@superlinked/sie-sdk 0.4.1 → 0.5.0

package/dist/index.d.cts

@@ -108,6 +108,37 @@ interface ModelDims {
     sparse?: number;
     multivector?: number;
 }
+/**
+ * Advertised model capabilities.
+ *
+ * Mirrors the gateway `capabilities` object on each `/v1/models` entry
+ * (`ModelCapabilitiesWire`). All fields are optional; their presence
+ * depends on what the model config declares. `grammar` lists the
+ * supported grammar kinds ("json_schema" | "regex" | "ebnf").
+ * `code`/`sql`/`guard` are informational flags advertising validated
+ * generation jobs that back the model="code"/"sql"/"guard" aliases.
+ *
+ * These flags mean the model *supports* a task — they are NOT a
+ * precision-independent quality SLA. A flag is true at the model level even
+ * when quality is profile/precision-dependent (e.g. `sql` quality regresses
+ * under FP8; route SQL-critical traffic to a BF16 bundle via the `sql` alias).
+ */
+interface ModelCapabilities {
+    /** Supported grammar kinds: ["json_schema", "regex", "ebnf"] */
+    grammar?: string[];
+    /** Whether the model supports tool / function calling */
+    tools?: boolean;
+    /** Union of LoRA served-names across profiles (display summary) */
+    lora_adapters?: string[];
+    /** Per-profile LoRA served-names, keyed by profile name */
+    profile_lora_adapters?: Record<string, string[]>;
+    /** Validated for code generation; backs model="code" */
+    code?: boolean;
+    /** Supports text-to-SQL; backs model="sql". Precision-sensitive (FP8 regresses SQL) — a support flag, not a per-profile quality guarantee. */
+    sql?: boolean;
+    /** Generative guard model; backs model="guard" */
+    guard?: boolean;
+}
 /**
  * Information about a model returned by listModels().
  */
@@ -124,6 +155,8 @@ interface ModelInfo {
     dims?: ModelDims;
     /** Maximum sequence length the model supports */
     maxSequenceLength?: number;
+    /** Advertised model capabilities (grammar, tools, code/sql/guard, LoRA adapters) */
+    capabilities?: ModelCapabilities;
 }
 /**
  * A single score entry from reranking.
@@ -1235,7 +1268,7 @@ declare class SIEClient {
     private detectEndpointType;
 }
 
-declare const SDK_VERSION = "0.4.1";
+declare const SDK_VERSION = "0.5.0";
 
 /**
  * Helpers for converting SIE encode results to plain JavaScript types.
@@ -1559,7 +1592,7 @@ declare function unpackMessage<T = unknown>(data: Uint8Array): T;
 /**
  * Image handling utilities for the SIE TypeScript SDK.
  *
- * Per design.md Section 4.3, images are serialized as bytes for transport.
+ * Images are serialized as bytes for transport.
  * This module handles conversion from various input formats to Uint8Array.
  *
  * Supported input formats:
@@ -1589,7 +1622,6 @@ declare function unpackMessage<T = unknown>(data: Uint8Array): T;
 type ImageInput = Uint8Array | ArrayBuffer | Blob | string;
 /**
  * Wire format for images sent to the server.
- * Per design.md Section 4.3.
  */
 interface ImageWireFormat {
     data: Uint8Array;
@@ -1620,7 +1652,7 @@ declare function toImageBytes(input: ImageInput): Promise<Uint8Array>;
 /**
  * Convert image bytes to wire format for transport.
  *
- * Per design.md Section 4.3, images are sent as:
+ * Images are sent as:
  * `{ data: <bytes>, format: "jpeg" | "png" | "webp" }`
  *
  * @param input - Image data in any supported format
@@ -1636,4 +1668,4 @@ declare function toImageWireFormat(input: ImageInput, format?: "jpeg" | "png" |
  */
 declare function detectImageFormat(bytes: Uint8Array): "jpeg" | "png" | "webp" | "unknown";
 
