@qvac/llm-llamacpp 0.59.0 → 0.60.0

package/README.md

@@ -198,8 +198,8 @@ const config = {
 | split-mode        | `"none"`, `"layer"`, or `"tensor"` | `"none"`                     | How to split the model across GPUs. `"tensor"` is EXPERIMENTAL and desktop-only; `"row"` is rejected ([details](./docs/multi-gpu.md)) |
 | tensor-split      | comma-separated proportions (e.g. `"1,1"`)  | —                            | GPU split ratios for the multi-GPU split modes ([details](./docs/multi-gpu.md)) |
 | parallel          | integer                                     | 1                            | Concurrent sequence slots for continuous batching. Values `>= 2` enable batch `run()` and split the KV cache uniformly across slots ([details](./docs/continuous-batching.md)) |
-| cache_checkpoints | 0 – 1024                                    | 1                            | Per-sequence cap on process-local checkpoints kept for cached requests on hybrid / recurrent models and DeepSeek V4, one per committed request at the end of its history. Each holds only the state a tail trim cannot rebuild (about 20 MB on Qwen3.5-0.8B). `1` serves the next turn and a regenerate; editing the user message k-th from the end (1 = the last) without reprocessing the whole conversation needs `k + 1`, so `2` for the last one; `0` disables them. Also accepted as `cache-checkpoints` ([details](./docs/cache-api.md), [how it works](./docs/cache-lifecycle.md)) |
-| cache_checkpoints_max_bytes | integer (bytes)                 | 0 (unlimited)                | Byte budget for those checkpoints, enforced before the count. The load fails with `InvalidArgument` when it cannot hold `cache_checkpoints` checkpoints of the largest size the context allows. Also accepted as `cache-checkpoints-max-bytes` |
+| cache_checkpoints | 0 – 1024                                    | 1                            | Per-sequence cap on process-local checkpoints kept for cached requests on hybrid / recurrent models, DeepSeek V4 and sliding-window models (Gemma 3/4, gpt-oss), one per committed request at the end of its history. Each holds only the state a tail trim cannot rebuild (about 20 MB on Qwen3.5-0.8B; on sliding-window models the window cells, which reach hundreds of MB on larger models). `1` serves the next turn and a regenerate; editing the user message k-th from the end (1 = the last) without reprocessing the whole conversation needs `k + 1`, so `2` for the last one; `0` disables them. The newest one is also saved in the `cacheKey` file, so a conversation loaded in a new process continues its next turn from it. Also accepted as `cache-checkpoints` ([details](./docs/cache-api.md), [how it works](./docs/cache-lifecycle.md)) |
+| cache_checkpoints_max_bytes | integer (bytes)                 | 0 (unlimited)                | Byte budget for those checkpoints, enforced before the count. The load fails with `InvalidArgument` when it cannot hold `cache_checkpoints` checkpoints of the largest size the context allows (on a sliding-window model, a full window). Also accepted as `cache-checkpoints-max-bytes` |
 | cache_checkpoint_storage | `"memory"` or `"disk"`             | `"memory"`                   | Where checkpoints and the per-request rollback snapshot live. `"memory"` keeps them in host RAM so a cached chat never touches the disk; `"disk"` writes them to a private directory inside `cache_checkpoint_dir`, which it requires. Also accepted as `cache-checkpoint-storage` |
 | cache_checkpoint_dir | directory path                              | none                         | **Required with `cache_checkpoint_storage: "disk"`**, and refused without it: where checkpoint files go, inside a private directory the addon creates in it (mode 0700; on Windows, the directory itself). On mobile, pass the app's cache directory. The load fails with `InvalidArgument` when it is missing, empty, or unusable. Also accepted as `cache-checkpoint-dir` |
 | cache_ram_mib | 0 – 1048576 (MiB)                               | 0 (off)                      | Host-RAM budget for conversations that are not running. A `cacheKey` switch or a batch slot eviction moves the conversation there instead of writing its file, and the next request on that key restores it. A write-back cache: unsaved turns reach the file on `saveCache()`, when the budget pushes the entry out, or at unload; ephemeral entries are dropped instead. Also accepted as `cache-ram-mib` ([details](./docs/cache-api.md#keep-switched-away-conversations-in-ram)) |

package/addon.d.ts

@@ -90,7 +90,6 @@ export type MappedAddonEvent = {
  * `FinetuneProgress`, or `null` to drop it.
  */
 export declare function mapAddonEvent(rawEvent: unknown, rawData: unknown, rawError: unknown): MappedAddonEvent | null;
-export declare function resolveBackendsDir(): string;
 /**
  * An interface between Bare addon in C++ and JS runtime.
  */

