Connector request — In-house DPI (child psychiatry) who already reads structured Heidi models
G
Grégor Mamou
Hello, I have had the approval of my manager to consider Heidi Cabinet if you can assure me that the integration will be possible with our DPI. Thank you for coming back.
G
Grégor Mamou
Thanks for your response! Here are the details:
- What we expect: both ways.
- To Heidi: create or link to the patient's session from the DPI (identifier, name, age) and send the context of the file (what we are pasting today into the Context tab). The agenda is not essential.
- To the DPI: retrieve the consultation observation and the output of a given model, launched on the session from the DPI, in plain text (or structured if possible), ideally with a notification (webhook) when the document is ready.
- At what time: during the consultation for the transcription, then just after for the observation, which is reviewed and validated in the DPI. The “dossier” models (synthesis, diagnostics, genogram, report) are launched later, on demand, often without any consultation: it is the context of the file that is then used as the source, not a transcript.
- The models: these are custom models that we created in Heidi for child psychiatry, with strict format guidelines for the DPI to analyze them:
- Patient summary: fixed sections (diagnosis, anamnesis, treatment, treatment, stakeholders, evolution, risks, next steps), updated from the previous summary;
- Diagnostics: one line per diagnosis “entitled — ICD-11 code — DSM-5-TR code — principal/secondary — justification”;
- Genogram: a text format (persons, couples, children, children, relationships, homes, with stable identifiers P1, C1...) that the DPI draws;
- Follow-up report: text under fixed titles, integrated into a PDF.
The consultation observation uses our own report model (similar to a SOAP adapted to child psychiatry).
The limit today: everything involves copy and paste. For each model, you must copy the context from the DPI, open the correct session, paste into Context, launch the model, then copy the result into the DPI. It's slow and prone to errors: bad session or bad patient, incomplete or truncated context, partially copied result. Without a shared ID, there is also no guarantee that the Heidi session is a good match for the PGD patient.
I can provide you with the exact text of our models and sample outputs if that helps with the framing.
Dr Gregor Mamou

Heidi Team
Thanks for this feedback, Grégor Mamou — it's very useful for us to fully understand the need for home DPI. So that we can frame the right type of connector, could you specify: (1) what exactly you expect from the integration (import of patient data/agenda to Heidi, export of Heidi notes to the DPI, or both); (2) at what point in the care journey you would like to use Heidi in your PGD (during the consultation, just after, or rather for proofreading/validation later); and (3) when you say that your DPI “already reads structured Heidi models”, what models/structures are they (e.g. SOAP, RAPID, model) child psychiatry), and what is the limit today without a connector?
G
Grégor Mamou
I know that there is no open API and that the connectors are developed on demand. Here's what we're doing today with Heidi. An adapted connector would also justify switching from my practitioner package to a pro or office package.
Current use
Heidi writes the observation with our model; I paste it into the DPI, reread it and validate it. We also created strict format models, which the DPI automatically analyzes:
- Patient summary: updated from the previous one, in fixed sections.
- Diagnostics: “title — ICD-11 — DSM-5-TR — principal/secondary — justification”, codes checked and then proposed to the practitioner.
- Genogram: structured text (persons, couples, children, children, relationships, households, stable identifiers P1, C1...) that the DPI draws; Heidi complements the existing genogram.
- Follow-up report: text with fixed titles, integrated into a PDF with diagnoses, genogram, treatments and exams.
The problem
Everything involves copy and paste: the DPI prepares the context, I paste it into the Context tab, I launch the model, then I copy the result into the DPI. The results are excellent, but it is long and error-prone (bad patient, incomplete context, truncated copy).
What a connector would allow
- Associate a PGD patient in a Heidi session.
- Send the context of the file in the session.
- Launch a template and retrieve its output (plain or structured text).
- Retrieve the observation at the end of the consultation, ideally via webhook.
- Optional: synchronize our models.
I develop the DPI myself: the integration on our side is my responsibility and I adapt to your format (REST, OAuth, webhooks...). I can share the specifications of our models. Nothing enters the file without validation from the practitioner.
Would you be open to this type of connector for a non-commercial DPI?