Explore a Patient
Ingest the synthetic Jeanne Tremblay dataset in the Context Engine Studio, browse the resulting patient graph, and ask an agent about her care.
In this tutorial you ingest the synthetic Jeanne Tremblay dataset into your workspace, follow the ingestion until it finishes, browse the patient's virtual file system, read a condition story, and ask the chat assistant a clinical question. It takes about fifteen minutes and needs nothing beyond a running Context Engine Studio.
Prerequisites
You need the studio running against your workspace, as set up in Install and Connect to a Workspace. The tutorial adds a patient to that workspace, which you can delete afterwards with Delete patient on the patient's page.
Steps
Download the dataset
Download jeanne-tremblay.zip and unzip it into a folder named jeanne-tremblay:
unzip jeanne-tremblay.zip -d jeanne-tremblayThe zip holds the FHIR R4 bundle (bundle.json) and the C-CDA documents (cda/) at its root, so the folder name is what the studio suggests as the patient ID. Synthetic Patient Datasets describes who Jeanne is and what her record contains.
Create the patient
Drop the jeanne-tremblay folder on the Add your first patient dropzone, check that Patient ID reads jeanne-tremblay, and select Create patient & ingest. The studio sends every .json and .xml file in the folder and skips the rest. In a workspace that already has patients, the same form opens from Create patient in the patient switcher.
Wait for the ingestion to finish
Select View ingestions in the notice that confirms the files were accepted, and wait until every row reads Succeeded, since ingestion runs in the background. Each row is one submission. Its Staged column counts the facts, events and narratives the engine staged from it. Ingestion explains what each status means.
Browse the virtual file system
Expand Virtual file system in the sidebar and open a few folders, for example /conditions/active/, /medications/current/ and one of the year folders under /encounters/. Each file opens in three views: Narrative and Compact, which render as Markdown, and Structured, which shows the JSON. VFS Paths lists every folder and file the tree can hold.
Read a condition story
Open /conditions/active/, then the COPD folder, then _story.md. The story follows the condition from its onset through treatments, monitoring labs and complications. It is the same text an agent gets when it calls the read_patient MCP tool on that path.
The _narratives/ folder beside it holds the narratives from Jeanne's documents that the engine linked to the condition. Condition Stories explains how the engine assembles a story.
Ask the assistant a question
Select New chat in the sidebar and ask a question that needs more than one file to answer, for example:
Given the recent acute COPD exacerbation, is the current inhaler regimen appropriate and is bone health being monitored?
The assistant browses and reads Jeanne's files over MCP before it answers and cites the files it used. Select any tool call in the thread to see its input and output.
Chats are listed per patient under Chats in the sidebar and reopen where you left them.
Next steps
- Connect an AI Agent to give your own agent the MCP tools the assistant used.
- VFS Paths for every folder and file the tree can hold.
- The Patient Graph for how the engine merged the FHIR and CDA records into one patient.