package/addon.js

@@ -2,11 +2,6 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.LlamaInterface = void 0;
 exports.mapAddonEvent = mapAddonEvent;
-exports.resolveBackendsDir = resolveBackendsDir;
-/* eslint-disable @typescript-eslint/no-require-imports -- Bare modules expose CommonJS export shapes. */
-const fs = require("bare-fs");
-const path = require("bare-path");
-const fabricBackends = require("@qvac/fabric/backends");
 // Index-matched to the C++ GenerationStopReason enum (SequenceDriver.hpp).
 const STOP_REASONS = [
     "none",
@@ -72,22 +67,6 @@ function mapAddonEvent(rawEvent, rawData, rawError) {
     }
     return { type: type, data: rawData, error: rawError };
 }
-// The ggml compute backends ship next to the @qvac/fabric runtime
-// (<root>/<host>/qvac__fabric). We deliberately do not copy them into this
-// addon to avoid duplicating tens of MB per fabric consumer. On desktop,
-// @qvac/fabric/backends resolves that root in whichever package holds the
-// runtime. On mobile the package tree isn't resolvable at runtime (the worklet
-// runs from a packed bundle), so fall back to this addon's own prebuilds, where
-// the mobile packaging stages the backends. The native side appends
-// BACKENDS_SUBDIR ("<host>/qvac__fabric") to whichever root we return.
-function resolveBackendsDir() {
-    // fabric's resolver only checks that the platform package resolves, not that
-    // its prebuilds are on disk.
-    const fabricRoot = fabricBackends.resolveBackendsDir();
-    if (fabricRoot !== null && fs.existsSync(fabricRoot))
-        return fabricRoot;
-    return path.join(__dirname, "prebuilds");
-}
 /**
  * An interface between Bare addon in C++ and JS runtime.
  */
@@ -99,9 +78,6 @@ class LlamaInterface {
         if (!configurationParams.config) {
             configurationParams.config = {};
         }
-        if (!configurationParams.config.backendsDir) {
-            configurationParams.config.backendsDir = resolveBackendsDir();
-        }
         this._handle = this._binding.createInstance(this, configurationParams, outputCb, null);
     }
     loadWeights(weightsData) {
@@ -208,5 +184,4 @@ exports.LlamaInterface = LlamaInterface;
 module.exports = {
     LlamaInterface,
     mapAddonEvent,
-    resolveBackendsDir,
 };

package/binding.js

@@ -1,6 +1,6 @@
 // Pre-load @qvac/fabric so its shared .bare module (the llama.cpp + ggml
 // runtime) is registered with the bare runtime before our addon resolves its
-// DT_NEEDED dependency qvac__fabric@0.bare.
+// DT_NEEDED dependency qvac__fabric-<host>@0.bare (shipped by @qvac/fabric-<host>).
 require('@qvac/fabric')
 
 module.exports = require.addon()

package/fit.js

@@ -1,7 +1,6 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.assessFit = assessFit;
-const addon_1 = require("./addon");
 /**
  * Projects one model against the memory free right now, reading GGUF metadata
  * and never weight data. The registry's weightless copy of a model answers the
@@ -22,9 +21,5 @@ const addon_1 = require("./addon");
 function assessFit(request) {
     // eslint-disable-next-line @typescript-eslint/no-require-imports -- native binding is resolved lazily from package prebuilds.
     const binding = require('./binding.js');
-    const config = { ...request.config };
-    if (config['backendsDir'] === undefined || config['backendsDir'] === '') {
-        config['backendsDir'] = (0, addon_1.resolveBackendsDir)();
-    }
-    return binding.assessFit({ ...request, config });
+    return binding.assessFit({ ...request, config: { ...request.config } });
 }

package/index.d.ts

@@ -300,7 +300,10 @@ declare namespace LlmLlamacpp {
          * conversation needs at least k + 1; with the default 1, editing the last
          * user message reprocesses everything. `0` keeps and takes none (every
          * divergent turn is a cold prefill), the maximum is 1024.
-         * Ignored on pure-attention models, which never take checkpoints.
+         * Sliding-window models (Gemma 3/4, gpt-oss, without `swa_full`) take
+         * them too: their checkpoints hold the window cells, restored when a
+         * turn diverges behind the window. Ignored on other pure-attention
+         * models, which never take checkpoints.
          * Also accepted as `cache-checkpoints`; supplying both is an error.
          */
         cache_checkpoints?: NumericLike;
@@ -309,8 +312,9 @@ declare namespace LlmLlamacpp {
          * `cache_checkpoints`: the oldest checkpoints are dropped until the total
          * payload fits. `0` (default) is unlimited. When set, the model load fails
          * with `InvalidArgument` if the budget cannot hold `cache_checkpoints`
-         * checkpoints of the largest size the context allows, so a misconfigured
-         * budget is reported up front rather than discovered mid-conversation.
+         * checkpoints of the largest size the context allows (on a sliding-window
+         * model, a full window), so a misconfigured budget is reported up front
+         * rather than discovered mid-conversation.
          * Also accepted as `cache-checkpoints-max-bytes`.
          */
         cache_checkpoints_max_bytes?: NumericLike;
@@ -320,7 +324,9 @@ declare namespace LlmLlamacpp {
          * touches the disk, `'disk'` writes them to files in
          * `cache_checkpoint_dir`, which it requires. Each snapshot holds only the
          * state a tail trim cannot rebuild, a size fixed by the model (about 20 MB
-         * on Qwen3.5-0.8B). Also accepted as `cache-checkpoint-storage`.
+         * on Qwen3.5-0.8B; on sliding-window models the window cells, which reach
+         * hundreds of MB on larger ones). Also accepted as
+         * `cache-checkpoint-storage`.
          */
         cache_checkpoint_storage?: "disk" | "memory";
         /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@qvac/llm-llamacpp",
-  "version": "0.59.0",
+  "version": "0.60.0",
   "description": "llama addon for qvac",
   "addon": true,
   "scripts": {
@@ -73,15 +73,15 @@
   "bugs": "https://github.com/tetherto/qvac/issues",
   "homepage": "https://qvac.tether.io",
   "devDependencies": {
-    "@qvac/fabric-android-arm64": "^0.20.0",
-    "@qvac/fabric-ios": "^0.20.0",
+    "@qvac/fabric-android-arm64": "0.21.1",
+    "@qvac/fabric-ios": "0.21.1",
     "@types/node": "^24.2.1",
     "bare-crypto": "^1.15.3",
     "bare-process": "^4.2.2",
     "bare-subprocess": "^6.1.0",
     "bare-url": "^2.1.6",
     "brittle": "^3.16.5",
-    "cmake-bare": "^1.7.5",
+    "cmake-bare": "^1.9.0",
     "cmake-vcpkg": "^1.1.0",
     "eslint": "^9.39.4",
     "eslint-import-resolver-typescript": "^4.4.5",
@@ -96,16 +96,34 @@
     "typescript-eslint": "^8.63.0"
   },
   "dependencies": {
-    "@qvac/fabric": "^0.20.2",
+    "@qvac/fabric": "^0.21.1",
     "@qvac/infer-base": "^0.6.2",
     "@qvac/logging": "^0.1.0",
     "bare-fs": "^4.5.1",
     "bare-path": "^3.0.0"
   },
+  "optionalDependencies": {
+    "@qvac/fabric-linux-x64": "^0.21.1",
+    "@qvac/fabric-linux-arm64": "^0.21.1",
+    "@qvac/fabric-darwin-arm64": "^0.21.1",
+    "@qvac/fabric-darwin-x64": "^0.21.1",
+    "@qvac/fabric-win32-x64": "^0.21.1"
+  },
+  "peerDependencies": {
+    "@qvac/fabric-android-arm64": "^0.21.1",
+    "@qvac/fabric-ios": "^0.21.1"
+  },
+  "peerDependenciesMeta": {
+    "@qvac/fabric-android-arm64": {
+      "optional": true
+    },
+    "@qvac/fabric-ios": {
+      "optional": true
+    }
+  },
   "engines": {
     "bare": ">=1.24.0"
   },
-  "peerDependencies": {},
   "exports": {
     "./package": "./package.json",
     ".": {