-export { type CapacityInfo, type ChatChoice, type ChatChunkChoice, type ChatCompletion, type ChatCompletionChunk, type ChatCompletionRequest, type ChatDelta, type ChatFinishReason, type ChatMessage, type ChatUsage, type Classification, type ClusterStatusMessage, type ClusterSummary, type ClusterWorkerInfo, type DType, type DetectedObject, type EncodeOptions, type EncodeResult, type Entity, type ExtractOptions, type ExtractResult, type FinishReason, type GPUMetrics, type GenerateChunk, type GenerateOptions, type GenerateResult, type GenerationUsage, type ImageInput, type ImageWireFormat, InputTooLongError, type Item, LoraLoadingError, type ModelConfig, type ModelDims, type ModelInfo, ModelLoadFailedError, ModelLoadingError, type ModelState, type ModelStatus, type ModelSummary, type OutputType, PoolError, type PoolInfo, type PoolSpec, type PoolStatus, ProvisioningError, type Relation, RequestError, type ResponseFormat, SDK_VERSION, SIEClient, type SIEClientOptions, SIEConnectionError, SIEError, SIEStreamError, type ScoreEntry, type ScoreOptions, type ScoreResult, ServerError, type ServerInfo, type SparseResult, type SparseVector, type StatusMessage, type TimingInfo, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolSpec, type WorkerInfo, type WorkerStatusMessage, denseEmbedding, detectImageFormat, multivectorEmbedding, normalizeSparseVector, packMessage, sparseEmbedding, sparseEmbeddingMap, toFloat32Array, toImageBytes, toImageWireFormat, toNumberArray, unpackMessage };
+export { type CapacityInfo, type ChatChoice, type ChatChunkChoice, type ChatCompletion, type ChatCompletionChunk, type ChatCompletionRequest, type ChatDelta, type ChatFinishReason, type ChatMessage, type ChatUsage, type Classification, type ClusterStatusMessage, type ClusterSummary, type ClusterWorkerInfo, type DType, type DetectedObject, type EncodeOptions, type EncodeResult, type Entity, type ExtractOptions, type ExtractResult, type FinishReason, type GPUMetrics, type GenerateChunk, type GenerateOptions, type GenerateResult, type GenerationUsage, type ImageInput, type ImageWireFormat, InputTooLongError, type Item, LoraLoadingError, type ModelCapabilities, type ModelConfig, type ModelDims, type ModelInfo, ModelLoadFailedError, ModelLoadingError, type ModelState, type ModelStatus, type ModelSummary, type OutputType, PoolError, type PoolInfo, type PoolSpec, type PoolStatus, ProvisioningError, type Relation, RequestError, type ResponseFormat, SDK_VERSION, SIEClient, type SIEClientOptions, SIEConnectionError, SIEError, SIEStreamError, type ScoreEntry, type ScoreOptions, type ScoreResult, ServerError, type ServerInfo, type SparseResult, type SparseVector, type StatusMessage, type TimingInfo, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolSpec, type WorkerInfo, type WorkerStatusMessage, denseEmbedding, detectImageFormat, multivectorEmbedding, normalizeSparseVector, packMessage, sparseEmbedding, sparseEmbeddingMap, toFloat32Array, toImageBytes, toImageWireFormat, toNumberArray, unpackMessage };

