC,MAAM,EAAE,KAAK,KAAK,QAAQ,CAAC;AAAA,CACvF;AAED,SAAS,UAAU,CAAC,MAA+B,EAAE,MAAuB,EAAc;IACzF,IAAI,CAAC,MAAM;QAAE,OAAO,GAAG,EAAE,CAAC,EAAC,CAAC,CAAC;IAC7B,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC;QACpB,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;QAC5B,OAAO,GAAG,EAAE,CAAC,EAAC,CAAC,CAAC;IACjB,CAAC;IACD,MAAM,KAAK,GAAG,GAAG,EAAE,CAAC,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;IAChD,MAAM,CAAC,gBAAgB,CAAC,OAAO,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC;IACxD,OAAO,GAAG,EAAE,CAAC,MAAM,CAAC,mBAAmB,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;AAAA,CACxD;AAED,SAAS,KAAK,CAAC,EAAU,EAAE,MAAoB,EAAiB;IAC/D,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,CAAC;QACvC,IAAI,MAAM,EAAE,OAAO,EAAE,CAAC;YACrB,MAAM,CAAC,MAAM,CAAC,MAAM,IAAI,IAAI,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;YAChD,OAAO;QACR,CAAC;QACD,MAAM,KAAK,GAAG,GAAG,EAAE,CAAC;YACnB,YAAY,CAAC,OAAO,CAAC,CAAC;YACtB,MAAM,CAAC,MAAM,EAAE,MAAM,IAAI,IAAI,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;QAAA,CACjD,CAAC;QACF,MAAM,OAAO,GAAG,UAAU,CAAC,GAAG,EAAE,CAAC;YAChC,MAAM,EAAE,mBAAmB,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;YAC5C,OAAO,EAAE,CAAC;QAAA,CACV,EAAE,EAAE,CAAC,CAAC;QACP,MAAM,EAAE,gBAAgB,CAAC,OAAO,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC;IAAA,CACzD,CAAC,CAAC;AAAA,CACH;AAED,SAAS,iBAAiB,CAAC,IAAa,EAA6B;IACpE,IAAI,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,KAAK,IAAI;QAAE,OAAO,SAAS,CAAC;IAChE,MAAM,QAAQ,GAAI,IAA+B,CAAC,QAAQ,CAAC;IAC3D,IAAI,OAAO,QAAQ,KAAK,QAAQ,IAAI,QAAQ,KAAK,IAAI;QAAE,OAAO,SAAS,CAAC;IACxE,MAAM,KAAK,GAAG,QAAqF,CAAC;IACpG,MAAM,KAAK,GACV,OAAO,KAAK,CAAC,OAAO,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/G,MAAM,MAAM,GAAG,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC;QACzC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,EAAmB,EAAE,CAAC,OAAO,KAAK,KAAK,QAAQ,CAAC;QAC5E,CAAC,CAAC,EAAE,CAAC;IACN,MAAM,UAAU,GAAG,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACvG,IAAI,KAAK,GAAG,UAAU,CAAC;IACvB,IAAI,KAAK,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAChC,MAAM,KAAK,GAAG,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;QACpC,IAAI,KAAK,IAAI,CAAC;YAAE,KAAK,GAAG,CAAC,KAAK,GAAG,CAAC,UAAU,IAAI,CAAC,CAAC,CAAC,GAAG,MAAM,CAAC,MAAM,CAAC;IACrE,CAAC;IACD,OAAO;QACN,OAAO,EAAE,KAAK,CAAC,CAAC,CAAC,WAAW,KAAK,CAAC,UAAU,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,eAAe;QAC1E,KAAK;KACL,CAAC;AAAA,CACF;AAED,SAAS,qBAAqB,CAAC,IAAa,EAA6B;IACxE,IAAI,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,KAAK,IAAI;QAAE,OAAO,SAAS,CAAC;IAChE,MAAM,MAAM,GAAI,IAA+B,CAAC,QAAQ,CAAC;IACzD,MAAM,KAAK,GAAG,OAAO,MAAM,KAAK,QAAQ,IAAI,MAAM,KAAK,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC;IAC5E,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,KAAK,MAAM,KAAK,IAAI,MAAM,CAAC,MAAM,CAAC,KAAgC,CAAC,EAAE,CAAC;QACrE,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,KAAK,IAAI;YAAE,SAAS;QAC1D,MAAM,KAAK,GAAG,KAA4C,CAAC;QAC3D,IAAI,OAAO,KAAK,CAAC,IAAI,KAAK,QAAQ,IAAI,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ;YAAE,SAAS;QAChF,IAAI,IAAI,KAAK,CAAC,IAAI,CAAC;QACnB,KAAK,IAAI,KAAK,CAAC,KAAK,CAAC;IACtB,CAAC;IACD,IAAI,KAAK,IAAI,CAAC;QAAE,OAAO,SAAS,CAAC;IACjC,OAAO;QACN,OAAO,EAAE,mBAAmB;QAC5B,KAAK,EAAE,IAAI,GAAG,KAAK;QACnB,MAAM,EAAE,GAAG,WAAW,CAAC,IAAI,CAAC,MAAM,WAAW,CAAC,KAAK,CAAC,EAAE;KACtD,CAAC;AAAA,CACF;AAED,MAAM,UAAU,WAAW,CAAC,KAAa,EAAU;IAClD,IAAI,KAAK,GAAG,IAAI;QAAE,OAAO,GAAG,KAAK,IAAI,CAAC;IACtC,MAAM,KAAK,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,CAAC,CAAC;IAC3C,IAAI,KAAK,GAAG,KAAK,GAAG,IAAI,CAAC;IACzB,IAAI,IAAI,GAAG,KAAK,CAAC,CAAC,CAAE,CAAC;IACrB,KAAK,IAAI,KAAK,GAAG,CAAC,EAAE,KAAK,GAAG,KAAK,CAAC,MAAM,IAAI,KAAK,IAAI,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC;QACpE,KAAK,IAAI,IAAI,CAAC;QACd,IAAI,GAAG,KAAK,CAAC,KAAK,CAAE,CAAC;IACtB,CAAC;IACD,OAAO,GAAG,KAAK,IAAI,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,IAAI,EAAE,CAAC;AAAA,CACtE;AAED,MAAM,UAAU,uBAAuB,CAAC,KAAa,EAAU;IAC9D,MAAM,GAAG,GAAG,IAAI,GAAG,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC;IAClC,IAAI,GAAG,CAAC,QAAQ,KAAK,OAAO,IAAI,GAAG,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;QAC3D,MAAM,IAAI,KAAK,CAAC,mCAAmC,CAAC,CAAC;IACtD,CAAC;IACD,GAAG,CAAC,IAAI,GAAG,EAAE,CAAC;IACd,GAAG,CAAC,MAAM,GAAG,EAAE,CAAC;IAChB,GAAG,CAAC,QAAQ,GAAG,GAAG,CAAC,QAAQ,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