Skip to content

🧠 Datasets ​

Datasets are the foundation of your AI model training process.
Each dataset consists of question–answer pairs that help the model learn how to respond correctly.
You can create datasets manually or automatically from uploaded files.


📋 Overview ​

Each dataset contains:

  • A Question (example user query)
  • A Good Answer (ideal AI response)
  • An optional Bad Answer (used for DPO fine-tuning)

Datasets can come from:

  1. Files you uploaded in the Training Files section (PDFs or images)
  2. Manual entries you create yourself

🧩 Creating Datasets from Files ​

If you’ve already uploaded files (PDFs or images) in the Training Files section and they’ve been successfully parsed, you can automatically create datasets from them.

Steps ​

  1. Go to Datasets.
  2. Click Create Dataset on the top-right corner.
  3. In the Select dropdown, choose Create From File.
  4. Pick one of your parsed files from the list.
  5. Click Create From File — Ainisa will start generating question–answer pairs from that file.

This process may take some time depending on file size.
Once parsing completes, the datasets will automatically appear in your list.

Create dataset from file


✍️ Creating Datasets Manually ​

If you prefer to create your own question–answer pairs manually, follow these steps:

Steps ​

  1. Go to Datasets.
  2. Click Create Dataset.
  3. In the Select dropdown, choose Create Manually.
  4. Fill in the fields:
    • Question — enter your example question.
    • Good Answer to question — ideal response from AI.
    • Bad Answer to question — optional; required only for DPO fine-tuning.
  5. Click Save to store your dataset.

Create dataset manually


🧾 Viewing Datasets ​

After creating datasets (manually or from files), you’ll see them listed in the Datasets table:

  • Question — the text of your training prompt.
  • Dataset File ID — the source file name (if it was created from a file).
  • Datasets created manually will have an empty Dataset File ID column.
  • You can manage or delete datasets anytime.

Datasets list


🧠 Notes ​

  • A minimum of 10 datasets is required to fine-tune a model.
  • Datasets created from files will show the file name (e.g. agentic-rag.jpeg or partnership.pdf).
  • Manually created datasets will not show a file name.
  • All created datasets are securely stored and linked to your business account.
  • Each dataset counts toward your monthly training data limit (based on your subscription plan).

đź’ˇ Plan limits:

  • Each subscription plan on Ainisa.com has its own dataset limits.
  • In the Free package, you can create up to 20 datasets.
  • Creating datasets from files (PDFs or images) is available starting from the Pro plan.
  • For more information, visit the Pricing page.

âś… Tip: For more details about fine-tuning methods, check out: