@juspay/neurolink 9.79.1 → 9.79.3

package/dist/lib/providers/anthropicImageBlocks.js

@@ -0,0 +1,223 @@
+/**
+ * Pure converters from multimodal content parts into native Anthropic
+ * Messages-API content blocks (image / document).
+ *
+ * Two input shapes reach the native Anthropic surface:
+ *
+ *  1. NeuroLink/V3 multimodal parts — `{ type: "image", image }` produced by
+ *     the multimodal message builder (base64, data URL, https URL, or bytes).
+ *
+ *  2. AI-SDK LanguageModel prompt parts — `{ type: "file", mediaType, data }`.
+ *     This is how `ai@6` encodes BOTH images and PDFs in the prompt that the
+ *     provider's `doGenerate(options.prompt)` receives. Before this module the
+ *     converter only handled `type:"image"`, so on the tool-using generate
+ *     path (which always normalises images to `type:"file"`) the image part
+ *     was silently dropped — the model never saw the image ("no image
+ *     detected"). Gemini/Vertex are immune because they read `input.images`
+ *     directly in their own builders instead of going through this conversion.
+ *
+ * Media type is taken from the AI-SDK-provided `mediaType` when present, then
+ * sniffed from magic bytes, and only then defaulted — a hardcoded `image/png`
+ * default silently corrupts JPEG/GIF/WebP uploads (the Anthropic API rejects a
+ * mislabeled base64 image with HTTP 400 and the image vanishes).
+ */
+// The base64 image media types the Anthropic Messages API accepts are exactly
+// `Anthropic.Messages.Base64ImageSource["media_type"]` — referenced inline below
+// rather than redeclared as a local alias (project rule: type aliases live in
+// src/lib/types/, and this provider-internal shape does not warrant a barrel entry).
+const SUPPORTED_IMAGE_MEDIA_TYPES = new Set([
+    "image/jpeg",
+    "image/png",
+    "image/gif",
+    "image/webp",
+]);
+/** Map a caller-provided MIME hint onto a supported image media type (or undefined). */
+const normalizeImageMediaType = (hint) => {
+    if (!hint) {
+        return undefined;
+    }
+    const h = hint.toLowerCase().split(";")[0].trim();
+    if (h === "image/jpg") {
+        return "image/jpeg";
+    }
+    return SUPPORTED_IMAGE_MEDIA_TYPES.has(h)
+        ? h
+        : undefined;
+};
+/**
+ * Detect a supported image media type from a buffer's magic bytes. Returns
+ * undefined when the bytes are not one of the four Anthropic-supported formats.
+ */
+export function sniffImageMediaType(bytes) {
+    // PNG: 89 50 4E 47 0D 0A 1A 0A
+    if (bytes.length >= 8 &&
+        bytes[0] === 0x89 &&
+        bytes[1] === 0x50 &&
+        bytes[2] === 0x4e &&
+        bytes[3] === 0x47) {
+        return "image/png";
+    }
+    // JPEG: FF D8 FF
+    if (bytes.length >= 3 &&
+        bytes[0] === 0xff &&
+        bytes[1] === 0xd8 &&
+        bytes[2] === 0xff) {
+        return "image/jpeg";
+    }
+    // GIF: "GIF"
+    if (bytes.length >= 6 &&
+        bytes[0] === 0x47 &&
+        bytes[1] === 0x49 &&
+        bytes[2] === 0x46) {
+        return "image/gif";
+    }
+    // WebP: "RIFF"...."WEBP"
+    if (bytes.length >= 12 &&
+        bytes[0] === 0x52 &&
+        bytes[1] === 0x49 &&
+        bytes[2] === 0x46 &&
+        bytes[3] === 0x46 &&
+        bytes[8] === 0x57 &&
+        bytes[9] === 0x45 &&
+        bytes[10] === 0x42 &&
+        bytes[11] === 0x50) {
+        return "image/webp";
+    }
+    return undefined;
+}
+// The longest signature `sniffImageMediaType` checks is WebP — "RIFF"…"WEBP",
+// which spans the first 12 decoded bytes. Base64 packs 3 bytes per 4 chars, so
+// 16 chars already cover those 12 bytes; we slice 32 (a clean 4-char boundary
+// that decodes to 24 bytes) for a comfortable margin over every signature.
+const SNIFF_BASE64_CHARS = 32;
+/** Sniff the leading bytes of a base64 payload (best-effort, never throws). */
+const sniffBase64 = (b64) => {
+    try {
+        return sniffImageMediaType(Buffer.from(b64.slice(0, SNIFF_BASE64_CHARS), "base64"));
+    }
+    catch {
+        return undefined;
+    }
+};
+/**
+ * Convert an image part (data URL, bare base64, https URL, byte array, or
+ * ArrayBuffer) into an Anthropic image block. Honors `mediaTypeHint` (the
+ * AI-SDK `mediaType`), falls back to magic-byte sniffing, then to image/png.
+ * Returns undefined for unusable inputs.
+ */
+export function toAnthropicImageBlock(data, mediaTypeHint) {
+    if (data instanceof ArrayBuffer) {