,OAAO,CAAC,QAAQ,EAAE,EAAE,CAAC,IAAI,GAAG,CAAC;IAC9E,OAAO,GAAG,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;AAAA,CAC1C;AAED,MAAM,UAAU,iBAAiB,CAAC,SAAiB,EAAU;IAC5D,OAAO,GAAG,uBAAuB,CAAC,SAAS,CAAC,KAAK,CAAC;AAAA,CAClD;AAED,MAAM,OAAO,WAAW;IACd,SAAS,CAAS;IACV,MAAM,CAAqB;IAE5C,YAAY,SAAiB,EAAE,MAAe,EAAE;QAC/C,IAAI,CAAC,SAAS,GAAG,uBAAuB,CAAC,SAAS,CAAC,CAAC;QACpD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;IAAA,CACrB;IAEO,KAAK,CAAC,OAAO,CAAC,IAAY,EAAE,IAAI,GAAgB,EAAE,EAAoB;QAC7E,MAAM,OAAO,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;QAC1C,IAAI,IAAI,CAAC,IAAI,KAAK,SAAS;YAAE,OAAO,CAAC,GAAG,CAAC,cAAc,EAAE,kBAAkB,CAAC,CAAC;QAC7E,IAAI,IAAI,CAAC,MAAM;YAAE,OAAO,CAAC,GAAG,CAAC,eAAe,EAAE,UAAU,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC;QACvE,MAAM,OAAO,GAAG,WAAW,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QAC5C,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC;QAC/E,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,IAAI,CAAC,SAAS,GAAG,IAAI,EAAE,EAAE,EAAE,GAAG,IAAI,EAAE,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC;QACvF,IAAI,OAAgB,CAAC;QACrB,IAAI,CAAC;YACJ,OAAO,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;QACjC,CAAC;QAAC,MAAM,CAAC;YACR,OAAO,GAAG,SAAS,CAAC;QACrB,CAAC;QACD,IAAI,CAAC,QAAQ,CAAC,EAAE;YAAE,MAAM,IAAI,KAAK,CAAC,YAAY,CAAC,OAAO,EAAE,2BAA2B,QAAQ,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;QACvG,OAAO,OAAO,CAAC;IAAA,CACf;IAED,KAAK,CAAC,IAAI,CAAC,OAAO,GAA+C,EAAE,EAA6B;QAC/F,MAAM,OAAO,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,UAAU,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;QAC9G,IAAI,OAAO,OAAO,KAAK,QAAQ,IAAI,OAAO,KAAK,IAAI,IAAI,CAAC,KAAK,CAAC,OAAO,CAAE,OAA8B,CAAC,IAAI,CAAC,EAAE,CAAC;YAC7G,MAAM,IAAI,KAAK,CAAC,6CAA6C,CAAC,CAAC;QAChE,CAAC;QACD,MAAM,IAAI,GAAI,OAA+B,CAAC,IAAI,CAAC;QACnD,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC;YAAE,MAAM,IAAI,KAAK,CAAC,gDAAgD,CAAC,CAAC;QAChG,OAAO,IAAI,CAAC;IAAA,CACZ;IAED,KAAK,CAAC,KAAK,CAAC,OAAO,GAA6B,EAAE,EAA6B;QAC9E,MAAM,OAAO,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;QACzE,IAAI,OAAO,OAAO,KAAK,QAAQ,IAAI,OAAO,KAAK,IAAI;YAAE,OAAO,EAAE,CAAC;QAC/D,MAAM,EAAE,eAAe,EAAE,cAAc,EAAE,GAAG,OAAkC,CAAC;QAC/E,OAAO,OAAO,cAAc,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,eAAe,EAAE,cAAc,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;IAAA,CACtF;IAED,KAAK,CAAC,IAAI,CAAC,KAAa,EAAE,MAAoB,EAAiB;QAC9D,MAAM,IAAI,CAAC,OAAO,CAAC,cAAc,EAAE,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC;IAAA,CAChG;IAED,KAAK,CAAC,MAAM,CAAC,KAAa,EAAE,MAAoB,EAAiB;QAChE,MAAM,IAAI,CAAC,OAAO,CAAC,gBAAgB,EAAE,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC;IAAA,CAClG;IAED,KAAK,CAAC,aAAa,CAAC,KAAa,EAAE,MAAoB,EAAiB;QACvE,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;QACjC,OAAO,IAAI,EAAE,CAAC;YACb,MAAM,KAAK,GAAG,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,CAAC,EAAE,KAAK,KAAK,CAAC,CAAC;YACxF,IAAI,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,CAAC,KAAK,KAAK,UAAU;gBAAE,OAAO;YACxD,MAAM,KAAK,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;QAC1B,CAAC;IAAA,CACD;IAED,KAAK,CAAC,QAAQ,CAAC,KAAa,EAAE,MAAoB,EAAiB;QAClE,MAAM,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC;IAAA,CAC3F;IAED,KAAK,CAAC,KAAK,CAAC,OAAyC,EAAE,MAAoB,EAAiB;QAC3F,MAAM,OAAO,GAAG,IAAI,OAAO,EAAE,CAAC;QAC9B,IAAI,IAAI,CAAC,MAAM;YAAE,OAAO,CAAC,GAAG,CAAC,eAAe,EAAE,UAAU,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC;QACvE,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,IAAI,CAAC,SAAS,aAAa,EAAE,EAAE,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC;QAClF,IAAI,CAAC,QAAQ,CAAC,EAAE,IAAI,CAAC,QAAQ,CAAC,IAAI;YAAE,MAAM,IAAI,KAAK,CAAC,+BAA+B,QAAQ,CAAC,MAAM,EAAE,CAAC,CAAC;QACtG,MAAM,MAAM,GAAG,QAAQ,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC;QACzC,MAAM,OAAO,GAAG,IAAI,WAAW,EAAE,CAAC;QAClC,IAAI,MAAM,GAAG,EAAE,CAAC;QAChB,OAAO,IAAI,EAAE,CAAC;YACb,MAAM,KAAK,GAAG,MAAM,MAAM,CAAC,IAAI,EAAE,CAAC;YAClC,IAAI,KAAK,CAAC,IAAI;gBAAE,MAAM;YACtB,MAAM,IAAI,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAC,UAAU,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;YACjF,IAAI,QAAQ,GAAG,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;YACtC,OAAO,QAAQ,IAAI,CAAC,EAAE,CAAC;gBACtB,MAAM,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,QAAQ,CAAC,CAAC;gBACxC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,QAAQ,GAAG,CAAC,CAAC,CAAC;gBACpC,MAAM,IAAI,GAAG,KAAK;qBAChB,KAAK,CAAC,IAAI,CAAC;qBACX,MAAM,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,UAAU,CAAC,OAAO,CAAC,CAAC;qBAC1C,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,EAAE,CAAC;qBACxC,IAAI,CAAC,IAAI,CAAC,CAAC;gBACb,IAAI,IAAI,EAAE,CAAC;oBACV,IAAI,CAAC;wBACJ,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAoB,CAAC;wBAClD,IAAI,KAAK,IAAI,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,IAAI,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ;4BAAE,OAAO,CAAC,KAAK,CAAC,CAAC;oBACjG,CAAC;oBAAC,MAAM,CAAC;wBACR,kEAAkE;oBACnE,CAAC;gBACF,CAAC;gBACD,QAAQ,GAAG,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;YACnC,CAAC;QACF,CAAC;IAAA,CACD;IAED,KAAK,CAAC,WAAW,CAChB,KAAa,EACb,UAA6C,EAC7C,MAAoB,EACM;QAC1B,MAAM,OAAO,GAAG,IAAI,eAAe,EAAE,CAAC;QACtC,MAAM,MAAM,GAAG,UAAU,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC3C,IAAI,WAAW,GAAG,KAAK,CAAC;QACxB,IAAI,UAA8B,CAAC;QACnC,KAAK,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC;YAC1B,IAAI,KAAK,CAAC,KAAK,KAAK,KAAK;gBAAE,OAAO;YAClC,IAAI,KAAK,CAAC,KAAK,KAAK,cAAc,IAAI,KAAK,CAAC,KAAK,KAAK,eAAe;gBAAE,OAAO;YAC9E,MAAM,IAAI,GAAG,KAAK,CAAC,IAAwC,CAAC;YAC5D,IAAI,IAAI,EAAE,MAAM,KAAK,QAAQ;gBAAE,WAAW,GAAG,IAAI,CAAC;YAClD,IAAI,IAAI,EAAE,MAAM,KAAK,UAAU;gBAAE,UAAU,GAAG,sBAAsB,CAAC;YACrE,MAAM,QAAQ,GAAG,iBAAiB,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;YAC/C,IAAI,QAAQ;gBAAE,UAAU,CAAC,QAAQ,CAAC,CAAC;QAAA,CACnC,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,EAAC,CAAC,CAAC,CAAC;QACnC,IAAI,CAAC;YACJ,MAAM,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;YAC/B,UAAU,CAAC,EAAE,OAAO,EAAE,eAAe,EAAE,CAAC,CAAC;YACzC,OAAO,IAAI,EAAE,CAAC;gBACb,IAAI,MAAM,EAAE,OAAO;oBAAE,MAAM,MAAM,CAAC,MAAM,IAAI,IAAI,KAAK,CAAC,WAAW,CAAC,CAAC;gBACnE,MAAM,KAAK,GAAG,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,CAAC,EAAE,KAAK,KAAK,CAAC,CAAC;gBACxF,IAAI,KAAK,EAAE,MAAM,CAAC,KAAK,KAAK,QAAQ;oBAAE,OAAO,KAAK,CAAC;gBACnD,IAAI,WAAW,IAAI,CAAC,KAAK;oBAAE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,EAAE,KAAK,EAAE,QAAQ,EAAE,EAAE,CAAC;gBAC7E,IAAI,KAAK,EAAE,MAAM,CAAC,MAAM,IAAI,UAAU,EAAE,CAAC;oBACxC,MAAM,IAAI,KAAK,CACd,KAAK,EAAE,MAAM,CAAC,SAAS,KAAK,SAAS;wBACpC,CAAC,CAAC,CAAC,UAAU,IAAI,sBAAsB,CAAC;wBACxC,CAAC,CAAC,0BAA0B,KAAK,CAAC,MAAM,CAAC,SAAS,EAAE,CACrD,CAAC;gBACH,CAAC;gBACD,MAAM,KAAK,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;YAC1B,CAAC;QACF,CAAC;gBAAS,CAAC;YACV,MAAM,EAAE,CAAC;YACT,OAAO,CAAC,KAAK,EAAE,CAAC;QACjB,CAAC;IAAA,CACD;IAED,KAAK,CAAC,eAAe,CACpB,KAAa,EACb,UAA6C,EAC7C,MAAoB,EACQ;QAC5B,MAAM,OAAO,GAAG,IAAI,eAAe,EAAE,CAAC;QACtC,MAAM,MAAM,GAAG,UAAU,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC3C,IAAI,QAAQ,GAAG,KAAK,CAAC;QACrB,IAAI,OAA2B,CAAC;QAChC,IAAI,cAAc,GAAG,KAAK,CAAC;QAC3B,IAAI,KAAK,GAAG,CAAC,CAAC;QACd,KAAK,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC;YAC1B,IAAI,KAAK,CAAC,KAAK,KAAK,KAAK;gBAAE,OAAO;YAClC,IAAI,KAAK,CAAC,KAAK,KAAK,mBAAmB;gBAAE,QAAQ,GAAG,IAAI,CAAC;YACzD,IAAI,KAAK,CAAC,KAAK,KAAK,iBAAiB;gBAAE,OAAO,GAAG,YAAY,CAAC,KAAK,CAAC,IAAI,EAAE,iBAAiB,CAAC,CAAC;YAC7F,IAAI,KAAK,CAAC,KAAK,KAAK,mBAAmB,EAAE,CAAC;gBACzC,cAAc,GAAG,IAAI,CAAC;gBACtB,MAAM,QAAQ,GAAG,qBAAqB,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBACnD,IAAI,QAAQ;oBAAE,UAAU,CAAC,QAAQ,CAAC,CAAC;YACpC,CAAC;QAAA,CACD,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,EAAC,CAAC,CAAC,CAAC;QACnC,IAAI,CAAC;YACJ,MAAM,IAAI,CAAC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;YACnC,UAAU,CAAC,EAAE,OAAO,EAAE,mBAAmB,EAAE,CAAC,CAAC;YAC7C,OAAO,IAAI,EAAE,CAAC;gBACb,IAAI,MAAM,EAAE,OAAO;oBAAE,MAAM,MAAM,CAAC,MAAM,IAAI,IAAI,KAAK,CAAC,WAAW,CAAC,CAAC;gBACnE,IAAI,OAAO;oBAAE,MAAM,IAAI,KAAK,CAAC,OAAO,CAAC,CAAC;gBACtC,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC;gBAC3C,KAAK,EAAE,CAAC;gBACR,MAAM,KAAK,GAAG,MAAM,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,CAAC,EAAE,KAAK,KAAK,CAAC,CAAC;gBACjE,IAAI,KAAK,EAAE,MAAM,CAAC,KAAK,KAAK,aAAa,EAAE,CAAC;oBAC3C,cAAc,GAAG,IAAI,CAAC;oBACtB,MAAM,QAAQ,GAAG,qBAAqB,CAAC,KAAK,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;oBAC9D,IAAI,QAAQ;wBAAE,UAAU,CAAC,QAAQ,CAAC,CAAC;gBACpC,CAAC;qBAAM,IAAI,QAAQ,IAAI,CAAC,KAAK,IAAI,CAAC,cAAc,IAAI,KAAK,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC;oBAClE,OAAO,IAAI,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,CAAC,CAAC;gBAC5C,CAAC;gBACD,MAAM,KAAK,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;YAC1B,CAAC;QACF,CAAC;gBAAS,CAAC;YACV,MAAM,EAAE,CAAC;YACT,OAAO,CAAC,KAAK,EAAE,CAAC;QACjB,CAAC;IAAA,CACD;CACD","sourcesContent":["export