Settings & ConfigurationExports
Exports
Get your AI Evals data out — on-demand batched exports from any data table, or scheduled incremental exports to your own cloud storage.
Exports
AI Evals offers two ways to get your trace, observation, score, and experiment data out of the platform for offline analysis, archival, fine-tuning, and warehousing.
Batched Exports
On-demand CSV / JSON / JSONL downloads from any data table, emailed to you.
Cloud Storage Exports
Scheduled, incremental exports into an AWS S3 bucket you own.
Which export should I use?
| Batched Exports | Cloud Storage Exports | |
|---|---|---|
| Trigger | On demand — click Export on a data table | Automatic — runs on a schedule you set |
| Scope | A single table view, optionally filtered | A data type (traces, observations, or scores), optionally filtered |
| Coverage | A point-in-time snapshot (up to 10,000 rows) | Incremental — only rows added since the previous run |
| Destination | Download link delivered by email | Your own AWS S3 bucket |
| Formats | CSV, JSON, JSONL | JSON Lines (gzip), CSV, JSON |
| Best for | Ad-hoc analysis, audits, one-off datasets | Continuous warehousing, analytics pipelines, large-scale fine-tuning |
| Where configured | Any data table | Settings → Exports |
Use Batched Exports when you need a quick, filtered snapshot for analysis in Excel/Pandas or a one-off fine-tuning dataset. Use Cloud Storage Exports when you want your data to land in your own infrastructure automatically and continuously.