Low-Latency Matching External Records

Match tens to hundreds of incoming or external records at low latency (sub-second) speeds to existing records or clusters in a Mastering project without committing them to Tamr.

Checklist before proceeding

• At least one mastering project exists (Creating a Project ).
• At least one dataset is added to the project and is schema mapped to the project's unified dataset (Adding a Dataset ).
• Generate Record Pairs and the Update Results jobs have both been executed OR a mastering model has been imported (Importing a Mastering Model).

📘

Updating

To make the latest results of a Mastering project available for low-latency matching you must Update low-latency match.

The results of a Mastering project are typically changed when

• the unified dataset is updated.
• the mastering model is updated.
• the record pairs or clusters are updated.

Using Low-Latency Match with Published Clusters

To use low-latency match with published clusters:

1. Update low-latency match as needed (Updating Low-Latency Match with Published Clusters), see Updating scenarios above.
2. Match incoming external records at low latency (Low-Latency Matching).

Preparing Low-Latency Match with Published Clusters

To prepare a Mastering project to use the low-latency match service with published clusters:

1. Wait For Queryable: GET /match/v1/projects/{project}:isQueryable
   Poll the status of the update job submitted in Step 1 until the response true is received.
2. Publish Clusters: POST /v1/projects/{project}/publishedClusters:refresh
   Publish clusters to update the Mastering project's data queryable for low-latency matching.

Updating Low-Latency Match with Published Clusters

To update a Mastering project's data queryable for low-latency matching:

1. Publish Clusters: POST /v1/projects/{project}/publishedClusters:refresh
   Publish clusters to update the Mastering project's data queryable for low-latency matching.

Using Low-Latency Match Service without Published Clusters

To use low-latency match without published clusters:

1. Match incoming external records at low latency (Low-Latency Matching).

Updating Low-Latency Match without Published Clusters

1. Update the Low-Latency Match: POST /v1/projects/{project}:updateLLM
   Update the low-latency match service, where project is the name of the given Mastering project.
2. Wait For Operation: GET /v1/operations/{operationId}
   Using the captured id from Step 1 above, poll the status state of the operation until status.state="SUCCEEDED" received.
3. Wait For Queryable: GET /match/v1/projects/{project}:isQueryable
   Poll the status of the LLM service until the response true received.

Low-Latency Matching

To match incoming external records at low latency:

1. Low-Latency Match: POST /match/v1/projects/{project}:match
   Submit records in a stream for low-latency matching against the latest queryable records or clusters, where project is the name of a Mastering project.