Reference
Supported data sources
What you can connect today, what's coming, and the limits of each.
Files
| Format | Status | Notes |
|---|---|---|
| CSV | ✅ Supported | UTF-8 preferred. Auto-detects delimiter, quoting, and headers. |
| Excel (.xlsx) | ✅ Supported | Multiple sheets supported; each sheet becomes a separate source. |
| TSV | ✅ Supported | Treated as a CSV variant. |
| JSON / NDJSON | ⏳ Coming | Roadmap. |
| Parquet | ⏳ Coming | Roadmap; large-data workflows recommended via BigQuery for now. |
File limits
- Individual files: up to ~500 MB practical limit. Larger files work via BigQuery.
- Multiple files in one upload: supported; they upload in parallel.
- Versions per file: unlimited; the platform stores the diff between versions.
Tips for clean ingestion
- Single header row at the top.
- One row per "thing" — orders flat, customers flat, etc. Avoid pivoted summaries.
- Use ISO date format (
YYYY-MM-DD) where you can. - Real column names, not abbreviations only you understand.
- Don't pre-aggregate — give the platform raw rows.
Warehouses
| Source | Status | Notes |
|---|---|---|
| BigQuery | ✅ Supported | Direct connection. Queries run live; no data downloaded. |
| Snowflake | ⏳ Coming | Roadmap. |
| Postgres / Redshift | ⏳ Coming | Roadmap. |
BigQuery specifics
- Service-account authentication.
- The platform reads schema and a representative sample row-set for analysis.
- Production queries run against your live warehouse — you pay your normal BigQuery costs.
- Row-level security in BigQuery is respected.
Documents
| Type | Status | Notes |
|---|---|---|
| Google Drive | ✅ Supported | Connect a folder; the platform reads documents into the ontology. |
| ✅ Supported | Text-extractable PDFs work best; scanned PDFs require OCR (auto-applied). | |
| DOCX | ✅ Supported | Extracted to text. |
| Markdown / TXT | ✅ Supported | Direct text. |
| Notion | ⏳ Coming | Roadmap. |
Audio
| Type | Status | Notes |
|---|---|---|
| Call recordings (.mp3 / .wav / .m4a) | ✅ Supported | Auto-transcribed; transcript becomes the source. |
| Meeting recordings | ✅ Supported | Same path as call recordings. |
App connectors
The Apps tab lists third-party services you can connect. The catalogue grows continuously. If you need a connector that isn't there, ask your admin — many connectors are easy to add via the platform's tool-extension surface.
Live ingestion (webhooks)
For systems that can push data, the platform exposes a signed HTTP endpoint. Your admin sets up the integration once and pushed payloads become source updates automatically.
What the platform does NOT support
- Direct connections to authenticated SaaS APIs without a connector (use Drive/file uploads, or ask for the connector).
- Streaming sources (Kafka, Pub/Sub) — push via the webhook endpoint instead.
- Database write-back — the platform reads, it doesn't write to your sources.
See also
- Connect data sources — the how-to.
- Sources concept — what happens after a source is connected.