@aithos/sdk 0.1.0-alpha.17 → 0.1.0-alpha.19

package/dist/src/compute.d.ts

@@ -118,6 +118,75 @@ export interface InvokeImageResult {
     /** Audit log id for traceability. */
     readonly auditId: string;
 }
+export interface InvokeBedrockVisionArgs {
+    readonly mandateId?: string;
+    /**
+     * Model id. Sonnet 4.6 is the default — it's vision-capable and
+     * returns reliable structured JSON when prompted.
+     */
+    readonly model?: string;
+    /** Source image — Blob (recommended) or raw base64. */
+    readonly image: Blob | {
+        readonly base64: string;
+        readonly contentType: string;
+    };
+    /** Text prompt accompanying the image. */
+    readonly prompt: string;
+    /** Optional system prompt. */
+    readonly system?: string;
+    readonly maxTokens?: number;
+    readonly temperature?: number;
+    readonly idempotencyKey?: string;
+    readonly signal?: AbortSignal;
+}
+export interface InvokeBedrockVisionResult {
+    readonly content: string;
+    readonly stopReason: StopReason;
+    readonly usage: {
+        readonly inputTokens: number;
+        readonly outputTokens: number;
+    };
+    readonly creditsCharged: number;
+    readonly walletBalance: number;
+    readonly auditId: string;
+}
+export interface InvokeSegmentationArgs {
+    /** Mandate id (optional for owner sessions — see InvokeImageArgs). */
+    readonly mandateId?: string;
+    /** Source image. Blob (recommended) or raw base64. */
+    readonly image: Blob | {
+        readonly base64: string;
+        readonly contentType: string;
+    };
+    /**
+     * Text phrase describing what to segment. Florence-2 is robust with
+     * natural-language descriptions: "the torso of the robot", "the
+     * dog's head", "the chest of the character".
+     */
+    readonly textInput: string;
+    readonly idempotencyKey?: string;
+    readonly signal?: AbortSignal;
+}
+export interface SegmentPolygon {
+    readonly points: ReadonlyArray<{
+        readonly x: number;
+        readonly y: number;
+    }>;
+}
+export interface InvokeSegmentationResult {
+    /** All polygons Florence-2 returned (typically 1, sometimes a few when the prompt matches multiple regions). */
+    readonly polygons: readonly SegmentPolygon[];
+    /** Bbox of the first polygon for callers that only need a coarse target. */
+    readonly bbox: {
+        readonly left: number;
+        readonly top: number;
+        readonly right: number;
+        readonly bottom: number;
+    } | null;
+    readonly creditsCharged: number;
+    readonly walletBalance: number;
+    readonly auditId: string;
+}
 export interface ComputeNamespaceDeps {
     readonly auth: AithosAuth;
     readonly appDid: string;
@@ -162,6 +231,16 @@ export declare class ComputeNamespace {
      *   `mandate_revoked`, `insufficient_credits`, …).
      */
     invokeBedrock(args: InvokeBedrockArgs): Promise<InvokeBedrockResult>;
+    /**
+     * Multimodal Bedrock invoke — image + text → text response.
+     * Default model: `claude-sonnet-4-6` (vision-capable, reliable JSON).
+     *
+     * Use when you need a VLM to reason about an image: locating
+     * features, structured extraction, semantic Q&A. Prompt the model
+     * to return JSON if you need structured output (the API itself is
+     * unstructured).
+     */
+    invokeBedrockVision(args: InvokeBedrockVisionArgs): Promise<InvokeBedrockVisionResult>;
     /**
      * Generate one or more images through the Aithos compute proxy
      * (currently powered by fal.ai FLUX models). Spec mirror of
@@ -182,5 +261,18 @@ export declare class ComputeNamespace {
      *   - flux-pro-1.1-ultra: 60 000 mc + fee per image
      */
     invokeImage(args: InvokeImageArgs): Promise<InvokeImageResult>;
+    /**
+     * Run text-prompted segmentation (Florence-2 referring-expression)
+     * on a source image. Returns one or more polygons hugging the
+     * region matching the text prompt.
+     *
+     * Use cases: locate the chest/torso area of a generated mascot
+     * for logo compositing, find the face zone for a thumbnail crop,
+     * extract a product from a marketing shot — anything that needs
+     * a precise mask + bbox from natural-language description.
+     *
+     * Pricing: flat 5 000 mc per call (~$0.005 — Florence-2 is cheap).
+     */
+    invokeSegmentation(args: InvokeSegmentationArgs): Promise<InvokeSegmentationResult>;
 }
 //# sourceMappingURL=compute.d.ts.map