type LlamaModelStatus = \"unloaded\" | \"loading\" | \"loaded\" | \"downloading\" | \"sleeping\";\n\nexport interface LlamaModelInfo {\n\tid: string;\n\taliases?: string[];\n\tstatus: {\n\t\tvalue: LlamaModelStatus;\n\t\targs?: string[];\n\t\tfailed?: boolean;\n\t\texit_code?: number;\n\t\tprogress?: Record<string, { done: number; total: number }>;\n\t};\n\tarchitecture?: {\n\t\tinput_modalities?: string[];\n\t\toutput_modalities?: string[];\n\t};\n\tsource?: string;\n\tmeta?: {\n\t\tn_ctx?: number;\n\t\tn_ctx_train?: number;\n\t\tsize?: number;\n\t\tftype?: string;\n\t};\n}\n\nexport interface LlamaModelsResponse {\n\tdata: LlamaModelInfo[];\n\tobject?: string;\n}\n\nexport interface LlamaServerProps {\n\tmodels_autoload?: boolean;\n}\n\nexport interface LlamaModelEvent {\n\tmodel: string;\n\tevent: string;\n\tdata?: unknown;\n}\n\nexport interface LlamaProgress {\n\tmessage: string;\n\tratio?: number;\n\tdetail?: string;\n}\n\nfunction errorMessage(payload: unknown, fallback: string): string {\n\tif (typeof payload !== \"object\" || payload === null) return fallback;\n\tconst error = (payload as { error?: unknown }).error;\n\tif (typeof error !== \"object\" || error === null) return fallback;\n\tconst message = (error as { message?: unknown }).message;\n\treturn typeof message === \"string\" && message ? message : fallback;\n}\n\nfunction isModelInfo(value: unknown): value is LlamaModelInfo {\n\tif (typeof value !== \"object\" || value === null) return false;\n\tconst candidate = value as { id?: unknown; status?: { value?: unknown } };\n\treturn typeof candidate.id === \"string\" && typeof candidate.status?.value === \"string\";\n}\n\nfunction linkSignal(source: AbortSignal | undefined, target: AbortController): () => void {\n\tif (!source) return () => {};\n\tif (source.aborted) {\n\t\ttarget.abort(source.reason);\n\t\treturn () => {};\n\t}\n\tconst abort = () => target.abort(source.reason);\n\tsource.addEventListener(\"abort\", abort, { once: true });\n\treturn () => source.removeEventListener(\"abort\", abort);\n}\n\nfunction sleep(ms: number, signal?: AbortSignal): Promise<void> {\n\treturn new Promise((resolve, reject) => {\n\t\tif (signal?.aborted) {\n\t\t\treject(signal.reason ?? new Error(\"Cancelled\"));\n\t\t\treturn;\n\t\t}\n\t\tconst abort = () => {\n\t\t\tclearTimeout(timeout);\n\t\t\treject(signal?.reason ?? new Error(\"Cancelled\"));\n\t\t};\n\t\tconst timeout = setTimeout(() => {\n\t\t\tsignal?.removeEventListener(\"abort\", abort);\n\t\t\tresolve();\n\t\t}, ms);\n\t\tsignal?.addEventListener(\"abort\", abort, { once: true });\n\t});\n}\n\nfunction parseLoadProgress(data: unknown): LlamaProgress | undefined {\n\tif (typeof data !== \"object\" || data === null) return undefined;\n\tconst progress = (data as { progress?: unknown }).progress;\n\tif (typeof progress !== \"object\" || progress === null) return undefined;\n\tconst value = progress as { stages?: unknown; current?: unknown; stage?: unknown; value?: unknown };\n\tconst stage =\n\t\ttypeof value.current === \"string\" ? value.current : typeof value.stage === \"string\" ? value.stage : undefined;\n\tconst stages = Array.isArray(value.stages)\n\t\t? value.stages.filter((entry): entry is string => typeof entry === \"string\")\n\t\t: [];\n\tconst stageRatio = typeof value.value === \"number\" ? Math.max(0, Math.min(1, value.value)) : undefined;\n\tlet ratio = stageRatio;\n\tif (stage && stages.length > 0) {\n\t\tconst index = stages.indexOf(stage);\n\t\tif (index >= 0) ratio = (index + (stageRatio ?? 0)) / stages.length;\n\t}\n\treturn {\n\t\tmessage: stage ? `Loading ${stage.replaceAll(\"_\", \" \")}` : \"Loading model\",\n\t\tratio,\n\t};\n}\n\nfunction parseDownloadProgress(data: unknown): LlamaProgress | undefined {\n\tif (typeof data !== \"object\" || data === null) return undefined;\n\tconst nested = (data as { progress?: unknown }).progress;\n\tconst files = typeof nested === \"object\" && nested !== null ? nested : data;\n\tlet done = 0;\n\tlet total = 0;\n\tfor (const value of Object.values(files as Record<string, unknown>)) {\n\t\tif (typeof value !== \"object\" || value === null) continue;\n\t\tconst entry = value as { done?: unknown; total?: unknown };\n\t\tif (typeof entry.done !