package/prebuilds/darwin-arm64/qvac__llm-llamacpp.bare.exports

@@ -7,7 +7,7 @@
     ],
     "current_versions": [
       {
-        "version": "0.59"
+        "version": "0.60"
       }
     ],
     "exported_symbols": [
@@ -61,10 +61,12 @@
             "__ZGVN28qvac_lib_inference_addon_cpp6logger8JsLogger12admin_mutex_E",
             "__ZN28qvac_lib_inference_addon_cpp6logger8JsLogger16async_initiated_E",
             "__ZNSt3__19__any_imp17__unique_typeinfoIN28qvac_lib_inference_addon_cpp6Output6LogMsgEE4__idE",
+            "__ZN30qvac_lib_inference_addon_llama5cache24CHECKPOINT_SECTION_MAGICE",
             "__ZNSt3__19__any_imp17__unique_typeinfoI22FinetuneTerminalResultE4__idE",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface13instancesMtx_E",
             "__ZN30qvac_lib_inference_addon_llama22BATCH_OUTPUT_TYPE_NAMEE",
             "__ZN28qvac_lib_inference_addon_cpp6logger8JsLogger10log_queue_E",
+            "__ZN30qvac_lib_inference_addon_llama5cache26CHECKPOINT_SECTION_VERSIONE",
             "__ZNSt3__19__any_imp17__unique_typeinfoINS_6vectorIN10LlamaModel6PromptENS_9allocatorIS4_EEEEE4__idE",
             "__ZNSt3__16ranges5__cpo4findE",
             "__ZN30qvac_lib_inference_addon_llama10JsBatchIds11kMintPrefixE",
@@ -87,8 +89,8 @@
             "__ZN30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher10addRequestEONSt3__16vectorIiNS2_9allocatorIiEEEERj",
             "__ZN30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher19sampleAndAppendIdleERKNSt3__18functionIFijiEEE",
             "__ZN10LlamaModel30discardFinetuneCancelSaveModesERKNSt3__16vectorIyNS0_9allocatorIyEEEE",
-            "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler29writeStateToFileOrThrowLockedEjRKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEERKNS2_6vectorIiNS6_IiEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching7RequestC1EjO11PrefillPlanji",
+            "__ZN30qvac_lib_inference_addon_llama5utils25partialSnapshotFromStreamERNSt3__113basic_istreamIcNS1_11char_traitsIcEEEEyiNS0_15SnapshotStorageERKNS1_12basic_stringIcS4_NS1_9allocatorIcEEEERKNS1_8functionIFvPKhmEEERNS0_21SequenceStateSnapshotE",
             "__ZNK13ModelMetaData10isU32OneOfEPKcSt16initializer_listIjE",
             "__ZN10LlamaModelD0Ev",
             "__ZN24llama_finetuning_helpers18schedulerLrForStepERKNS_20LoraLrSchedulerStateEx",
@@ -96,20 +98,22 @@
             "__ZN12CacheManager28persistedBackingStoreMissingERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZThn48_N14TextLlmContext13onLogitsReadyEijRKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEEP10LlamaBatch",
             "__ZN14MtmdLlmContext23restoreCacheStateTokensERKNSt3__16vectorIiNS0_9allocatorIiEEEE",
-            "__ZN14TextLlmContext19acceptRestoredStateERKNSt3__16vectorIiNS0_9allocatorIiEEEERKNS0_12basic_stringIcNS0_11char_traitsIcEENS2_IcEEEE",
+            "__ZN14MtmdLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEb",
             "__ZN12CacheManager28saveActiveCacheForTransitionEv",
             "__ZN14MtmdLlmContext15advanceTextSpanEi",
             "__ZN14MtmdLlmContext27discardPendingResidentTokenEv",
             "__ZN14TextLlmContext4stopEv",
-            "__ZN30qvac_lib_inference_addon_llama5utils32configureTemplateDerivedSamplingER13common_paramsRKNSt3__18functionIFNS3_6vectorIiNS3_9allocatorIiEEEERKNS3_12basic_stringIcNS3_11char_traitsIcEENS6_IcEEEEEEERKNS0_18PromptRenderResultEbRKNS3_8optionalINS0_13ReasoningTagsEEE",
+            "__ZN14TextLlmContext19acceptRestoredStateERKNSt3__16vectorIiNS0_9allocatorIiEEEERKNS0_12basic_stringIcNS0_11char_traitsIcEENS2_IcEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler26accumulateSlotRuntimeStatsERKNS1_9SlotStateERKNS0_7RequestE",
             "__ZN17backend_selection20splitModeDeviceNamesERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN30qvac_lib_inference_addon_llama5utils21SequenceStateSnapshotD1Ev",
             "__ZThn48_N10LlamaModelD0Ev",
-            "__ZThn16_NK10LlamaModel6cancelEv",
+            "__ZN30qvac_lib_inference_addon_llama5utils32configureTemplateDerivedSamplingER13common_paramsRKNSt3__18functionIFNS3_6vectorIiNS3_9allocatorIiEEEERKNS3_12basic_stringIcNS3_11char_traitsIcEENS6_IcEEEEEEERKNS0_18PromptRenderResultEbRKNS3_8optionalINS0_13ReasoningTagsEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler11clearLockedEv",
-            "__ZThn48_N14TextLlmContext24snapshotPreRequestCursorEv",
+            "__ZN30qvac_lib_inference_addon_llama5utils21streamPartialSnapshotERKNS0_21SequenceStateSnapshotERKNSt3__18functionIFvPKhmEEE",
+            "__ZThn16_NK10LlamaModel6cancelEv",
             "__ZN17backend_selection30preferredBackendTypeFromStringERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
