ChatMDR API
Looking for the API reference?
We are introducing the Application Programming Interface (API) to ChatMDR. You are now able to integrate ChatMDR or ChatIVDR into your own applications, like requirement management tools, software development environments, word processing for regulatory documents or instructions for use – whatever you can think of. Here are some possible use cases:
- Validation of a rationale
- Rewriting text to use the correct terminology from MDR
- Proofreading a response to an audit finding by the notified body
- Checking completeness of a GSPR checklist
- Checking references to articles and MDCG guidances
- Assessing significance of a change based on release notes
In short, you can use the API to automate a lot of the tasks that you would now switch back and forth to ChatMDR or ChatIVDR for. Paste the text, add a prompt, copy the response. With a little bit of software development, you can now streamline this workflow.
While the API is available to everyone, also for personal use, it will find most use with software integrators that add ChatMDR functionality to their own tools.
Availability
The ChatMDR API is currently available in beta. This is a testing phase in which we gather customer feedback before we go live. This is expected to happen in the last quarter of 2024.
During the beta, please request your API key using the Contact Form. We will be able to keep closely in touch with you and notify you by email of any updates or technical issues.
Pricing
During the beta period, access to the ChatMDR API is free! 🍻 We will be employing a fair use policy to protect our servers.
After the beta, the price of the API calls will be measured in tokens. In an AI large language model such as the ones underlying ChatMDR and ChatIVDR, every word is broken up into tokens. Roughly, a token is four characters long – so a 1,000 character response is about 250 tokens.
Prices are per 1 million (1M) tokens after the moment of public release in Q4 2024. Prices exclude Dutch VAT at 21%.
| Model type | Cost per 1M tokens (EUR/€) |
| Fast Model (chatmdr-fast-openai, chatmdr-fast-llama3_70b + IVDR equivalents) | 1.15 |
| Smart Model (chatmdr-smart-openai, chativdr-smart-openai) | 1.98 |
Example
A provider of technical file management software adds ChatMDR to their online application. Users can rewrite part of their documentation to more easily comply with MDR, by referring to the regulation and using similar wording. The tool is used by 100 organizations that, on average, submit 50 requests a day in total. Each request has 1,000 tokens of input and 750 tokens of output, plus a prompt from the manufacturer that tells ChatMDR what to do of 150 tokens. For the highest output quality, the manufacturer chooses to use the ChatMDR Smart model.
Per month, the total cost comes to:
50 requests / day x 30 days /month x 1,900 tokens / request x 1.98 EUR / 1M tokens = 5.64 EUR / month
Technical details
The ChatMDR API uses a subset of the OpenAI API. This means it’s compatible with existing tools and libraries in languages like Python and Node.js, meaning integration with your software should be smooth.
The API is at https://api.chatmdr.eu/ . For example, here is a GET request you can open in your browser to get the list of supported models as JSON: https://api.chatmdr.eu/v1/models
Please see the API reference and its Models section for more information.