== \"number\" || typeof entry.total !== \"number\") continue;\n\t\tdone += entry.done;\n\t\ttotal += entry.total;\n\t}\n\tif (total <= 0) return undefined;\n\treturn {\n\t\tmessage: \"Downloading model\",\n\t\tratio: done / total,\n\t\tdetail: `${formatBytes(done)} / ${formatBytes(total)}`,\n\t};\n}\n\nexport function formatBytes(bytes: number): string {\n\tif (bytes < 1024) return `${bytes} B`;\n\tconst units = [\"KiB\", \"MiB\", \"GiB\", \"TiB\"];\n\tlet value = bytes / 1024;\n\tlet unit = units[0]!;\n\tfor (let index = 1; index < units.length && value >= 1024; index++) {\n\t\tvalue /= 1024;\n\t\tunit = units[index]!;\n\t}\n\treturn `${value >= 10 ? value.toFixed(1) : value.toFixed(2)} ${unit}`;\n}\n\nexport function normalizeLlamaServerUrl(value: string): string {\n\tconst url = new URL(value.trim());\n\tif (url.protocol !== \"http:\" && url.protocol !== \"https:\") {\n\t\tthrow new Error(\"Server URL must use http or https\");\n\t}\n\turl.hash = \"\";\n\turl.search = \"\";\n\turl.pathname = url.pathname.replace(/\\/+$/u, \"\").replace(/\\/v1$/u, \"\") || \"/\";\n\treturn url.toString().replace(/\\/$/u, \"\");\n}\n\nexport function llamaInferenceUrl(serverUrl: string): string {\n\treturn `${normalizeLlamaServerUrl(serverUrl)}/v1`;\n}\n\nexport class LlamaClient {\n\treadonly serverUrl: string;\n\tprivate readonly apiKey: string | undefined;\n\n\tconstructor(serverUrl: string, apiKey?: string) {\n\t\tthis.serverUrl = normalizeLlamaServerUrl(serverUrl);\n\t\tthis.apiKey = apiKey;\n\t}\n\n\tprivate async request(path: string, init: RequestInit = {}): Promise<unknown> {\n\t\tconst headers = new Headers(init.headers);\n\t\tif (init.body !== undefined) headers.set(\"Content-Type\", \"application/json\");\n\t\tif (this.apiKey) headers.set(\"Authorization\", `Bearer ${this.apiKey}`);\n\t\tconst timeout = AbortSignal.timeout(15_000);\n\t\tconst signal = init.signal ? AbortSignal.any([init.signal, timeout]) : timeout;\n\t\tconst response = await fetch(`${this.serverUrl}${path}`, { ...init, headers, signal });\n\t\tlet payload: unknown;\n\t\ttry {\n\t\t\tpayload = await response.json();\n\t\t} catch {\n\t\t\tpayload = undefined;\n\t\t}\n\t\tif (!response.ok) throw new Error(errorMessage(payload, `llama.cpp returned HTTP ${response.status}`));\n\t\treturn payload;\n\t}\n\n\tasync list(options: { reload?: boolean; signal?: AbortSignal } = {}): Promise<LlamaModelInfo[]> {\n\t\tconst payload = await this.request(`/models${options.reload ? \"?reload=1\" : \"\"}`, { signal: options.signal });\n\t\tif (typeof payload !== \"object\" || payload === null || !Array.isArray((payload as { data?: unknown }).data)) {\n\t\t\tthrow new Error(\"llama.cpp returned an invalid model catalog\");\n\t\t}\n\t\tconst data = (payload as { data: unknown[] }).data;\n\t\tif (!data.every(isModelInfo)) throw new Error(\"Server is not running in llama.cpp router mode\");\n\t\treturn data;\n\t}\n\n\tasync props(options: { signal?: AbortSignal } = {}): Promise<LlamaServerProps> {\n\t\tconst payload = await this.request(\"/props\", { signal: options.signal });\n\t\tif (typeof payload !== \"object\" || payload === null) return {};\n\t\tconst { models_autoload: modelsAutoload } = payload as Record<string, unknown>;\n\t\treturn typeof modelsAutoload === \"boolean\" ? { models_autoload: modelsAutoload } : {};\n\t}\n\n\tasync load(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.request(\"/models/load\", { method: \"POST\", body: JSON.stringify({ model }), signal });\n\t}\n\n\tasync unload(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.request(\"/models/unload\", { method: \"POST\", body: JSON.stringify({ model }), signal });\n\t}\n\n\tasync unloadAndWait(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.unload(model, signal);\n\t\twhile (true) {\n\t\t\tconst entry = (await this.list({ signal })).find((candidate) => candidate.id === model);\n\t\t\tif (!entry || entry.status.value === \"unloaded\") return;\n\t\t\tawait sleep(100, signal);\n\t\t}\n\t}\n\n\tasync download(model: string, signal?: AbortSignal): Promise<void> {\n\t\tawait this.request(\"/models\", { method: \"POST\", body: JSON.stringify({ model }), signal });\n\t}\n\n\tasync watch(onEvent: (event: LlamaModelEvent) => void, signal?: AbortSignal): Promise<void> {\n\t\tconst headers = new Headers();\n\t\tif (this.apiKey) headers.set(\"Authorization\", `Bearer ${this.apiKey}`);\n\t\tconst response = await fetch(`${this.serverUrl}/models/sse`, { headers, signal });\n\t\tif (!response.ok || !response.body) throw new Error(`llama.cpp SSE returned HTTP ${response.status}`);\n\t\tconst reader = response.body.getReader();\n\t\tconst decoder = new TextDecoder();\n\t\tlet buffer = \"\";\n\t\twhile (true) {\n\t\t\tconst chunk = await reader.read();\n\t\t\tif (chunk.done) break;\n\t\t\tbuffer += decoder.decode(chunk.value, { stream: true }).replaceAll(\"\\r\\n\", \"\\n\");\n\t\t\tlet boundary = buffer.indexOf(\"\\n\\n\");\n\t\t\twhile (boundary >= 0) {\n\t\t\t\tconst frame = buffer.slice(0, boundary);\n\t\t\t\tbuffer = buffer.slice(boundary + 2);\n\t\t\t\tconst data = frame\n\t\t\t\t\t.split(\"\\n\")\n\t\t\t\t\t.filter((line) => line.startsWith(\"data:\"))\n\t\t\t\t\t.map((line) => line.slice(5).trimStart())\n\t\t\t\t\t.join(\"\\n\");\n\t\t\t\tif (data) {\n\t\t\t\t\ttry {\n\t\t\t\t\t\tconst event = JSON.parse(data) as LlamaModelEvent;\n\t\t\t\t\t\tif (event && typeof event.model === \"string\" && typeof event.event === \"string\") onEvent(event);\n\t\t\t\t\t} catch {\n\t\t\t\t\t\t// Ignore malformed events; catalog polling remains authoritative.\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\tboundary = buffer.indexOf(\"\\n\\n\");\n\t\t\t}\n\t\t}\n\t}\n\n\tasync loadAndWait(\n\t\tmodel: string,\n\t\tonProgress: (progress: LlamaProgress) => void,\n\t\tsignal?: AbortSignal,\n\t): Promise<LlamaModelInfo> {\n\t\tconst watcher = new AbortController();\n\t\tconst unlink = linkSignal(signal, watcher);\n\t\tlet eventLoaded = false;\n\t\tlet eventError: string | undefined;\n\t\tvoid this.watch((event) => {\n\t\t\tif (event.model !== model) return;\n\t\t\tif (event.event !== \"model_status\" && event.event !== \"status_change\") return;\n\t\t\tconst data = event.data as { status?: unknown } | undefined;\n\t\t\tif (data?.status === \"loaded\") eventLoaded = true;\n\t\t\tif (data?.status === \"unloaded\") eventError = \"Model failed to load\";\n\t\t\tconst progress = parseLoadProgress(event.data);\n\t\t\tif (progress) onProgress(progress);\n\t\t}, watcher.signal).catch(() => {});\n\t\ttry {\n\t\t\tawait this.load(model, signal);\n\t\t\tonProgress({ message: \"Loading model\" });\n\t\t\twhile (true) {\n\t\t\t\tif (signal?.aborted) throw signal.reason ?? new Error(\"Cancelled\");\n\t\t\t\tconst entry = (await this.list({ signal })).find((candidate) => candidate.id === model);\n\t\t\t\tif (entry?.status.value === \"loaded\") return entry;\n\t\t\t\tif (eventLoaded && !entry) return { id: model, status: { value: \"loaded\" } };\n\t\t\t\tif (entry?.status.failed || eventError) {\n\t\t\t\t\tthrow new Error(\n\t\t\t\t\t\tentry?.status.exit_code === undefined\n\t\t\t\t\t\t\t? (eventError ?? \"Model failed to load\")\n\t\t\t\t\t\t\t: `Model exited with code ${entry.status.exit_code}`,\n\t\t\t\t\t);\n\t\t\t\t}\n\t\t\t\tawait sleep(250, signal);\n\t\t\t}\n\t\t} finally {\n\t\t\tunlink();\n\t\t\twatcher.abort();\n\t\t}\n\t}\n\n\tasync downloadAndWait(\n\t\tmodel: string,\n\t\tonProgress: (progress: LlamaProgress) => void,\n\t\tsignal?: AbortSignal,\n\t): Promise<LlamaModelInfo[]> {\n\t\tconst watcher = new AbortController();\n\t\tconst unlink = linkSignal(signal, watcher);\n\t\tlet finished = false;\n\t\tlet failure: string | undefined;\n\t\tlet sawDownloading = false;\n\t\tlet polls = 0;\n\t\tvoid this.watch((event) => {\n\t\t\tif (event.model !== model) return;\n\t\t\tif (event.event === \"download_finished\") finished = true;\n\t\t\tif (event.event === \"download_failed\") failure = errorMessage(event.data, \"Download failed\");\n\t\t\tif (event.event === \"download_progress\") {\n\t\t\t\tsawDownloading = true;\n\t\t\t\tconst progress = parseDownloadProgress(event.data);\n\t\t\t\tif (progress) onProgress(progress);\n\t\t\t}\n\t\t}, watcher.signal).catch(() => {});\n\t\ttry {\n\t\t\tawait this.download(model, signal);\n\t\t\tonProgress({ message: \"Downloading model\" });\n\t\t\twhile (true) {\n\t\t\t\tif (signal?.aborted) throw signal.reason ?? new Error(\"Cancelled\");\n\t\t\t\tif (failure) throw new Error(failure);\n\t\t\t\tconst models = await this.list({ signal });\n\t\t\t\tpolls++;\n\t\t\t\tconst entry = models.find((candidate) => candidate.id === model);\n\t\t\t\tif (entry?.status.value === \"downloading\") {\n\t\t\t\t\tsawDownloading = true;\n\t\t\t\t\tconst progress = parseDownloadProgress(entry.status.progress);\n\t\t\t\t\tif (progress) onProgress(progress);\n\t\t\t\t} else if (finished || (entry && (sawDownloading || polls >= 2))) {\n\t\t\t\t\treturn this.list({ reload: true, signal });\n\t\t\t\t}\n\t\t\t\tawait sleep(500, signal);\n\t\t\t}\n\t\t} finally {\n\t\t\tunlink();\n\t\t\twatcher.abort();\n\t\t}\n\t}\n}\n"]}

