Multimodality: Upload PDFs and Images (#135)

2025-05-20 10:23:21 -07:00
parent f86630075e cb2d216973
commit 06e0de6f85
10 changed files with 1947 additions and 1214 deletions
--- a/next.config.mjs
+++ b/next.config.mjs
@@ -1,4 +1,10 @@
 /** @type {import('next').NextConfig} */
-const nextConfig = {};
+const nextConfig = {
+  experimental: {
+    serverActions: {
+      bodySizeLimit: "10mb",
+    },
+  },
+};

 export default nextConfig;
--- a/package.json
+++ b/package.json
@@ -53,7 +53,7 @@
    "tailwind-merge": "^3.0.2",
    "tailwindcss-animate": "^1.0.7",
    "use-stick-to-bottom": "^1.0.46",
-    "uuid": "^11.0.5",
+    "uuid": "^11.1.0",
    "zod": "^3.24.2"
  },
  "devDependencies": {
@@ -64,6 +64,7 @@
    "@types/react": "^19.0.8",
    "@types/react-dom": "^19.0.3",
    "@types/react-syntax-highlighter": "^15.5.13",
+    "@types/uuid": "^10.0.0",
    "autoprefixer": "^10.4.20",
    "dotenv": "^16.4.7",
    "eslint": "^9.19.0",
@@ -81,8 +82,7 @@
    "typescript-eslint": "^8.22.0"
  },
  "overrides": {
-    "react-is": "^19.0.0-rc-69d4b800-20241021",
-    "@langchain/langgraph-checkpoint": "^0.0.16"
+    "react-is": "^19.0.0-rc-69d4b800-20241021"
  },
  "packageManager": "pnpm@10.5.1"
 }
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
--- a/src/components/thread/ContentBlocksPreview.tsx
+++ b/src/components/thread/ContentBlocksPreview.tsx
@@ -0,0 +1,37 @@
+import React from "react";
+import type { Base64ContentBlock } from "@langchain/core/messages";
+import { MultimodalPreview } from "../ui/MultimodalPreview";
+import { cn } from "@/lib/utils";
+
+interface ContentBlocksPreviewProps {
+  blocks: Base64ContentBlock[];
+  onRemove: (idx: number) => void;
+  size?: "sm" | "md" | "lg";
+  className?: string;
+}
+
+/**
+ * Renders a preview of content blocks with optional remove functionality.
+ * Uses cn utility for robust class merging.
+ */
+export const ContentBlocksPreview: React.FC<ContentBlocksPreviewProps> = ({
+  blocks,
+  onRemove,
+  size = "md",
+  className,
+}) => {
+  if (!blocks.length) return null;
+  return (
+    <div className={cn("flex flex-wrap gap-2 p-3.5 pb-0", className)}>
+      {blocks.map((block, idx) => (
+        <MultimodalPreview
+          key={idx}
+          block={block}
+          removable
+          onRemove={() => onRemove(idx)}
+          size={size}
+        />
+      ))}
+    </div>
+  );
+};
--- a/src/components/thread/index.tsx
+++ b/src/components/thread/index.tsx
@@ -21,6 +21,8 @@ import {
  PanelRightClose,
  SquarePen,
  XIcon,
+  Plus,
+  CircleX,
 } from "lucide-react";
 import { useQueryState, parseAsBoolean } from "nuqs";
 import { StickToBottom, useStickToBottomContext } from "use-stick-to-bottom";
@@ -36,6 +38,8 @@ import {
  TooltipProvider,
  TooltipTrigger,
 } from "../ui/tooltip";
+import { useFileUpload } from "@/hooks/use-file-upload";
+import { ContentBlocksPreview } from "./ContentBlocksPreview";
 import {
  useArtifactOpen,
  ArtifactContent,
@@ -122,6 +126,14 @@ export function Thread() {
    parseAsBoolean.withDefault(false),
  );
  const [input, setInput] = useState("");
+  const {
+    contentBlocks,
+    setContentBlocks,
+    handleFileUpload,
+    dropRef,
+    removeBlock,
+    resetBlocks,
+  } = useFileUpload();
  const [firstTokenReceived, setFirstTokenReceived] = useState(false);
  const isLargeScreen = useMediaQuery("(min-width: 1024px)");

