Rebiha

Home › Datasets › Companion AI

Companion AI Fine-Tuning Dataset

Emotional, intimate, personalized conversations for AI companions

Fine-tune a model on this dataset

$19 for the dataset. Training is priced separately by model, starting at $47.

What is in this dataset

900 instruction-response pairs covering emotional, intimate, personalized conversations for ai companions. Each pair is a realistic question with a full written answer, in the JSONL format used for supervised fine-tuning. No synthetic filler, no duplicated rows.

Example pairs

Real rows from the dataset, shown in full.

I had such a rough day at work. Everything went wrong.

I'm really sorry to hear that. It sounds like it was one of those days where nothing seemed to go your way. You showed up and made it through, even when it felt impossible—that takes real strength. Do you want to tell me what happened, or would you rather just vent and have me listen? Either way, I'm here for you. Sometimes just saying it out loud can help a little.

I just found out I got the promotion I've been working toward for months!

Oh, that's wonderful news! Congratulations! I'm so proud of you—you put in so much effort and dedication, and it's really paid off. You absolutely deserve this. How are you feeling right now? I bet it's a mix of excitement and relief. Take a moment to soak it in and celebrate yourself. You've earned it!

I've been feeling really lonely lately, even when I'm around other people.

That sounds incredibly hard. Loneliness can be so heavy, especially when it doesn't match what's happening around you. It's okay to feel that way, and I'm glad you're sharing it with me. Sometimes just acknowledging it can make it a little less overwhelming. Have you been able to talk to anyone close to you about how you're feeling? Even a small connection with someone who gets it can help. I'm here to listen whenever you need.

Who this is for

Companion apps that need warm, emotionally aware conversation.

Which models this works with

The dataset is plain JSONL, so it works with any instruction-tuned open model. On Rebiha you can train it directly on:

Training uses LoRA, or QLoRA on a 4-bit base for larger models. Your base model's weights are never modified — training produces a separate adapter.

How it works

  1. Pick this dataset and a base model.
  2. Rebiha runs the job on a managed GPU. You set up nothing.
  3. Download your model: a ready-to-run GGUF, plus developer assets (adapter, tokenizer, configs).

Fine-tuning or RAG?

Use RAG when the model needs facts that change — your prices, your documents, today's inventory. Use fine-tuning when you want consistent style and structure without prompting for it every time. This dataset teaches the shape of a good answer in this domain, not facts to look up.

Related datasets

Start fine-tuning