+            "__ZThn48_N14TextLlmContext24snapshotPreRequestCursorEv",
             "__ZN10LlamaModel9saveCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZNK10LlamaModel15consumeJobStatsEy",
             "__ZN10LlamaModel10getContextEv",
@@ -142,6 +146,7 @@
             "__ZN30qvac_lib_inference_addon_llama8batching7Request11isOptActiveERKNSt3__18optionalIS1_EE",
             "__ZThn48_N14TextLlmContext12syncPositionEi",
             "__ZNK30qvac_lib_inference_addon_llama8batching20RuntimeStatsSnapshot16avgConcurrentSeqEv",
+            "__ZThn48_NK14TextLlmContext28appendCheckpointsToCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZThn8_N10LlamaModel17setWeightsForFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEONS0_10unique_ptrINS0_15basic_streambufIcS3_EENS0_14default_deleteISB_EEEE",
             "__ZN14MtmdLlmContext24holdPendingResidentTokenEii",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler16discardKeyLockedERKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEE",
@@ -159,15 +164,16 @@
             "__ZN30qvac_lib_inference_addon_llama5utils16isGemma4BasenameENSt3__117basic_string_viewIcNS1_11char_traitsIcEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler22storeCheckpointsLockedERKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEONS2_5dequeINS_5cache10CheckpointENS6_ISD_EEEE",
             "__ZN14MtmdLlmContext24captureHistoryCheckpointEi",
-            "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler16cancelSlotLockedEjNS1_15SaveCachePolicyE",
+            "__ZN30qvac_lib_inference_addon_llama5cache23appendCheckpointSectionERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEERKNS1_5dequeINS0_10CheckpointENS5_ISB_EEEE",
             "__ZN10LlamaModel4initEb",
             "__ZNK10LlamaModel21requestFinetuneCancelEy",
+            "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler16cancelSlotLockedEjNS1_15SaveCachePolicyE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler22clearUnparkedSequencesEv",
-            "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler12runtimeStatsEv",
             "__ZN10LlamaModel17processPromptImplERKNS_6PromptE",
             "__Z30applyGenerationParamsToContextR13common_paramsRNSt3__110unique_ptrI14common_sampler20CommonSamplerDeleterEEP11llama_modelRK16GenerationParams",
             "__ZN30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher5clearERKNSt3__18functionIFvjEEE",
             "__ZNK30qvac_lib_inference_addon_llama8batching7Request13exceededLimitEv",
+            "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler12runtimeStatsEv",
             "__ZN14MtmdLlmContext10resetStateEb",
             "__ZN14TextLlmContext27confirmPendingResidentTokenEi",
             "__ZN14LlamaFinetuner23initializeCheckpointingERKN30qvac_lib_inference_addon_llama21LlamaFinetuningParamsEP18llama_adapter_loraPN24llama_finetuning_helpers20LoraLrSchedulerStateE",
@@ -184,7 +190,6 @@
             "__ZNK10LlamaModel12runtimeStatsEv",
             "__ZN14MtmdLlmContext10resetMediaEv",
             "__ZN30qvac_lib_inference_addon_llama8batching22finalizeTerminalDriverER14SequenceDriverNS0_10StopReasonEbRKNSt3__18functionIFvRKNS4_12basic_stringIcNS4_11char_traitsIcEENS4_9allocatorIcEEEEEEE",
-            "__ZN14TextLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN30qvac_lib_inference_addon_llama5utils21SequenceStateSnapshot10adoptEmptyEi",
             "__ZN22load_fit_normalization36canonicalizeCpuTensorBufferOverridesERNSt3__16vectorI32llama_model_tensor_buft_overrideNS0_9allocatorIS2_EEEEP24ggml_backend_buffer_type",
             "__ZN24llama_finetuning_helpers16parseLoraModulesERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
@@ -223,24 +228,24 @@
             "__ZNK30qvac_lib_inference_addon_llama8batching7Request21chunkConsumesAllUnfedEj",
             "__ZN10LlamaModel7processERKNSt3__13anyE",
             "__ZThn48_N14MtmdLlmContext8onCancelERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
+            "__ZN14MtmdLlmContext29adoptCheckpointsFromCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEy",
             "__ZN10LlamaModel32setFinetuneCancelSavesCheckpointEbRKNSt3__16vectorIyNS0_9allocatorIyEEEE",
             "__ZN13ModelMetaData5parseERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEERK10GGUFShardsbPKc",
             "__ZNK10LlamaModel15cancelInferenceEv",
             "__ZN14LlamaFinetuner17clearPauseRequestEv",
+            "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler29writeStateToFileOrThrowLockedEjRKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEERKNS2_6vectorIiNS6_IiEEEERKNS2_5dequeINS_5cache10CheckpointENS6_ISI_EEEE",
             "__ZN14LlamaFinetuner12requestPauseEb",
             "__ZN10LlamaModel17resolveShardPathsER10GGUFShardsRKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEE",