package/dist/index.d.ts

@@ -108,6 +108,37 @@ interface ModelDims {
     sparse?: number;
     multivector?: number;
 }
+/**
+ * Advertised model capabilities.
+ *
+ * Mirrors the gateway `capabilities` object on each `/v1/models` entry
+ * (`ModelCapabilitiesWire`). All fields are optional; their presence
+ * depends on what the model config declares. `grammar` lists the
+ * supported grammar kinds ("json_schema" | "regex" | "ebnf").
+ * `code`/`sql`/`guard` are informational flags advertising validated
+ * generation jobs that back the model="code"/"sql"/"guard" aliases.
+ *
+ * These flags mean the model *supports* a task — they are NOT a
+ * precision-independent quality SLA. A flag is true at the model level even
+ * when quality is profile/precision-dependent (e.g. `sql` quality regresses
+ * under FP8; route SQL-critical traffic to a BF16 bundle via the `sql` alias).
+ */
+interface ModelCapabilities {
+    /** Supported grammar kinds: ["json_schema", "regex", "ebnf"] */
+    grammar?: string[];
+    /** Whether the model supports tool / function calling */
+    tools?: boolean;
+    /** Union of LoRA served-names across profiles (display summary) */
+    lora_adapters?: string[];
+    /** Per-profile LoRA served-names, keyed by profile name */
+    profile_lora_adapters?: Record<string, string[]>;
+    /** Validated for code generation; backs model="code" */
+    code?: boolean;
+    /** Supports text-to-SQL; backs model="sql". Precision-sensitive (FP8 regresses SQL) — a support flag, not a per-profile quality guarantee. */
+    sql?: boolean;
+    /** Generative guard model; backs model="guard" */
+    guard?: boolean;
+}
 /**
  * Information about a model returned by listModels().
  */
@@ -124,6 +155,8 @@ interface ModelInfo {
     dims?: ModelDims;
     /** Maximum sequence length the model supports */
     maxSequenceLength?: number;
+    /** Advertised model capabilities (grammar, tools, code/sql/guard, LoRA adapters) */
+    capabilities?: ModelCapabilities;
 }
 /**
  * A single score entry from reranking.
@@ -1235,7 +1268,7 @@ declare class SIEClient {
     private detectEndpointType;
 }
 
-declare const SDK_VERSION = "0.4.1";
+declare const SDK_VERSION = "0.5.0";
 
 /**
  * Helpers for converting SIE encode results to plain JavaScript types.
@@ -1559,7 +1592,7 @@ declare function unpackMessage<T = unknown>(data: Uint8Array): T;
 /**
  * Image handling utilities for the SIE TypeScript SDK.
  *
- * Per design.md Section 4.3, images are serialized as bytes for transport.
+ * Images are serialized as bytes for transport.
  * This module handles conversion from various input formats to Uint8Array.
  *
  * Supported input formats:
@@ -1589,7 +1622,6 @@ declare function unpackMessage<T = unknown>(data: Uint8Array): T;
 type ImageInput = Uint8Array | ArrayBuffer | Blob | string;
 /**
  * Wire format for images sent to the server.
- * Per design.md Section 4.3.
  */
 interface ImageWireFormat {
     data: Uint8Array;
@@ -1620,7 +1652,7 @@ declare function toImageBytes(input: ImageInput): Promise<Uint8Array>;
 /**
  * Convert image bytes to wire format for transport.
  *
- * Per design.md Section 4.3, images are sent as:
+ * Images are sent as:
  * `{ data: <bytes>, format: "jpeg" | "png" | "webp" }`
  *
  * @param input - Image data in any supported format
@@ -1636,4 +1668,4 @@ declare function toImageWireFormat(input: ImageInput, format?: "jpeg" | "png" |
  */
 declare function detectImageFormat(bytes: Uint8Array): "jpeg" | "png" | "webp" | "unknown";
 