@@ -183,13 +195,17 @@ export function Thread() {

  const handleSubmit = (e: FormEvent) => {
    e.preventDefault();
-    if (!input.trim() || isLoading) return;
+    if ((input.trim().length === 0 && contentBlocks.length === 0) || isLoading)
+      return;
    setFirstTokenReceived(false);

    const newHumanMessage: Message = {
      id: uuidv4(),
      type: "human",
-      content: input,
+      content: [
+        ...(input.trim().length > 0 ? [{ type: "text", text: input }] : []),
+        ...contentBlocks,
+      ] as Message["content"],
    };

    const toolMessages = ensureToolCallsHaveResponses(stream.messages);
@@ -214,6 +230,7 @@ export function Thread() {
    );

    setInput("");
+    setContentBlocks([]);
  };

  const handleRegenerate = (
@@ -423,11 +440,18 @@ export function Thread() {

                  <ScrollToBottom className="animate-in fade-in-0 zoom-in-95 absolute bottom-full left-1/2 mb-4 -translate-x-1/2" />

-                  <div className="bg-muted relative z-10 mx-auto mb-8 w-full max-w-3xl rounded-2xl border shadow-xs">
+                  <div
+                    ref={dropRef}
+                    className="bg-muted relative z-10 mx-auto mb-8 w-full max-w-3xl rounded-2xl border shadow-xs"
+                  >
                    <form
                      onSubmit={handleSubmit}
                      className="mx-auto grid max-w-3xl grid-rows-[1fr_auto] gap-2"
                    >
+                      <ContentBlocksPreview
+                        blocks={contentBlocks}
+                        onRemove={removeBlock}
+                      />
                      <textarea
                        value={input}
                        onChange={(e) => setInput(e.target.value)}
@@ -448,7 +472,7 @@ export function Thread() {
                        className="field-sizing-content resize-none border-none bg-transparent p-3.5 pb-0 shadow-none ring-0 outline-none focus:ring-0 focus:outline-none"
                      />

-                      <div className="flex items-center justify-between p-2 pt-4">
+                      <div className="flex items-center gap-6 p-2 pt-4">
                        <div>
                          <div className="flex items-center space-x-2">
                            <Switch
@@ -464,10 +488,28 @@ export function Thread() {
                            </Label>
                          </div>
                        </div>
+                        <Label
+                          htmlFor="file-input"
+                          className="flex cursor-pointer items-center gap-2"
+                        >
+                          <Plus className="size-5 text-gray-600" />
+                          <span className="text-sm text-gray-600">
+                            Upload PDF or Image
+                          </span>
+                        </Label>
+                        <input
+                          id="file-input"
+                          type="file"
+                          onChange={handleFileUpload}
+                          multiple
+                          accept="image/jpeg,image/png,image/gif,image/webp,application/pdf"
+                          className="hidden"
+                        />
                        {stream.isLoading ? (
                          <Button
                            key="stop"
                            onClick={() => stream.stop()}
+                            className="ml-auto"
                          >
                            <LoaderCircle className="h-4 w-4 animate-spin" />
                            Cancel
@@ -475,8 +517,11 @@ export function Thread() {
                        ) : (
                          <Button
                            type="submit"
-                            className="shadow-md transition-all"
-                            disabled={isLoading || !input.trim()}
+                            className="ml-auto shadow-md transition-all"
+                            disabled={
+                              isLoading ||
+                              (!input.trim() && contentBlocks.length === 0)
+                            }
                          >
                            Send
                          </Button>
--- a/src/components/thread/messages/human.tsx
+++ b/src/components/thread/messages/human.tsx
@@ -5,6 +5,8 @@ import { getContentString } from "../utils";
 import { cn } from "@/lib/utils";
 import { Textarea } from "@/components/ui/textarea";
 import { BranchSwitcher, CommandBar } from "./shared";
+import { MultimodalPreview } from "@/components/ui/MultimodalPreview";
+import type { Base64ContentBlock } from "@langchain/core/messages";

 function EditableContent({
  value,
@@ -32,6 +34,36 @@ function EditableContent({
  );
 }

+// Type guard for Base64ContentBlock
+function isBase64ContentBlock(block: unknown): block is Base64ContentBlock {
+  if (typeof block !== "object" || block === null || !("type" in block))
+    return false;
+  // file type (legacy)
+  if (
+    (block as { type: unknown }).type === "file" &&
+    "source_type" in block &&
+    (block as { source_type: unknown }).source_type === "base64" &&
+    "mime_type" in block &&
+    typeof (block as { mime_type?: unknown }).mime_type === "string" &&
+    ((block as { mime_type: string }).mime_type.startsWith("image/") ||
+      (block as { mime_type: string }).mime_type === "application/pdf")
+  ) {
+    return true;
+  }
+  // image type (new)
+  if (
+    (block as { type: unknown }).type === "image" &&
+    "source_type" in block &&
+    (block as { source_type: unknown }).source_type === "base64" &&
+    "mime_type" in block &&
+    typeof (block as { mime_type?: unknown }).mime_type === "string" &&
+    (block as { mime_type: string }).mime_type.startsWith("image/")
+  ) {
+    return true;
+  }
+  return false;
+}
+
 export function HumanMessage({
  message,
  isLoading,
@@ -84,9 +116,34 @@ export function HumanMessage({
            onSubmit={handleSubmitEdit}
          />
        ) : (
-          <p className="bg-muted ml-auto w-fit rounded-3xl px-4 py-2 whitespace-pre-wrap">
-            {contentString}
-          </p>
+          <div className="flex flex-col gap-2">
+            {/* Render images and files if no text */}
+            {Array.isArray(message.content) && message.content.length > 0 && (
+              <div className="flex flex-col items-end gap-2">
+                {message.content.reduce<React.ReactNode[]>(
+                  (acc, block, idx) => {
+                    if (isBase64ContentBlock(block)) {
+                      acc.push(
+                        <MultimodalPreview
+                          key={idx}
+                          block={block}
+                          size="md"
+                        />,
+                      );
+                    }
+                    return acc;
+                  },
+                  [],
+                )}
+              </div>
+            )}
+            {/* Render text if present, otherwise fallback to file/image name */}
+            {contentString ? (
+              <p className="bg-muted ml-auto w-fit rounded-3xl px-4 py-2 text-right whitespace-pre-wrap">
+                {contentString}
+              </p>
+            ) : null}
+          </div>
        )}

        <div
--- a/src/components/thread/utils.ts
+++ b/src/components/thread/utils.ts
@@ -1,5 +1,11 @@
 import type { Message } from "@langchain/langgraph-sdk";

+/**
+ * Extracts a string summary from a message's content, supporting multimodal (text, image, file, etc.).
+ * - If text is present, returns the joined text.
+ * - If not, returns a label for the first non-text modality (e.g., 'Image', 'Other').
+ * - If unknown, returns 'Multimodal message'.
+ */
 export function getContentString(content: Message["content"]): string {
  if (typeof content === "string") return content;
  const texts = content
--- a/src/components/ui/MultimodalPreview.tsx
+++ b/src/components/ui/MultimodalPreview.tsx
@@ -0,0 +1,129 @@
+import React from "react";
+import { File, Image as ImageIcon, X as XIcon } from "lucide-react";
+import type { Base64ContentBlock } from "@langchain/core/messages";
+import { cn } from "@/lib/utils";
+import Image from "next/image";
+export interface MultimodalPreviewProps {
+  block: Base64ContentBlock;
+  removable?: boolean;
+  onRemove?: () => void;
+  className?: string;
+  size?: "sm" | "md" | "lg";
+}
+
+export const MultimodalPreview: React.FC<MultimodalPreviewProps> = ({
+  block,
+  removable = false,
+  onRemove,
+  className,
+  size = "md",
+}) => {
+  // Sizing
+  const sizeMap = {
+    sm: "h-10 w-10 text-base",
+    md: "h-16 w-16 text-lg",
+    lg: "h-24 w-24 text-xl",
+  };
+  const iconSize: string =
+    typeof sizeMap[size] === "string" ? sizeMap[size] : sizeMap["md"];
+
+  // Image block
+  if (
+    block.type === "image" &&
+    block.source_type === "base64" &&
+    typeof block.mime_type === "string" &&
+    block.mime_type.startsWith("image/")
+  ) {
+    const url = `data:${block.mime_type};base64,${block.data}`;
+    let imgClass: string = "rounded-md object-cover h-16 w-16 text-lg";
+    if (size === "sm") imgClass = "rounded-md object-cover h-10 w-10 text-base";
+    if (size === "lg") imgClass = "rounded-md object-cover h-24 w-24 text-xl";
+    return (
+      <div className={cn("relative inline-block", className)}>
+        <Image
+          src={url}
+          alt={String(block.metadata?.name || "uploaded image")}
+          className={imgClass}
+          width={size === "sm" ? 16 : size === "md" ? 32 : 48}
+          height={size === "sm" ? 16 : size === "md" ? 32 : 48}
+        />
+        {removable && (
+          <button
+            type="button"
+            className="absolute top-1 right-1 z-10 rounded-full bg-gray-500 text-white hover:bg-gray-700"
+            onClick={onRemove}
+            aria-label="Remove image"
+          >
+            <XIcon className="h-4 w-4" />
+          </button>
+        )}
+      </div>
+    );
+  }
+
+  // PDF block
+  if (
+    block.type === "file" &&
+    block.source_type === "base64" &&
+    block.mime_type === "application/pdf"
+  ) {
+    const filename =
+      block.metadata?.filename || block.metadata?.name || "PDF file";
+    return (
+      <div
+        className={cn(
+          "relative flex items-center gap-2 rounded-md border bg-gray-100 px-3 py-2",
+          className,
+        )}
+      >
+        <File
+          className={cn(
+            "flex-shrink-0 text-teal-700",
+            size === "sm" ? "h-5 w-5" : "h-7 w-7",
+          )}
+        />
+        <span
+          className={cn(
+            "truncate text-sm text-gray-800",
+            size === "sm" ? "max-w-[80px]" : "max-w-[160px]",
+          )}
+        >
+          {String(filename)}
+        </span>
+        {removable && (
+          <button
+            type="button"
+            className="ml-2 rounded-full bg-gray-200 p-1 text-teal-700 hover:bg-gray-300"
+            onClick={onRemove}
+            aria-label="Remove PDF"
+          >
+            <XIcon className="h-4 w-4" />
+          </button>
+        )}
+      </div>