             "__ZN12CacheManager13saveForCallerERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN14TextLlmContext18commitCacheRequestEv",
             "__ZN14TextLlmContext9onFailureERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
             "__ZN17backend_selection26getEffectiveGpuDeviceCountERKNS_16BackendInterfaceE",
-            "__ZN30qvac_lib_inference_addon_llama5utils25getThinkingForcedOpenTextERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEES9_",
             "__ZN24llama_finetuning_helpers19savePauseCheckpointEP16ggml_opt_contextRNS_23TrainingCheckpointStateEbx",
             "__ZN14MtmdLlmContext14setCacheTokensEi",
             "__ZN14TextLlmContext15checkAntipromptEv",
-            "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler22writeStateToFileLockedEjRKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEERKNS2_6vectorIiNS6_IiEEEE",
+            "__ZN30qvac_lib_inference_addon_llama5utils25getThinkingForcedOpenTextERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEES9_",
             "__ZN30qvac_lib_inference_addon_llama5utils21SequenceStateSnapshotC1EOS1_",
             "__ZN24llama_finetuning_helpers25setCurrentCheckpointStateEPNS_23TrainingCheckpointStateE",
-            "__ZThn48_N14TextLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN10LlamaModel17processConcurrentERKNS_6PromptEy",
             "__ZN14MtmdLlmContext19appendResidentTokenEi",
             "__ZN24llama_finetuning_helpers12readTextFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
@@ -254,9 +259,10 @@
             "__ZN14MtmdLlmContext8onCancelERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
             "__ZN14MtmdLlmContext16evalMediaSegmentEmi",
             "__ZNK30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher14firstFreeSeqIdEv",
-            "__ZN30qvac_lib_inference_addon_llama5utils26generationPromptTailLengthEP13llama_contextRKNSt3__112basic_stringIcNS3_11char_traitsIcEENS3_9allocatorIcEEEERKNS3_6vectorIiNS7_IiEEEE",
+            "__ZN30qvac_lib_inference_addon_llama5cache21readCheckpointSectionERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEEyRKNS0_6LedgerERKNS0_16CheckpointPolicyE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler17cancelGroupQueuedEy",
             "__ZN30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher15markAllFinishedENS0_10StopReasonE",
+            "__ZN30qvac_lib_inference_addon_llama5utils26generationPromptTailLengthEP13llama_contextRKNSt3__112basic_stringIcNS3_11char_traitsIcEENS3_9allocatorIcEEEERKNS3_6vectorIiNS7_IiEEEE",
             "__ZN30qvac_lib_inference_addon_llama5utils34isQwen3ReasoningFamilyArchitectureENSt3__117basic_string_viewIcNS1_11char_traitsIcEEEE",
             "__ZNK14TextLlmContext10ctxCeilingEv",
             "__ZN10LlamaModel12formatPromptERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
@@ -284,6 +290,7 @@
             "__ZN14TextLlmContext21initializeCommonStateEv",
             "__ZNK14LlamaFinetuner31getCurrentCheckpointStateSharedEv",
             "__ZThn16_N10LlamaModelD1Ev",
+            "__ZThn48_N14MtmdLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEb",
             "__ZThn48_N14MtmdLlmContext32snapshotPreRequestRollbackAnchorEv",
             "__ZN10LlamaModelC1EONSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEES7_ONS0_13unordered_mapIS6_S6_NS0_4hashIS6_EENS0_8equal_toIS6_EENS4_INS0_4pairIKS6_S6_EEEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler25finalizeFinishedSequencesEv",
@@ -291,12 +298,12 @@
             "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler7hasWorkEv",
             "__ZN14TextLlmContext22rollbackCurrentRequestERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
             "__ZN14TextLlmContext27discardPendingResidentTokenEv",
+            "__ZThn48_N14TextLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEb",
             "__ZN10LlamaModel13createContextEONSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEER13common_paramsNS0_10unique_ptrI18common_init_resultNS0_14default_deleteISB_EEEE",
             "__ZN14TextLlmContext20onGenerationFinishedERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE20GenerationStopReason",
             "__Z23validateByteBufferCountRKNSt3__16vectorI12PlannedMediaNS_9allocatorIS1_EEEEm",
             "__ZN17backend_selection23getSplitDeviceSelectionEv",
             "__ZN19LlamaBackendsHandleC1EOS_",
-            "__ZThn48_N14MtmdLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN12CacheManager11handleCacheER19ParsedPromptPayloadRKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEENS2_8functionIFS0_SA_EEESA_b",
