Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.ajayji.com/llms.txt

Use this file to discover all available pages before exploring further.

For power users who want to go a step further, aj provides a unique workflow to create your own custom AI models, fine-tuned on your own data. This feature acts as an orchestrator, guiding you through a professional-grade training process while leveraging powerful external tools. The Goal: To create a specialized model (in .gguf format) that understands your unique terminology, data, and needs, which you can then use locally and privately within the application. How it Works: The app guides you through a persistent, step-by-step process:
  1. Connect Your Tools: The workflow starts by connecting to your Google Drive and your Label Studio instance, a popular open-source data labeling tool.
  2. Prepare Your Data: aj helps you configure a “fine-tuning run” by selecting a base model to start from and the dataset you’ve prepared in Label Studio.
  3. Orchestrate the Job: The app then prepares and organizes all the necessary files in your Google Drive, including your dataset and a specialized training notebook.
  4. Train in Google Colab: You’ll be directed to open the notebook in Google Colab, a powerful environment that provides free access to GPUs for the computationally-intensive training process.
  5. Download Your Custom Model: Once training is complete, the app helps you download the resulting model file directly from your Google Drive and registers it within aj, making it immediately available for use in your workflows.
Because this can be a long process, every step is saved in a “Fine-Tuning Run.” You can manage multiple training jobs at once and resume a run at any time, picking up exactly where you left off. This advanced feature puts the power of custom AI directly in your hands.