Browse and Preview Source Data
Browse and connector preview are separate, connector-specific capabilities. Check the connector capability matrix before using this workflow. When preview is unsupported, validate data through a bounded poll read, an inbound subscription sample, or the file-import workflow instead.
Current browse and connector-preview support
| Connector | Browse | Connector preview |
|---|---|---|
| OPC UA | Node hierarchy. | Not implemented. |
| BACnet | Discovered devices and configured object-type groups. | Not implemented. |
| Modbus | Configured register map. | Not implemented. |
| MQTT | Not supported; configure topics manually. | Not implemented. |
| REST | Not supported; configure JSONPath fields manually. | Not implemented. |
| CSV / Excel | Not supported; inspect the imported file. | Not implemented. |
| JDBC | Tables, views, and columns from JDBC metadata. | Not implemented. |
| Fabric — OneLake | OneLake file and folder paths. | Not implemented; OneLake does not decode rows. |
| Fabric — SQL endpoint | Configured table and allowlisted columns. | Up to 25 tabular rows. |
Prerequisites
Use a supported browse or preview operation only after:
- the connector has been saved;
- test connection succeeds or the known failure is being investigated;
- source owner can confirm tag, topic, object, table, file, or field meaning;
- expected units, timestamp behavior, and update interval are available;
- the operator knows which downstream workflow needs the data.
When to use this page
Use this workflow when:
- a connector test succeeds;
- you need to find the correct tags, fields, topics, objects, tables, or files;
- the source owner needs to confirm units and meaning;
- mappings are being prepared for Inspector, AI Agent, predictive maintenance, or BI.
Browse workflow
- Open
Data Integration > Connectors. - Open the connector.
- Start the connector; connector browse and preview require an active instance.
- Open source browse only if the matrix marks it supported.
- Navigate through source hierarchy.
- Select candidate nodes or fields, or record the manually configured source path.
- Record source paths that should be mapped.
What "source hierarchy" means depends on connector type:
| Source type | Supported discovery result |
|---|---|
| OPC UA | Namespace, objects, variables, nodes. |
| BACnet | Discovered devices and configured object-type groups. |
| Modbus | Configured register map. |
| JDBC | Tables, views, and columns. |
| Fabric — OneLake | File and folder paths. |
| Fabric — SQL endpoint | Configured table and allowlisted columns. |
Preview workflow
- Confirm that connector preview is supported. Today this means Fabric SQL endpoint mode.
- Select an allowlisted SQL endpoint column and run preview.
- Check recent sample values.
- Confirm timestamp.
- Confirm data type.
- Confirm unit.
- Confirm expected range.
- Mark the field ready, uncertain, or rejected.
What to record
For every candidate field, record:
| Item | Why it matters |
|---|---|
| Source path | Needed for mapping. |
| Display name | Helps reviewers recognize the field. |
| Source owner | Needed when values look wrong. |
| Timestamp field | Required for time-based workflows. |
| Unit | Needed for conversion and quality checks. |
| Expected range | Helps detect impossible values. |
| Target asset or point | Needed for operational context. |
| Review state | Ready, uncertain, or rejected. |
Expected output
The output of discovery and value validation is not a production mapping. It is a reviewed source-field list that includes source path, sample value when available, timestamp behavior, unit, target context, and review state. Keep uncertain fields out of mappings until the source owner confirms them.
Preview review checklist
Approve a source field for mapping only when:
- the value is current enough for the workflow;
- timestamp is clear;
- equipment or asset identity is clear;
- unit is known;
- value range is plausible;
- source owner can explain the field;
- the same field is not duplicated under a better source path.
Common problems
| Symptom | Likely cause | Action |
|---|---|---|
| Browse fails | Connector is not running, source is unreachable, or permission is missing. | Start connector and recheck credentials/network. |
| Preview is unavailable | The connector does not implement connector preview, or Fabric is in OneLake mode. | Use the capability matrix and validate with poll read, subscription, file import, or Fabric SQL endpoint mode. |
| Supported preview is empty | The Fabric SQL path has no rows or its selection is too narrow. | Ask the source owner and check table, column allowlist, and query window. |
| Timestamp is missing | Source payload has no usable time field. | Add timestamp mapping or choose another field. |
| Values are stale | Upstream source stopped updating or interval is too long. | Check source system and sync strategy. |
| Values look impossible | Unit, scale factor, byte order, or field selection may be wrong. | Review connector configuration and mapping transform. |
Next step
After the supported validation path confirms the right source field, continue with Mapping Source Fields.