+    );
+  }
+
+  // Fallback for unknown types
+  return (
+    <div
+      className={cn(
+        "flex items-center gap-2 rounded-md border bg-gray-100 px-3 py-2 text-gray-500",
+        className,
+      )}
+    >
+      <File className="h-5 w-5 flex-shrink-0" />
+      <span className="truncate text-xs">Unsupported file type</span>
+      {removable && (
+        <button
+          type="button"
+          className="ml-2 rounded-full bg-gray-200 p-1 text-gray-500 hover:bg-gray-300"
+          onClick={onRemove}
+          aria-label="Remove file"
+        >
+          <XIcon className="h-4 w-4" />
+        </button>
+      )}
+    </div>
+  );
+};
--- a/src/hooks/use-file-upload.tsx
+++ b/src/hooks/use-file-upload.tsx
@@ -0,0 +1,164 @@
+import { useState, useRef, useEffect, ChangeEvent } from "react";
+import { toast } from "sonner";
+import type { Base64ContentBlock } from "@langchain/core/messages";
+import { fileToContentBlock } from "@/lib/multimodal-utils";
+
+export const SUPPORTED_FILE_TYPES = [
+  "image/jpeg",
+  "image/png",
+  "image/gif",
+  "image/webp",
+  "application/pdf",
+];
+
+interface UseFileUploadOptions {
+  initialBlocks?: Base64ContentBlock[];
+}
+
+export function useFileUpload({
+  initialBlocks = [],
+}: UseFileUploadOptions = {}) {
+  const [contentBlocks, setContentBlocks] =
+    useState<Base64ContentBlock[]>(initialBlocks);
+  const dropRef = useRef<HTMLDivElement>(null);
+
+  const isDuplicate = (file: File, blocks: Base64ContentBlock[]) => {
+    if (file.type === "application/pdf") {
+      return blocks.some(
+        (b) =>
+          b.type === "file" &&
+          b.mime_type === "application/pdf" &&
+          b.metadata?.filename === file.name,
+      );
+    }
+    if (SUPPORTED_FILE_TYPES.includes(file.type)) {
+      return blocks.some(
+        (b) =>
+          b.type === "image" &&
+          b.metadata?.name === file.name &&
+          b.mime_type === file.type,
+      );
+    }
+    return false;
+  };
+
+  const handleFileUpload = async (e: ChangeEvent<HTMLInputElement>) => {
+    const files = e.target.files;
+    if (!files) return;
+    const fileArray = Array.from(files);
+    const validFiles = fileArray.filter((file) =>
+      SUPPORTED_FILE_TYPES.includes(file.type),
+    );
+    const invalidFiles = fileArray.filter(
+      (file) => !SUPPORTED_FILE_TYPES.includes(file.type),
+    );
+    const duplicateFiles = validFiles.filter((file) =>
+      isDuplicate(file, contentBlocks),
+    );
+    const uniqueFiles = validFiles.filter(
+      (file) => !isDuplicate(file, contentBlocks),
+    );
+
+    if (invalidFiles.length > 0) {
+      toast.error(
+        "You have uploaded invalid file type. Please upload a JPEG, PNG, GIF, WEBP image or a PDF.",
+      );
+    }
+    if (duplicateFiles.length > 0) {
+      toast.error(
+        `Duplicate file(s) detected: ${duplicateFiles.map((f) => f.name).join(", ")}. Each file can only be uploaded once per message.`,
+      );
+    }
+
+    const newBlocks = uniqueFiles.length
+      ? await Promise.all(uniqueFiles.map(fileToContentBlock))
+      : [];
+    setContentBlocks((prev) => [...prev, ...newBlocks]);
+    e.target.value = "";
+  };
+
+  // Drag and drop handlers
+  useEffect(() => {
+    if (!dropRef.current) return;
+
+    const handleDragOver = (e: DragEvent) => {
+      e.preventDefault();
+      e.stopPropagation();
+    };
+
+    const handleDrop = async (e: DragEvent) => {
+      e.preventDefault();
+      e.stopPropagation();
+
+      if (!e.dataTransfer) return;
+
+      const files = Array.from(e.dataTransfer.files);
+      const validFiles = files.filter((file) =>
+        SUPPORTED_FILE_TYPES.includes(file.type),
+      );
