osFoundry / Platform
Bring your own datasets
Upload CSVs, JSONLs or stream from your warehouse. Same indexing pipeline as built-in data.
What it is
"Bring your own datasets" lets you feed osFoundry the data you already have. Drop in a CSV or JSONL file, or point it at your data warehouse and stream rows in. Whichever way the data arrives, it lands in the same place your other knowledge lives, ready for chat, agents, and apps to query and reason over.\n\nThe key idea is parity: your uploaded and streamed data goes through the exact same indexing pipeline as osFoundry's built-in data. There is no second-class path for "custom" data and no separate connector you have to babysit. A CSV of your customers, a JSONL export of support tickets, and a table streamed from your warehouse are all indexed, chunked, and made retrievable the same way, so a question that spans your data and osFoundry's data just works. Because osFoundry is a local-first desktop workspace, this happens on your machine rather than in someone else's multi-tenant store, and your dataset does not have to leave your device to be usable.\n\nThis matters most when your useful data is scattered and slightly awkward. Real data is a folder of CSV exports, a JSONL dump from an internal tool, and a warehouse table that changes every night. Rather than reshaping all of that to fit a rigid schema, you bring it as-is and let the same pipeline handle indexing. Combined with osFoundry's BYOK model, you also stay in control of which model reads your data and where it runs, whether that is a local model, a cloud API on your own key with no reseller markup, or a self-hosted endpoint.
Key capabilities
- Upload CSV and JSONL files directly into your workspace with no schema migration or preprocessing step required
- Stream rows continuously from your data warehouse so indexed data tracks the source instead of going stale after a one-time export
- Runs the identical indexing pipeline used for osFoundry's built-in data, so your data is a first-class citizen for retrieval
- Query your uploaded and streamed data from chat, agents, and apps in the same workspace, alongside built-in data
- Local-first and BYOK: index on your own device and choose which local, cloud, or self-hosted model reads the data
Use cases
- Ops analyst: Uploads a nightly JSONL export of support tickets and asks agents to summarize recurring issues without moving the data to a third-party tool
- Data engineer: Streams a warehouse table into the workspace so downstream agents always query current rows instead of a stale CSV snapshot
- Privacy-conscious founder: Brings a customer CSV that must stay on-device and indexes it locally, keeping sensitive records off any vendor's servers
Frequently asked questions
What file formats can I upload as a dataset?
CSV and JSONL files are supported directly. You can also stream data in from your warehouse instead of uploading a static file, which is the better choice when the source changes frequently and you want the indexed copy to stay current.
Is my custom data treated differently from osFoundry's built-in data?
No. Your uploaded and streamed datasets run through the same indexing pipeline as built-in data, so they are indexed and made retrievable the same way. A query can span your data and built-in data without any special handling.
Does my data leave my device?
osFoundry is a local-first desktop workspace, so your dataset can be indexed and stay on your own machine. Because it is bring-your-own-key, you also control which model reads the data and whether it runs locally, in the cloud on your own key, or on a self-hosted endpoint.