Hyperparam opens multi-gigabyte Parquet, JSONL, and CSV straight from S3, Hugging Face, or your laptop. Search by keyword, meaning, or SQL, then run a model over the rows that match. Nothing to install: drop a file on this page to preview it, or sign in for the full app.
Explore and transform training, eval, and log data at full size. Add AI-generated columns, filter, and export what you keep.
Open-source core: hyparquet · icebird · hightable · squirreling. The full app runs here when you sign in.
Records every agent session, finds what's breaking, and hands your agent the fix.
npm i -g hypawarehyp setupInstall HypAware using npm i -g hypaware, then run hyp setup --guide and follow its instructions.
Works with coding agents · agent harnesses · software factories · local models · MCP clients
For decades the data we stored was rows and columns you analyze by aggregation, and a whole industry grew up around it: databases, warehouses, dashboards. AI data is not that. Open an agent log or an LLM training set and you find a couple of structured columns and then one enormous text column where every cell is a whole conversation, a reasoning trace, a source file, and there are millions of them.
SQL can count the rows. It cannot read them. Understanding AI data needs infrastructure that both scales with the volume and reads the text the way a person would. That is what Hyperparam built, and both products run on it.
Search by keyword, by meaning, or with SQL, then run a language model over the rows that match. In Hyperparam it all runs in your browser, straight against the files in your bucket, reading only the bytes each query needs.
2.3M traces, loaded in under a minute, then filtered, clustered, and annotated with AI assistance.