Qwisine 14B

Model details

Intended use & limitations

• Conversational assistant for developers building on Convex.
• Drafting / Helping with convex orientated questions & tasks.
• Documentation chatbots or support assistants.

• Production‑critical decision making without human review.

Dataset

• Size  : 938 Q&A pairs
• Source: Convex official docs, example apps, public issues, community Discord, and synthetic edge‑cases.
• Question types (distilled)
  • what_is   – factual look‑ups (no reasoning)
  • why       – causal explanations (no reasoning)
  • task      – recipe‑style how‑to (with reasoning)
  • edge_case – tricky or undocumented scenarios (with reasoning)
  • v‑task    – verbose multi‑step tasks (with reasoning)

Training procedure -- will add later since i ran & experimented MANY RUNS 😭😭😭😭

• Epochs  : **
• Batch   : **
• LR / schedule : **
• Optimizer : **

Evaluation results

Think Mode

Non-Think Mode

How to run locally

Limitations & biases

• Training data is entirely Convex‑centred; the model may hallucinate.
• The dataset size is modest (938 samples); edge‑case coverage is still incomplete and so is more complex prompts like create project from scratch with multiple steps and instructions.

Future work

Citation

Acknowledgements