package/dist/src/compute.js

@@ -86,6 +86,56 @@ export class ComputeNamespace {
             signal: args.signal,
         });
     }
+    /**
+     * Multimodal Bedrock invoke — image + text → text response.
+     * Default model: `claude-sonnet-4-6` (vision-capable, reliable JSON).
+     *
+     * Use when you need a VLM to reason about an image: locating
+     * features, structured extraction, semantic Q&A. Prompt the model
+     * to return JSON if you need structured output (the API itself is
+     * unstructured).
+     */
+    async invokeBedrockVision(args) {
+        const { endpoints, fetch: fetchImpl } = this.#deps;
+        const choice = this.#resolveSigner(args.mandateId);
+        let imageBase64;
+        let imageContentType;
+        if ("base64" in args.image) {
+            imageBase64 = args.image.base64;
+            imageContentType = args.image.contentType;
+        }
+        else {
+            const buf = await args.image.arrayBuffer();
+            imageBase64 = arrayBufferToBase64(buf);
+            imageContentType = args.image.type || "image/png";
+        }
+        const url = computeInvokeUrl(endpoints);
+        const idempotencyKey = args.idempotencyKey ?? generateIdempotencyKey();
+        const model = args.model ?? "claude-sonnet-4-6";
+        const params = {
+            app_did: this.#deps.appDid,
+            mandate_id: this.#resolveMandateIdForWire(args.mandateId, choice),
+            model,
+            image_base64: imageBase64,
+            image_content_type: imageContentType,
+            prompt: args.prompt,
+            idempotency_key: idempotencyKey,
+        };
+        if (args.system !== undefined)
+            params.system = args.system;
+        if (args.maxTokens !== undefined)
+            params.max_tokens = args.maxTokens;
+        if (args.temperature !== undefined)
+            params.temperature = args.temperature;
+        return await this.#signAndPost({
+            url,
+            method: "aithos.compute_invoke_bedrock_vision",
+            params,
+            choice,
+            fetchImpl,
+            signal: args.signal,
+        });
+    }
     /**
      * Generate one or more images through the Aithos compute proxy
      * (currently powered by fal.ai FLUX models). Spec mirror of
@@ -138,6 +188,52 @@ export class ComputeNamespace {
             signal: args.signal,
         });
     }
+    /**
+     * Run text-prompted segmentation (Florence-2 referring-expression)
+     * on a source image. Returns one or more polygons hugging the
+     * region matching the text prompt.
+     *
+     * Use cases: locate the chest/torso area of a generated mascot
+     * for logo compositing, find the face zone for a thumbnail crop,
+     * extract a product from a marketing shot — anything that needs
+     * a precise mask + bbox from natural-language description.
+     *
+     * Pricing: flat 5 000 mc per call (~$0.005 — Florence-2 is cheap).
+     */
+    async invokeSegmentation(args) {
+        const { endpoints, fetch: fetchImpl } = this.#deps;
+        const choice = this.#resolveSigner(args.mandateId);
+        // Normalize image input to base64 + content type.
+        let imageBase64;
+        let imageContentType;
+        if ("base64" in args.image) {
+            imageBase64 = args.image.base64;
+            imageContentType = args.image.contentType;
+        }
+        else {
+            const buf = await args.image.arrayBuffer();
+            imageBase64 = arrayBufferToBase64(buf);
+            imageContentType = args.image.type || "image/png";
+        }
+        const url = computeInvokeUrl(endpoints);
+        const idempotencyKey = args.idempotencyKey ?? generateIdempotencyKey();
+        const params = {
+            app_did: this.#deps.appDid,
+            mandate_id: this.#resolveMandateIdForWire(args.mandateId, choice),
+            image_base64: imageBase64,
+            image_content_type: imageContentType,
+            text_input: args.textInput,
+            idempotency_key: idempotencyKey,
+        };
+        return await this.#signAndPost({
+            url,
+            method: "aithos.compute_invoke_segmentation",
+            params,
+            choice,
+            fetchImpl,
+            signal: args.signal,
+        });
+    }
     /**
      * Resolve the active signer (owner takes precedence over delegate).
      *
@@ -267,4 +363,19 @@ function generateIdempotencyKey() {
     }
     return hex;
 }
+/**
+ * Encode an ArrayBuffer as base64 in environments where `Buffer` is
+ * not available (browser). Uses btoa over a binary string — safe for
+ * the small payload sizes the SDK deals with (≤ a few MB).
+ */
+function arrayBufferToBase64(buf) {
+    const bytes = new Uint8Array(buf);
+    let bin = "";
+    // Process in chunks to avoid stack overflow on String.fromCharCode.apply
+    const CHUNK = 0x8000;
+    for (let i = 0; i < bytes.length; i += CHUNK) {
+        bin += String.fromCharCode.apply(null, Array.from(bytes.subarray(i, i + CHUNK)));
+    }
+    return btoa(bin);
+}
 //# sourceMappingURL=compute.js.map