             "__ZN30qvac_lib_inference_addon_llama5utils21snapshotSequenceStateEP13llama_contextiiRNS0_21SequenceStateSnapshotENS0_15SnapshotStorageENS0_13SnapshotScopeERKNSt3__112basic_stringIcNS7_11char_traitsIcEENS7_9allocatorIcEEEE",
             "__ZNK10LlamaModel16supportsBatchingEv",
@@ -306,12 +313,13 @@
             "__ZN14MtmdLlmContext17onPrefillCompleteEim",
             "__ZN17backend_selection25tryBackendOverrideFromMapERNSt3__113unordered_mapINS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEES7_NS0_4hashIS7_EENS0_8equal_toIS7_EENS5_INS0_4pairIKS7_S7_EEEEEE",
             "__ZN18AsyncWeightsLoader31lendFirstShardAndWaitForReleaseEONSt3__110unique_ptrINS0_15basic_streambufIcNS0_11char_traitsIcEEEENS0_14default_deleteIS5_EEEE",
+            "__ZN14TextLlmContext29adoptCheckpointsFromCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEy",
             "__ZN22load_fit_normalization17tuneLoadConfigMapERNSt3__113unordered_mapINS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEES7_NS0_4hashIS7_EENS0_8equal_toIS7_EENS5_INS0_4pairIKS7_S7_EEEEEERK13ModelMetaDataRKNS0_8optionalIiEERK23FinetuneConfigOverridesbbbbb",
-            "__ZNK14MtmdLlmContext10ctxCeilingEv",
             "__ZN30qvac_lib_inference_addon_llama8batching7Request22isOptGenerationPendingERKNSt3__18optionalIS1_EE",
-            "__ZNK30qvac_lib_inference_addon_llama8batching7Request16isPrefillPendingEv",
+            "__ZNK14MtmdLlmContext10ctxCeilingEv",
             "__ZThn48_N14TextLlmContext18adoptResidentStateERKNSt3__16vectorIiNS0_9allocatorIiEEEE",
             "__ZThn48_N14MtmdLlmContext13onSequenceEndERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
+            "__ZNK30qvac_lib_inference_addon_llama8batching7Request16isPrefillPendingEv",
             "__ZThn32_NK10LlamaModel10cancelByIdEy",
             "__ZN30qvac_lib_inference_addon_llama8batching20RuntimeStatsSnapshot14accumulateSlotExxRKNS0_7RequestE",
             "__ZNK14MtmdLlmContext25getToolDefinitionsDroppedEv",
@@ -338,6 +346,8 @@
             "__ZN14TextLlmContext23restoreCacheStateTokensERKNSt3__16vectorIiNS0_9allocatorIiEEEE",
             "__ZN18AsyncWeightsLoader20joinFirstShardWorkerEv",
             "__ZN30qvac_lib_inference_addon_llama8batching24generationBudgetExceededEjjij",
+            "__ZNK14MtmdLlmContext28appendCheckpointsToCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
+            "__ZThn48_N14TextLlmContext29adoptCheckpointsFromCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEy",
             "__ZN14TextLlmContext17beginCacheRequestEv",
             "__ZN10LlamaModel19resolveChatAndToolsERKNS_6PromptE",
             "__ZThn48_NK14MtmdLlmContext8getNPastEv",
@@ -351,8 +361,8 @@
             "__ZNK14MtmdLlmContext8getNPastEv",
             "__ZN14MtmdLlmContext18commitCacheRequestEv",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler28applyGroupQueuedCancelLockedEy",
+            "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler22writeStateToFileLockedEjRKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEERKNS2_6vectorIiNS6_IiEEEERKNS2_5dequeINS_5cache10CheckpointENS6_ISI_EEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler23recordPendingSlotCancelEjy",
-            "__ZNK14LlamaFinetuner17isFinetuneRunningEv",
             "__ZN30qvac_lib_inference_addon_llama5utils34selectReasoningTagsForArchitectureERKNSt3__18optionalINS1_12basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEEEE",
             "__ZN10LlamaModel18processPromptBatchERKNSt3__16vectorINS_6PromptENS0_9allocatorIS2_EEEE",
             "__ZN14LlamaFinetuner31setCurrentCheckpointStateSharedENSt3__110shared_ptrIN24llama_finetuning_helpers23TrainingCheckpointStateEEE",
@@ -366,16 +376,17 @@
             "__ZN30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher9fillBatchER10LlamaBatch",
             "__ZN14MtmdLlmContext16generateResponseERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler14failSlotLockedEjSt13exception_ptr",
+            "__ZNK14LlamaFinetuner17isFinetuneRunningEv",
             "__ZN14MtmdLlmContextC2ER13common_paramsNSt3__110unique_ptrI18common_init_resultNS2_14default_deleteIS4_EEEE",
             "__ZN30qvac_lib_inference_addon_llama5utils21updateReasoningBufferERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEERNS0_14ReasoningStateE",
             "__ZThn24_N10LlamaModelD1Ev",
             "__ZN17backend_selection23applyAdrenoRestrictionsERNS_20SplitDeviceSelectionERK13ModelMetaDatab",
             "__ZThn48_N14MtmdLlmContext20onGenerationFinishedERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE20GenerationStopReason",
             "__ZN14MtmdLlmContext13resetStopFlagEv",
-            "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler12parkedSeqIdsEv",
             "__ZN30qvac_lib_inference_addon_llama5utils24requireSnapshotDirectoryERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEE",
