AI File Agent

Analyze API and file data with natural language and a nice UI.

Features
Development
Launching-client
Debugging
Contributing

Features

AI File Agent is not only another AI chat UI, it allows you to upload any file and create custom components to render with each message, for example:

messageContext.updateMessage({
    role: "assistant",
    content: `File "${file.name}" uploaded successfully. What would you like to do with it?`,
    beforeContentComponent: (
        <Typography.Description>
        NOTE: This file is assigned to a temporary account.{" "}
        <Typography.Link href="#">Create an account</Typography.Link> to keep it.
        </Typography.Description>
    ),
    afterContentComponent: <MessageFileType.Options file={file} fieldName={FormFieldNames.message} />,
    type: "file",
    file,
    id: messageContext.transformId(file.upload!.uuid),
});

Then, in the MessageFileType.tsx component:

export const MessageFileType = ({ message, className }: MessageFileTypeProps) => {
  const isSimulationEnabled = message.role === "assistant" && !message.hasInnerHtml;

  const { simulationEnded } = useTypingSimulation(message.content, isSimulationEnabled, `#${message.id}`);

  const progress: number = useSubscription(0, message.file.progressObservable);

  return (
    <div className={clsx(styles["message-file-type"], className)}>
      <div>
        <div className={styles["message-file-type__avatar"]}>
          <div className={styles["message-file-type__avatar-box"]}>
            {progress === 100 ? (
              <Icon name="icon-file-check" />
            ) : (
              <CircularProgress color="#ffd74b" percentage={progress} fontSize="21px" />
            )}
          </div>
        </div>
        <div className={styles["message-file-type__content"]}>
          {message.beforeContentComponent && simulationEnded && message.beforeContentComponent}

          {!isSimulationEnabled ? (
            <Typography.Text>{message.content}</Typography.Text>
          ) : (
            <Typography.Text id={message.id} />
          )}

          {message.afterContentComponent && simulationEnded && message.afterContentComponent}
        </div>
      </div>
    </div>
  );
};

Plus

✅ Upload any file of any size, display a nice uploading animation all within the chat interface
✅ Standard GoogleAI and OpenAI choice responses. Should work with more LM's in the future
✅ Animated assistant replies (typing animation)
✅ Variable height textarea
✅ API authentication handler
✅ function_call message type handler to call server-side functions
✅ Custom label values to render TSX components upon a given AI reply
✅ Well-typed Typescript interfaces to keep your dev flow groovin'
✅ Mobile-first design
✅ Dark and Light theming
✅ i18n ready
✅ Optional Tailwind CSS components and optional Shadcn UI components
✅ Open-source with paid bounties! 🤑

Development

Follow app/README.md for a detailed explanation of launching the dev environment.

Launching the frontend client

The client is a NextJS application.

To launch on localhost:3003, first clone this repo and run:

git@github.com:aufacicenta/fileagent.git
cd fileagent
yarn
cd app
yarn
yarn dev:debug

You'll need these values in app/.env:

export NODE_ENV=test
export NEXT_PUBLIC_ORIGIN="http://localhost:3003"

export NEXT_PUBLIC_CHAT_AI_API="googleai" # "googleai" OR "openai"

# NANONETS, for OCR features: https://nanonets.com/
export NANONETS_API_KEY="..."

# OPEANAI, if you switch NEXT_PUBLIC_CHAT_AI_API to "openai"
export OPENAI_API_KEY="..."

# SUPABASE, used to store file content after it's read once: https://supabase.com/
export NEXT_PUBLIC_SUPABASE_URL="..."
export NEXT_PUBLIC_SUPABASE_ANON_KEY="..."

# DROPBOX, used for the Dropbox Sign API features: https://developers.hellosign.com/api/reference/signature-request/
export DROPBOX_CLIENT_ID="..."
export DROPBOX_CLIENT_SECRET="..."
export DROPBOX_REDIRECT_URI="..."

# SQUARE, used for the Square API features: https://developer.squareup.com/explorer/square
export SQUARE_APP_ID="..."
export SQUARE_APP_SECRET="..."
export SQUARE_ACCESS_TOKEN="..."
export SQUARE_OAUTH_ENDPOINT="..."

# GOOGLE, used for googleai authentication: https://cloud.google.com/docs/authentication/provide-credentials-adc
export GOOGLE_APPLICATION_CREDENTIALS="path to credentials"
export GOOGLE_PROJECT_ID="..."

# DATABASE, used for the Supabase postgres instance
export POSTGRES_DB_NAME=postgres
export POSTGRES_DB_USERNAME=postgres
export POSTGRES_DB_PASSWORD=...
export POSTGRES_DB_HOST=...
export POSTGRES_DB_PORT=5432

Optional VSCode debugging

Instead of yarn dev:debug, hit the F5 key, it should launch the VSCode Debugger in a new terminal session. You may create debug breakpoints on API endpoints, for example.

Contributing

Check the paid issues in the AI File Agent project board!

Udit-takkar / pulsemarkets