-export { type CapacityInfo, type ChatChoice, type ChatChunkChoice, type ChatCompletion, type ChatCompletionChunk, type ChatCompletionRequest, type ChatDelta, type ChatFinishReason, type ChatMessage, type ChatUsage, type Classification, type ClusterStatusMessage, type ClusterSummary, type ClusterWorkerInfo, type DType, type DetectedObject, type EncodeOptions, type EncodeResult, type Entity, type ExtractOptions, type ExtractResult, type FinishReason, type GPUMetrics, type GenerateChunk, type GenerateOptions, type GenerateResult, type GenerationUsage, type ImageInput, type ImageWireFormat, InputTooLongError, type Item, LoraLoadingError, type ModelConfig, type ModelDims, type ModelInfo, ModelLoadFailedError, ModelLoadingError, type ModelState, type ModelStatus, type ModelSummary, type OutputType, PoolError, type PoolInfo, type PoolSpec, type PoolStatus, ProvisioningError, type Relation, RequestError, type ResponseFormat, SDK_VERSION, SIEClient, type SIEClientOptions, SIEConnectionError, SIEError, SIEStreamError, type ScoreEntry, type ScoreOptions, type ScoreResult, ServerError, type ServerInfo, type SparseResult, type SparseVector, type StatusMessage, type TimingInfo, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolSpec, type WorkerInfo, type WorkerStatusMessage, denseEmbedding, detectImageFormat, multivectorEmbedding, normalizeSparseVector, packMessage, sparseEmbedding, sparseEmbeddingMap, toFloat32Array, toImageBytes, toImageWireFormat, toNumberArray, unpackMessage };
+export { type CapacityInfo, type ChatChoice, type ChatChunkChoice, type ChatCompletion, type ChatCompletionChunk, type ChatCompletionRequest, type ChatDelta, type ChatFinishReason, type ChatMessage, type ChatUsage, type Classification, type ClusterStatusMessage, type ClusterSummary, type ClusterWorkerInfo, type DType, type DetectedObject, type EncodeOptions, type EncodeResult, type Entity, type ExtractOptions, type ExtractResult, type FinishReason, type GPUMetrics, type GenerateChunk, type GenerateOptions, type GenerateResult, type GenerationUsage, type ImageInput, type ImageWireFormat, InputTooLongError, type Item, LoraLoadingError, type ModelCapabilities, type ModelConfig, type ModelDims, type ModelInfo, ModelLoadFailedError, ModelLoadingError, type ModelState, type ModelStatus, type ModelSummary, type OutputType, PoolError, type PoolInfo, type PoolSpec, type PoolStatus, ProvisioningError, type Relation, RequestError, type ResponseFormat, SDK_VERSION, SIEClient, type SIEClientOptions, SIEConnectionError, SIEError, SIEStreamError, type ScoreEntry, type ScoreOptions, type ScoreResult, ServerError, type ServerInfo, type SparseResult, type SparseVector, type StatusMessage, type TimingInfo, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolSpec, type WorkerInfo, type WorkerStatusMessage, denseEmbedding, detectImageFormat, multivectorEmbedding, normalizeSparseVector, packMessage, sparseEmbedding, sparseEmbeddingMap, toFloat32Array, toImageBytes, toImageWireFormat, toNumberArray, unpackMessage };

package/dist/index.js

@@ -269,7 +269,7 @@ function packMessage(data) {
   return encode(data, { extensionCodec });
 }
 function isNumpyArrayMap(obj) {
-  if (typeof obj !== "object" || obj === null) {
+  if (obj === null || typeof obj !== "object") {
     return false;
   }
   const map = obj;
@@ -796,7 +796,7 @@ function extractDataPayload(block) {
 }
 
 // src/version.ts
-var SDK_VERSION = "0.4.1";
+var SDK_VERSION = "0.5.0";
 
 // src/client.ts
 function sleep2(ms) {
@@ -2091,10 +2091,11 @@ var SIEClient = class {
   createWebSocket(url) {
     const headers = this.apiKey ? { Authorization: `Bearer ${this.apiKey}` } : void 0;
     try {
-      if (headers) {
-        return new WebSocket(url, [], { headers });
+      if (!headers) {
+        return new WebSocket(url);
       }
-      return new WebSocket(url);
+      const args = [url, [], { headers }];
+      return Reflect.construct(WebSocket, args);
     } catch (error) {
       if (headers) {
         throw new SIEConnectionError(