How Records Become Memory
How the Context Engine turns records from systems of record into a patient graph, and what you get back when you send them.
Recorded Memory is the memory the Context Engine forms from systems of record, such as EHRs, lab systems and claims systems. It is built to take in what those systems hold, whether structured data, documents or images. The ingestion endpoints read FHIR R4 and R5 bundles and CDA R2 documents.
What happens after you send records
Ingestion is asynchronous. The engine accepts a submission, returns a receipt and processes the submission in the background.
Processing turns the records into the patient graph. Each record is kept as an extracted copy of what its source stated. Records that describe the same thing, such as one condition coded differently by two sources, are merged into a single fact. Facts and clinical events are then linked to one another, for example a medication to the condition it treats. Prose from clinical documents is grouped by meaning into narratives, which are checked against the facts they speak to. Facts and events are also indexed so that agents can search them.
Each submission is resolved against everything the engine already holds for that patient, so a medication sent today is linked to a condition sent last year. The Patient Graph explains each of these steps.
What you get back
Each ingestion call returns a receipt carrying an ingestionId, which you poll to follow the submission until it settles, as described in Ingestion. Once it settles, everything it contributed is part of the patient graph. Agents then read it through context assembly.
Next steps
- Ingest a FHIR Bundle and Ingest a CDA Document to send a patient's records.
- Ingestion for what each status on the receipt means.
- The Patient Graph for how records are merged into facts, linked and kept traceable.
- How the Engine Thinks About Data for the model behind Recorded Memory.
- From Theory to Practice for the objects the API returns.
Synthetic Patient Datasets
Download and ingest fully synthetic FHIR R4 and C-CDA datasets — an older adult with COPD, a school-age surgical patient, an infant with one admission, and a ventilator-dependent preschooler born extremely preterm — to exercise the Clinia Context Engine without real patient data.
The Patient Graph
How the Context Engine builds the patient graph from ingested records and keeps every decision traceable.