+        return toAnthropicImageBlock(new Uint8Array(data), mediaTypeHint);
+    }
+    if (data instanceof Uint8Array) {
+        const media_type = normalizeImageMediaType(mediaTypeHint) ??
+            sniffImageMediaType(data) ??
+            "image/png";
+        return {
+            type: "image",
+            source: {
+                type: "base64",
+                media_type,
+                data: Buffer.from(data).toString("base64"),
+            },
+        };
+    }
+    if (typeof data !== "string" && !(data instanceof URL)) {
+        return undefined;
+    }
+    const str = data instanceof URL ? data.toString() : data;
+    const dataUrlMatch = str.match(/^data:(image\/[a-z0-9+.-]+);base64,(.+)$/i);
+    if (dataUrlMatch) {
+        const media_type = normalizeImageMediaType(dataUrlMatch[1]) ?? sniffBase64(dataUrlMatch[2]);
+        if (!media_type) {
+            return undefined;
+        }
+        return {
+            type: "image",
+            source: { type: "base64", media_type, data: dataUrlMatch[2] },
+        };
+    }
+    if (/^https?:\/\//i.test(str)) {
+        return { type: "image", source: { type: "url", url: str } };
+    }
+    // Bare base64 payload — prefer the hint, then sniff, then assume PNG
+    // (matches the OpenAI-compat client's historical default).
+    const media_type = normalizeImageMediaType(mediaTypeHint) ?? sniffBase64(str) ?? "image/png";
+    return {
+        type: "image",
+        source: { type: "base64", media_type, data: str },
+    };
+}
+/** Convert a PDF file part into an Anthropic document block. */
+const toAnthropicPdfBlock = (data) => {
+    if (data instanceof ArrayBuffer) {
+        return toAnthropicPdfBlock(new Uint8Array(data));
+    }
+    if (data instanceof Uint8Array) {
+        return {
+            type: "document",
+            source: {
+                type: "base64",
+                media_type: "application/pdf",
+                data: Buffer.from(data).toString("base64"),
+            },
+        };
+    }
+    if (data instanceof URL) {
+        return { type: "document", source: { type: "url", url: data.toString() } };
+    }
+    if (typeof data === "string") {
+        const m = data.match(/^data:application\/pdf;base64,(.+)$/i);
+        if (m) {
+            return {
+                type: "document",
+                source: { type: "base64", media_type: "application/pdf", data: m[1] },
+            };
+        }
+        if (/^https?:\/\//i.test(data)) {
+            return { type: "document", source: { type: "url", url: data } };
+        }
+        return {
+            type: "document",
+            source: { type: "base64", media_type: "application/pdf", data },
+        };
+    }
+    return undefined;
+};
+/**
+ * Convert an AI-SDK `{ type: "file", mediaType, data }` content part into the
+ * matching Anthropic content block: image/* → image block (media type honored,
+ * not hardcoded), application/pdf → document block. Returns undefined for
+ * unsupported media types (so the caller simply omits them rather than 400ing).
+ * When `mediaType` is absent, image bytes are still salvaged via sniffing.
+ */
+export function fileToAnthropicBlock(part) {
+    const data = part?.data;
+    if (data === undefined || data === null) {
+        return undefined;
+    }
+    const mediaType = typeof part.mediaType === "string"
+        ? part.mediaType.toLowerCase().split(";")[0].trim()
+        : "";
+    if (SUPPORTED_IMAGE_MEDIA_TYPES.has(mediaType) || mediaType === "image/jpg") {
+        // Hint is one of the four Anthropic-supported image types (or the jpg alias).
+        return toAnthropicImageBlock(data, mediaType);
+    }
+    if (mediaType === "application/pdf") {
+        return toAnthropicPdfBlock(data);
+    }
+    // No media type, unknown media type, OR an unsupported image/* hint
+    // (e.g. image/svg+xml, image/bmp) — attempt magic-byte salvage so that the
+    // part is included only when the actual bytes sniff to a supported format,
+    // otherwise omit it rather than mislabeling it as image/png.
+    const bytes = data instanceof ArrayBuffer
+        ? new Uint8Array(data)
+        : data instanceof Uint8Array
+            ? data
+            : undefined;
+    if (bytes && sniffImageMediaType(bytes)) {
+        return toAnthropicImageBlock(bytes);
+    }
+    return undefined;
+}
+//# sourceMappingURL=anthropicImageBlocks.js.map