+      const invalidFiles = files.filter(
+        (file) => !SUPPORTED_FILE_TYPES.includes(file.type),
+      );
+      const duplicateFiles = validFiles.filter((file) =>
+        isDuplicate(file, contentBlocks),
+      );
+      const uniqueFiles = validFiles.filter(
+        (file) => !isDuplicate(file, contentBlocks),
+      );
+
+      if (invalidFiles.length > 0) {
+        toast.error(
+          "You have uploaded invalid file type. Please upload a JPEG, PNG, GIF, WEBP image or a PDF.",
+        );
+      }
+      if (duplicateFiles.length > 0) {
+        toast.error(
+          `Duplicate file(s) detected: ${duplicateFiles.map((f) => f.name).join(", ")}. Each file can only be uploaded once per message.`,
+        );
+      }
+
+      const newBlocks = uniqueFiles.length
+        ? await Promise.all(uniqueFiles.map(fileToContentBlock))
+        : [];
+      setContentBlocks((prev) => [...prev, ...newBlocks]);
+    };
+
+    const handleDragEnter = (e: DragEvent) => {
+      e.preventDefault();
+      e.stopPropagation();
+    };
+
+    const handleDragLeave = (e: DragEvent) => {
+      e.preventDefault();
+      e.stopPropagation();
+    };
+
+    const element = dropRef.current;
+    element.addEventListener("dragover", handleDragOver);
+    element.addEventListener("drop", handleDrop);
+    element.addEventListener("dragenter", handleDragEnter);
+    element.addEventListener("dragleave", handleDragLeave);
+
+    return () => {
+      element.removeEventListener("dragover", handleDragOver);
+      element.removeEventListener("drop", handleDrop);
+      element.removeEventListener("dragenter", handleDragEnter);
+      element.removeEventListener("dragleave", handleDragLeave);
+    };
+  }, [contentBlocks]);
+
+  const removeBlock = (idx: number) => {
+    setContentBlocks((prev) => prev.filter((_, i) => i !== idx));
+  };
+
+  const resetBlocks = () => setContentBlocks([]);
+
+  return {
+    contentBlocks,
+    setContentBlocks,
+    handleFileUpload,
+    dropRef,
+    removeBlock,
+    resetBlocks,
+  };
+}
--- a/src/lib/multimodal-utils.ts
+++ b/src/lib/multimodal-utils.ts
@@ -0,0 +1,57 @@
+import type { Base64ContentBlock } from "@langchain/core/messages";
+import { toast } from "sonner";
+
+// Returns a Promise of a typed multimodal block for images or PDFs
+export async function fileToContentBlock(
+  file: File,
+): Promise<Base64ContentBlock> {
+  const supportedImageTypes = [
+    "image/jpeg",
+    "image/png",
+    "image/gif",
+    "image/webp",
+  ];
+  const supportedFileTypes = [...supportedImageTypes, "application/pdf"];
+
+  if (!supportedFileTypes.includes(file.type)) {
+    toast.error(
+      `Unsupported file type: ${file.type}. Supported types are: ${supportedFileTypes.join(", ")}`,
+    );
+    return Promise.reject(new Error(`Unsupported file type: ${file.type}`));
+  }
+
+  const data = await fileToBase64(file);
+
+  if (supportedImageTypes.includes(file.type)) {
+    return {
+      type: "image",
+      source_type: "base64",
+      mime_type: file.type,
+      data,
+      metadata: { name: file.name },
+    };
+  }
+
+  // PDF
+  return {
+    type: "file",
+    source_type: "base64",
+    mime_type: "application/pdf",
+    data,
+    metadata: { filename: file.name },
+  };
+}
+
+// Helper to convert File to base64 string
+export async function fileToBase64(file: File): Promise<string> {
+  return new Promise<string>((resolve, reject) => {
+    const reader = new FileReader();
+    reader.onloadend = () => {
+      const result = reader.result as string;
+      // Remove the data:...;base64, prefix
+      resolve(result.split(",")[1]);
+    };
+    reader.onerror = reject;
+    reader.readAsDataURL(file);
+  });
+}