             "__ZN24llama_finetuning_helpers24pauseCheckpointDirectoryERKNSt3__14__fs10filesystem4pathEx",
             "__Z21computeMediaLoadOrderRKNSt3__16vectorI12PlannedMediaNS_9allocatorIS1_EEEE",
+            "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler12parkedSeqIdsEv",
             "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler9occupancyEv",
             "__ZThn48_N14MtmdLlmContext24captureHistoryCheckpointEi",
             "__ZThn32_N10LlamaModelD1Ev",
@@ -453,6 +464,7 @@
             "__ZN30qvac_lib_inference_addon_llama26applyFinetuneParamHandlersEP8js_env_sRN28qvac_lib_inference_addon_cpp2js6ObjectERNS_21LlamaFinetuningParamsE",
             "__ZNK14MtmdLlmContext14getKvCellsUsedEv",
             "__ZNK30qvac_lib_inference_addon_llama8batching7Request15hasTokensToFeedEv",
+            "__ZN14TextLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEb",
             "__ZN30qvac_lib_inference_addon_llama8batching7Request19isOptPrefillPendingERKNSt3__18optionalIS1_EE",
             "__ZN14MtmdLlmContext21applyGenerationParamsERK16GenerationParams",
             "__ZN10LlamaModel6reloadENSt3__18optionalI23FinetuneConfigOverridesEE",
@@ -476,8 +488,8 @@
             "__ZN14TextLlmContext24rebuildSamplerFromLedgerERKN30qvac_lib_inference_addon_llama5cache6LedgerE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler28serviceCheckpointStopsLockedEv",
             "__ZN14MtmdLlmContext19acceptRestoredStateERKNSt3__16vectorIiNS0_9allocatorIiEEEERKNS0_12basic_stringIcNS0_11char_traitsIcEENS2_IcEEEE",
-            "__ZN24llama_finetuning_helpers20clearPauseCheckpointERKNSt3__14__fs10filesystem4pathE",
             "__ZN30qvac_lib_inference_addon_llama5utils20historyReasoningTagsEPK21common_chat_templatesPK11llama_modelb",
+            "__ZN24llama_finetuning_helpers20clearPauseCheckpointERKNSt3__14__fs10filesystem4pathE",
             "__ZN14MtmdLlmContext30capturePreRequestCacheSnapshotEv",
             "__ZN18AsyncWeightsLoader30extractIndividualStreamedFilesEv",
             "__ZN30qvac_lib_inference_addon_llama8batching20computeObservedStatsENSt3__16chrono10time_pointINS2_12steady_clockENS2_8durationIxNS1_5ratioILl1ELl1000000000EEEEEEERKNS0_7RequestENS1_8optionalI20GenerationStopReasonEE",
@@ -512,17 +524,19 @@
             "__ZNK30qvac_lib_inference_addon_llama8batching7Request16isGenerationIdleEv",
             "__ZNK30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler17slotOwnedByLockedEjy",
             "__ZN30qvac_lib_inference_addon_llama5utils21SequenceStateSnapshot19seedEmptyForTestingEi",
+            "__ZThn48_N14MtmdLlmContext29adoptCheckpointsFromCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEy",
             "__ZN18AsyncWeightsLoaderC1ERK10GGUFShardsR10InitLoaderRKNSt3__112basic_stringIcNS5_11char_traitsIcEENS5_9allocatorIcEEEEP13ModelMetaData",
             "__ZN17backend_selection20parseBackendOverrideERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN12CacheManager9loadCacheEv",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler14enqueueSaveJobERKNSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEEb",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler8freeSlotEj",
-            "__ZNK12CacheManager24wasCacheUsedInLastPromptEv",
             "__ZN14MtmdLlmContext6getCtxEv",
             "__ZN14LlamaFinetuner32clearPausedCheckpointStateSharedEv",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler17getOutputCallbackERNS1_9SlotStateEj",
+            "__ZNK12CacheManager24wasCacheUsedInLastPromptEv",
             "__ZN14TextLlmContextC2ERK13common_paramsRK15LlmModelContextii",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler6submitEONS0_13SubmitRequestE",
+            "__ZNK14TextLlmContext28appendCheckpointsToCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN12CacheManagerC1EP10LlmContextNSt3__18functionIFvbEEE",
             "__ZNK18AsyncWeightsLoader11isLastShardERKNSt3__14__fs10filesystem4pathE",
             "__ZN14TextLlmContext19appendResidentTokenEi",
@@ -602,11 +616,12 @@
             "__ZN30qvac_lib_inference_addon_llama7parsers26parseReasoningBudgetConfigENSt3__117basic_string_viewIcNS1_11char_traitsIcEEEE",
             "__ZN14MtmdLlmContext18adoptResidentStateERKNSt3__16vectorIiNS0_9allocatorIiEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching7RequestC2EjONSt3__16vectorIiNS2_9allocatorIiEEEEji",
-            "__ZN12CacheManager15savedCompletelyERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEm",
             "__ZNK10LlamaModel13isBitnetModelEv",
+            "__ZN12CacheManager15savedCompletelyERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEm",
+            "__ZThn48_NK14MtmdLlmContext25getToolDefinitionsDroppedEv",
             "__ZThn40_N10LlamaModelD0Ev",