package/dist/lib/providers/googleNativeGemini3.d.ts

@@ -96,6 +96,32 @@ export declare function buildNativeConfig(options: {
  * Compute a safe, clamped maxSteps value.
  */
 export declare function computeMaxSteps(rawMaxSteps?: number): number;
+/**
+ * Map a `@google/genai` `Candidate.finishReason` enum value onto NeuroLink's
+ * unified finish reason, mirroring anthropic.ts `mapAnthropicStopReason`.
+ *
+ * Enum values per `@google/genai` `FinishReason`: STOP, MAX_TOKENS, SAFETY,
+ * RECITATION, LANGUAGE, OTHER, BLOCKLIST, PROHIBITED_CONTENT, SPII,
+ * MALFORMED_FUNCTION_CALL, IMAGE_SAFETY, UNEXPECTED_TOOL_CALL,
+ * FINISH_REASON_UNSPECIFIED. Unknown / unset / non-terminal values default to
+ * "stop" (a clean completion is the safe assumption).
+ *
+ * Returns a plain string (not a `{ unified, raw }` object): the Vertex result
+ * builders and the consuming layer (neurolink.ts `finishReason || "unknown"`
+ * and `finishReason === "length"`) compare against plain strings.
+ */
+export declare function mapGeminiFinishReason(raw: string | null | undefined): "stop" | "length" | "tool-calls" | "content-filter";
+/**
+ * Append a step's text to a running cross-step accumulator, ignoring empty
+ * steps and inserting a single newline between non-empty contributions.
+ *
+ * The native Gemini loops overwrite per-step text into `lastStepText`, so when
+ * the loop is force-terminated by the step cap the intermediate tool-step prose
+ * is lost and a canned placeholder becomes the answer. Accumulating here mirrors
+ * the Vertex-Claude loop's `aggregatedTurnText += block.text` so the gathered
+ * text can be surfaced at the maxSteps-exhaustion exit instead of the placeholder.
+ */
+export declare function appendStepText(accumulated: string, stepText: string): string;
 /**
  * Process stream chunks to extract raw response parts, function calls, and usage metadata.
  *

package/dist/lib/providers/googleNativeGemini3.js

@@ -450,6 +450,54 @@ export function computeMaxSteps(rawMaxSteps) {
         ? Math.min(Math.floor(value), GEMINI3_NATIVE_MAX_STEPS)
         : Math.min(DEFAULT_MAX_STEPS, GEMINI3_NATIVE_MAX_STEPS);
 }
+/**
+ * Map a `@google/genai` `Candidate.finishReason` enum value onto NeuroLink's
+ * unified finish reason, mirroring anthropic.ts `mapAnthropicStopReason`.
+ *
+ * Enum values per `@google/genai` `FinishReason`: STOP, MAX_TOKENS, SAFETY,
+ * RECITATION, LANGUAGE, OTHER, BLOCKLIST, PROHIBITED_CONTENT, SPII,
+ * MALFORMED_FUNCTION_CALL, IMAGE_SAFETY, UNEXPECTED_TOOL_CALL,
+ * FINISH_REASON_UNSPECIFIED. Unknown / unset / non-terminal values default to
+ * "stop" (a clean completion is the safe assumption).
+ *
+ * Returns a plain string (not a `{ unified, raw }` object): the Vertex result
+ * builders and the consuming layer (neurolink.ts `finishReason || "unknown"`
+ * and `finishReason === "length"`) compare against plain strings.
+ */
+export function mapGeminiFinishReason(raw) {
+    switch (raw) {
+        case "MAX_TOKENS":
+            return "length";
+        case "MALFORMED_FUNCTION_CALL":
+        case "UNEXPECTED_TOOL_CALL":
+            return "tool-calls";
+        case "SAFETY":
+        case "RECITATION":
+        case "BLOCKLIST":
+        case "PROHIBITED_CONTENT":
+        case "SPII":
+        case "IMAGE_SAFETY":
+            return "content-filter";
+        default:
+            return "stop";
+    }
+}
+/**
+ * Append a step's text to a running cross-step accumulator, ignoring empty
+ * steps and inserting a single newline between non-empty contributions.
+ *
+ * The native Gemini loops overwrite per-step text into `lastStepText`, so when
+ * the loop is force-terminated by the step cap the intermediate tool-step prose
+ * is lost and a canned placeholder becomes the answer. Accumulating here mirrors
+ * the Vertex-Claude loop's `aggregatedTurnText += block.text` so the gathered
+ * text can be surfaced at the maxSteps-exhaustion exit instead of the placeholder.
+ */
+export function appendStepText(accumulated, stepText) {
+    if (!stepText) {
+        return accumulated;
+    }
+    return accumulated ? `${accumulated}\n${stepText}` : stepText;
+}
 /**
  * Process stream chunks to extract raw response parts, function calls, and usage metadata.
  *

package/dist/lib/providers/googleVertex.d.ts

@@ -160,6 +160,22 @@ export declare class GoogleVertexProvider extends BaseProvider {
      * This bypasses @ai-sdk/google-vertex to properly handle thought_signature
      */
     private executeNativeGemini3Generate;
+    /**
+     * One-shot, tools-disabled model call used when a native Gemini agentic loop
+     * is force-terminated by the step cap with no text produced. Lets the model
+     * synthesize a final answer from the function results already in `contents`
+     * instead of returning a canned placeholder (Bug 1, part b).
+     *
+     * Tools are disabled by OMITTING `config.tools` — the codebase's established
+     * mechanism. `@google/genai`'s `FunctionCallingConfigMode.NONE` is documented
+     * as equivalent to passing no function declarations, and `functionCallingConfig`
+     * is not used anywhere in this codebase. When the structured-output
+     * (`final_result`) pattern was active, a trailing instruction countermands the
+     * earlier "you MUST call final_result" directive so the model answers in plain
+     * text. Never throws — returns empty text so the caller falls back to the
+     * placeholder, guaranteeing no new failure path.
+     */
+    private synthesizeFinalAnswerWithoutTools;
     /**
      * Create native AnthropicVertex client for Claude models
      */