AI Tools
Building on top of language models brings its own set of chores: preparing training data in the exact JSONL shape a provider expects, keeping prompts consistent across environments, and keeping an eye on prompt length.
3 free tools
AI Tools
About these ai tools
These tools do not call any AI model. They are deterministic helpers that run locally, so you can check a dataset that contains private examples or draft a prompt that includes internal details without sending it anywhere.
Why use them
Dataset sanity checks
Find malformed lines, missing roles and empty messages in chat-format JSONL before you upload a fine-tuning job.
Reusable prompts
Detect {{variables}} in a template and fill them from a form or JSON object.
Honest estimates
Token counts are labelled as estimates — exact numbers depend on each model's tokenizer.
Frequently asked questions
Do these tools send my prompts to an AI service?
No. None of the tools in this category call a model or any external API. All processing is local to your browser.
Why is the token count an estimate?
Every model family uses its own tokenizer, and the same text can produce different token counts. The estimator uses documented character and word heuristics that are typically close for English text, but you should rely on your provider's tokenizer for billing-critical calculations.