package/dist/extensions/llama/huggingface.d.ts

@@ -0,0 +1,23 @@
+export interface HuggingFaceModel {
+    id: string;
+    downloads: number;
+}
+export interface HuggingFaceQuantization {
+    name: string;
+    size?: number;
+}
+export interface HuggingFaceModelDetails {
+    id: string;
+    gated: false | "auto" | "manual";
+    quantizations: HuggingFaceQuantization[];
+}
+export declare function findHuggingFaceToken(env?: NodeJS.ProcessEnv): Promise<string | undefined>;
+export declare class HuggingFaceClient {
+    private readonly token;
+    private readonly baseUrl;
+    constructor(token?: string, baseUrl?: string);
+    private request;
+    search(query: string, signal?: AbortSignal): Promise<HuggingFaceModel[]>;
+    details(id: string, signal?: AbortSignal): Promise<HuggingFaceModelDetails>;
+}
+//# sourceMappingURL=huggingface.d.ts.map

package/dist/extensions/llama/huggingface.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"huggingface.d.ts","sourceRoot":"","sources":["../../../src/extensions/llama/huggingface.ts"],"names":[],"mappings":"AASA,MAAM,WAAW,gBAAgB;IAChC,EAAE,EAAE,MAAM,CAAC;IACX,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,uBAAuB;IACvC,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,CAAC,EAAE,MAAM,CAAC;CACd;AAED,MAAM,WAAW,uBAAuB;IACvC,EAAE,EAAE,MAAM,CAAC;IACX,KAAK,EAAE,KAAK,GAAG,MAAM,GAAG,QAAQ,CAAC;IACjC,aAAa,EAAE,uBAAuB,EAAE,CAAC;CACzC;AAsBD,wBAAsB,oBAAoB,CAAC,GAAG,GAAE,MAAM,CAAC,UAAwB,GAAG,OAAO,CAAC,MAAM,GAAG,SAAS,CAAC,CAe5G;AAED,qBAAa,iBAAiB;IAC7B,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAqB;IAC3C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IAEjC,YAAY,KAAK,CAAC,EAAE,MAAM,EAAE,OAAO,SAA2B,EAG7D;YAEa,OAAO;IA4Bf,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,gBAAgB,EAAE,CAAC,CAgB7E;IAEK,OAAO,CAAC,EAAE,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,uBAAuB,CAAC,CAuChF;CACD","sourcesContent":["import { readFile } from \"node:fs/promises\";\nimport { homedir } from \"node:os\";\nimport { join } from \"node:path\";\n\nconst DEFAULT_HUGGING_FACE_URL = \"https://huggingface.co\";\nconst QUANTIZATION_PATTERN =\n\t/(?:^|[-_.])((?:UD-)?(?:IQ\\d(?:_[A-Z0-9]+)+|Q\\d(?:_[A-Z0-9]+)+|BF16|F16|F32|MXFP\\d(?:_[A-Z0-9]+)*))$/iu;\nconst SHARD_SUFFIX_PATTERN = /-\\d{5}-of-\\d{5}$/u;\n\nexport interface HuggingFaceModel {\n\tid: string;\n\tdownloads: number;\n}\n\nexport interface HuggingFaceQuantization {\n\tname: string;\n\tsize?: number;\n}\n\nexport interface HuggingFaceModelDetails {\n\tid: string;\n\tgated: false | \"auto\" | \"manual\";\n\tquantizations: HuggingFaceQuantization[];\n}\n\nfunction payloadError(payload: unknown, fallback: string): string {\n\tif (typeof payload !== \"object\" || payload === null) return fallback;\n\tconst error = (payload as { error?: unknown }).error;\n\treturn typeof error === \"string\" && error ? error : fallback;\n}\n\nfunction parseRateLimitDelay(value: string | null): number | undefined {\n\tconst match = value?.match(/(?:^|;)t=(\\d+)/u);\n\treturn match ? Number(match[1]) : undefined;\n}\n\nasync function readToken(path: string): Promise<string | undefined> {\n\ttry {\n\t\tconst token = (await readFile(path, \"utf8\")).trim();\n\t\treturn token || undefined;\n\t} catch {\n\t\treturn undefined;\n\t}\n}\n\nexport async function findHuggingFaceToken(env: NodeJS.ProcessEnv = process.env): Promise<string | undefined> {\n\tconst fromEnvironment = env.HF_TOKEN?.trim();\n\tif (fromEnvironment) return fromEnvironment;\n\n\tconst paths = [\n\t\tenv.HF_TOKEN_PATH,\n\t\tenv.HF_HOME ? join(env.HF_HOME, \"token\") : undefined,\n\t\tenv.XDG_CACHE_HOME ? join(env.XDG_CACHE_HOME, \"huggingface\", \"token\") : undefined,\n\t\tjoin(homedir(), \".cache\", \"huggingface\", \"token\"),\n\t].filter((path): path is string => Boolean(path));\n\tfor (const path of new Set(paths)) {\n\t\tconst token = await readToken(path);\n\t\tif (token) return token;\n\t}\n\treturn undefined;\n}\n\nexport class HuggingFaceClient {\n\tprivate readonly token: string | undefined;\n\tprivate readonly baseUrl: string;\n\n\tconstructor(token?: string, baseUrl = DEFAULT_HUGGING_FACE_URL) {\n\t\tthis.token = token;\n\t\tthis.baseUrl = baseUrl.replace(/\\/+$/u, \"\");\n\t}\n\n\tprivate async request(path: string, signal?: AbortSignal): Promise<unknown> {\n\t\tconst headers = new Headers();\n\t\tif (this.token) headers.set(\"Authorization\", `Bearer ${this.token}`);\n\t\tconst timeout = AbortSignal.timeout(15_000);\n\t\tconst response = await fetch(`${this.baseUrl}${path}`, {\n\t\t\theaders,\n\t\t\tsignal: signal ? AbortSignal.any([signal, timeout]) : timeout,\n\t\t});\n\t\tlet payload: unknown;\n\t\ttry {\n\t\t\tpayload = await response.json();\n\t\t} catch {\n\t\t\tpayload = undefined;\n\t\t}\n\t\tif (!response.ok) {\n\t\t\tconst fallback = `Hugging Face returned HTTP ${response.status}`;\n\t\t\tif (response.status === 429) {\n\t\t\t\tconst delay =\n\t\t\t\t\tNumber(response.headers.get(\"retry-after\")) || parseRateLimitDelay(response.headers.get(\"ratelimit\"));\n\t\t\t\tthrow new Error(\n\t\t\t\t\tdelay ? `Hugging Face rate limit reached; retry in ${delay}s` : \"Hugging Face rate limit reached\",\n\t\t\t\t);\n\t\t\t}\n\t\t\tthrow new Error(payloadError(payload, fallback));\n\t\t}\n\t\treturn payload;\n\t}\n\n\tasync search(query: string, signal?: AbortSignal): Promise<HuggingFaceModel[]> {\n\t\tconst params = new URLSearchParams({\n\t\t\tsearch: query,\n\t\t\tfilter: \"gguf\",\n\t\t\tsort: \"downloads\",\n\t\t\tdirection: \"-1\",\n\t\t\tlimit: \"20\",\n\t\t});\n\t\tconst payload = await this.request(`/api/models?${params}`, signal);\n\t\tif (!Array.isArray(payload)) throw new Error(\"Hugging Face returned invalid search results\");\n\t\treturn payload.flatMap((value) => {\n\t\t\tif (typeof value !== \"object\" || value === null || typeof (value as { id?: unknown }).id !== \"string\")\n\t\t\t\treturn [];\n\t\t\tconst model = value as { id: string; downloads?: unknown };\n\t\t\treturn [{ id: model.id, downloads: typeof model.downloads === \"number\" ? model.downloads : 0 }];\n\t\t});\n\t}\n\n\tasync details(id: string, signal?: AbortSignal): Promise<HuggingFaceModelDetails> {\n\t\tconst encodedId = id.split(\"/\").map(encodeURIComponent).join(\"/\");\n\t\tconst payload = await this.request(`/api/models/${encodedId}?blobs=true`, signal);\n\t\tif (typeof payload !== \"object\" || payload === null) {\n\t\t\tthrow new Error(\"Hugging Face returned invalid model details\");\n\t\t}\n\t\tconst model = payload as { id?: unknown; gated?: unknown; siblings?: unknown };\n\t\tconst sizes = new Map<string, { total: number; complete: boolean }>();\n\t\tif (Array.isArray(model.siblings)) {\n\t\t\tfor (const value of model.siblings) {\n\t\t\t\tif (typeof value !== \"object\" || value === null) continue;\n\t\t\t\tconst file = value as { rfilename?: unknown; size?: unknown };\n\t\t\t\tif (typeof file.rfilename !== \"string\" || !file.rfilename.toLowerCase().endsWith(\".gguf\")) continue;\n\t\t\t\tconst filename = file.rfilename.split(\"/\").at(-1)!;\n\t\t\t\tif (filename.toLowerCase().startsWith(\"mmproj\")) continue;\n\t\t\t\tconst stem = filename.slice(0, -5).replace(SHARD_SUFFIX_PATTERN, \"\");\n\t\t\t\tconst quantization = stem.match(QUANTIZATION_PATTERN)?.[1]?.toUpperCase();\n\t\t\t\tif (!quantization) continue;\n\t\t\t\tconst current = sizes.get(quantization) ?? { total: 0, complete: true };\n\t\t\t\tif (typeof file.size === \"number\") current.total += file.size;\n\t\t\t\telse current.complete = false;\n\t\t\t\tsizes.set(quantization, current);\n\t\t\t}\n\t\t}\n\t\tconst quantizations = [...sizes]\n\t\t\t.map(([name, size]) => ({ name, size: size.complete ? size.total : undefined }))\n\t\t\t.sort((left, right) => {\n\t\t\t\tif (left.name === \"Q4_K_M\") return -1;\n\t\t\t\tif (right.name === \"Q4_K_M\") return 1;\n\t\t\t\treturn (\n\t\t\t\t\t(left.size ?? Number.MAX_SAFE_INTEGER) - (right.size ?? Number.MAX_SAFE_INTEGER) ||\n\t\t\t\t\tleft.name.localeCompare(right.name)\n\t\t\t\t);\n\t\t\t});\n\t\treturn {\n\t\t\tid: typeof model.id === \"string\" ? model.id : id,\n\t\t\tgated: model.gated === \"auto\" || model.gated === \"manual\" ? model.gated : false,\n\t\t\tquantizations,\n\t\t};\n\t}\n}\n"]}