package/dist/src/index.d.ts

@@ -1,11 +1,11 @@
-export declare const VERSION = "0.1.0-alpha.17";
+export declare const VERSION = "0.1.0-alpha.19";
 export { AithosSDK } from "./sdk.js";
 export type { AithosSDKConfig } from "./types.js";
 export { AithosSDKError } from "./types.js";
 export { AithosRpcError } from "@aithos/protocol-client";
 export type { AithosSdkEndpoints } from "./endpoints.js";
 export { DEFAULT_SDK_ENDPOINTS } from "./endpoints.js";
-export type { ComputeMessage, ImageAspectRatio, ImageModelId, InvokeBedrockArgs, InvokeBedrockResult, InvokeImageArgs, InvokeImageImage, InvokeImageResult, StopReason, } from "./compute.js";
+export type { ComputeMessage, ImageAspectRatio, ImageModelId, InvokeBedrockArgs, InvokeBedrockResult, InvokeBedrockVisionArgs, InvokeBedrockVisionResult, InvokeImageArgs, InvokeImageImage, InvokeImageResult, InvokeSegmentationArgs, InvokeSegmentationResult, SegmentPolygon, StopReason, } from "./compute.js";
 export { ComputeNamespace } from "./compute.js";
 export type { CreditPackId, CreateTopupSessionArgs, CreateTopupSessionResult, GetBalanceArgs, GetBalanceResult, } from "./wallet.js";
 export { WalletNamespace } from "./wallet.js";

package/dist/src/index.js

@@ -17,7 +17,7 @@
 // Public types specific to the SDK (`AithosSDKConfig`, `AithosSDKError`)
 // are exported from here. Endpoint config (`AithosSdkEndpoints`,
 // `DEFAULT_SDK_ENDPOINTS`) likewise.
-export const VERSION = "0.1.0-alpha.17";
+export const VERSION = "0.1.0-alpha.19";
 export { AithosSDK } from "./sdk.js";
 export { AithosSDKError } from "./types.js";
 // Re-export protocol-client's JSON-RPC error type so consumers can

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aithos/sdk",
-  "version": "0.1.0-alpha.17",
+  "version": "0.1.0-alpha.19",
   "description": "Aithos SDK — high-level TypeScript developer kit for building agentic apps on the Aithos protocol. Wraps @aithos/protocol-client and exposes the Aithos compute proxy and wallet (Stripe top-up) endpoints.",
   "keywords": [
     "aithos",