             "__ZNK14TextLlmContext9saveCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
-            "__ZThn48_NK14MtmdLlmContext25getToolDefinitionsDroppedEv",
+            "__ZThn48_NK14MtmdLlmContext28appendCheckpointsToCacheFileERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching19MultiRequestBatcher6cancelEjRKNSt3__18functionIFvjEEE",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler18notifyDoneNoexceptEj",
             "__ZN30qvac_lib_inference_addon_llama8batching22aggregateObservedStatsERKNSt3__16vectorINS0_20ObservedRequestStatsENS1_9allocatorIS3_EEEE",
@@ -622,34 +637,33 @@
             "__ZN14MtmdLlmContext14preparePrefillERKNSt3__16vectorI15common_chat_msgNS0_9allocatorIS2_EEEERKNS1_I16common_chat_toolNS3_IS8_EEEERKNS1_INS1_IhNS3_IhEEEENS3_ISE_EEEERKNS1_I12PlannedMediaNS3_ISJ_EEEEbb",
             "__Z33qvacLibInferenceAddonLlamaExportsP8js_env_sP10js_value_s",
             "__ZN30qvac_lib_inference_addon_llama8batching24ContinuousBatchScheduler25ensureWorkerStartedLockedEv",
-            "__ZN14MtmdLlmContext13onSequenceEndERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE",
-            "__ZN14MtmdLlmContext9loadCacheERKNSt3__112basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEE"
+            "__ZN14MtmdLlmContext13onSequenceEndERKNSt3__18functionIFvRKNS0_12basic_stringIcNS0_11char_traitsIcEENS0_9allocatorIcEEEEEEE"
           ],
           "weak": [
             "__ZZZN28qvac_lib_inference_addon_cpp16OutputCallBackJs26initializeProcessingThreadENSt3__110shared_ptrINS_11OutputQueueEEEENKUlvE_clEvENUlP11uv_handle_sE_8__invokeES7_",
             "__ZN30qvac_lib_inference_addon_llama9assessFitEP8js_env_sP18js_callback_info_s",
             "__ZN30qvac_lib_inference_addon_llama6cancelEP8js_env_sP18js_callback_info_s",
+            "__ZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask13onEnvTeardownINSt3__19monostateEEEvP22js_deferred_teardown_sPv",
             "__ZN30qvac_lib_inference_addon_llama11activeSlotsEP8js_env_sP18js_callback_info_s",
-            "__ZZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask3runEP8js_env_sNSt3__18functionIFvvEEEENUlP11uv_handle_sE_8__invokeES9_",
             "__ZN28qvac_lib_inference_addon_cpp6logger8JsLogger13asyncCallbackEP10uv_async_s",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface6cancelEP8js_env_sP18js_callback_info_s",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface10activeJobsEP8js_env_sP18js_callback_info_s",
-            "__ZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask13onCloseHandleEP11uv_handle_s",
+            "__ZZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask5startINSt3__19monostateEEEP10js_value_sP8js_env_sNS1_12CallbackDataIT_E4WorkENSB_7MarshalEENUlP11uv_handle_sE_8__invokeESF_",
             "__ZN30qvac_lib_inference_addon_llama12discardCacheEP8js_env_sP18js_callback_info_s",
             "__ZN30qvac_lib_inference_addon_llama8finetuneEP8js_env_sP18js_callback_info_s",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface13releaseLoggerEP8js_env_sP18js_callback_info_s",
             "__ZZN28qvac_lib_inference_addon_cpp16OutputCallBackJs16drainAndTeardownEPNS0_5StateEbbENUlP11uv_handle_sE_8__invokeES4_",
+            "__ZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask13onCloseHandleINSt3__19monostateEEEvP11uv_handle_s",
+            "__ZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask10onCompleteINSt3__19monostateEEEvP10uv_async_s",
             "__ZZN28qvac_lib_inference_addon_cpp6logger8JsLogger22closeAsyncHandleLockedEvENUlP11uv_handle_sE_8__invokeES3_",
             "__ZN30qvac_lib_inference_addon_llama14createInstanceEP8js_env_sP18js_callback_info_s",
             "__ZN30qvac_lib_inference_addon_llama6runJobEP8js_env_sP18js_callback_info_s",
-            "__ZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask10onCompleteEP10uv_async_s",
             "__ZN30qvac_lib_inference_addon_llama9saveCacheEP8js_env_sP18js_callback_info_s",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface15destroyInstanceEP8js_env_sP18js_callback_info_s",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface8activateEP8js_env_sP18js_callback_info_s",
             "__ZN28qvac_lib_inference_addon_cpp6logger8JsLogger13onEnvTeardownEPv",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface9setLoggerEP8js_env_sP18js_callback_info_s",
             "__ZN28qvac_lib_inference_addon_cpp11JsInterface11loadWeightsEP8js_env_sP18js_callback_info_s",
-            "__ZN28qvac_lib_inference_addon_cpp2js11JsAsyncTask13onEnvTeardownEP22js_deferred_teardown_sPv",
             "__ZN28qvac_lib_inference_addon_cpp16OutputCallBackJs13onEnvTeardownEPv",
             "__ZN28qvac_lib_inference_addon_cpp16OutputCallBackJs16jsOutputCallbackEP10uv_async_s"
           ]