AI in Moliri
What the AI can do
This page lists every function in Moliri that has a model behind it, with the place where it runs and what it needs. Each function is described in its own place in the handbook; the name in the first column leads there. What the places mean is explained under Where the AI runs .
Capturing
| Function | What it does | Where it runs | Requirement |
|---|---|---|---|
| Speech input | Dictation into any text field, through the microphone in the field or a key combination | Transcribe on this PC, in the browser, at ElevenLabs or on a server of your own | Switched on in the profile |
| Recognise text | Reads the text in images and PDFs, on a file’s Text tab | In the browser, with Transcribe, on the server or by the AI service | None; the AI service needs a model for text recognition in the AI account |
| Transcription | Turns audio and video into text with timestamps and speakers | On the server | None |
| Text recognition by rules | Transcribes or reads a file as soon as it gets a certain keyword | On the server or by the AI service | Rules set by the administration |
| Suggest fields with AI | Assigns the texts read from a document to fields in the OCR import | The instance’s chat model | AI account, write access in the project |
For recordings that are not in Moliri, the desktop app Transcribe by Moliri transcribes on your computer.
Searching
| Function | What it does | Where it runs | Requirement |
|---|---|---|---|
| Search by meaning | Finds objects by what is meant: vessels for liquids finds jugs and amphorae. For a single field as the comparison Meaning in the advanced search | On the server | The project has been read in under Preparation |
| Field search | Finds fields in every field picker that have a different name but mean the same: Künstler (artist) finds Urheber (creator) | In the browser or with Transcribe | Switched on in the profile |
| Column header in the Excel import | Suggests a field for open columns by their header, English headers included | Like the field search | Field search switched on, model already loaded |
Checking and upkeep
| Function | What it does | Where it runs | Requirement |
|---|---|---|---|
| Similar values, AI run | Finds values that mean the same but are written differently, such as Florenz and Firenze | The instance’s embedding model and chat model | AI account with an embedding model |
| Dimensions from free text | Reads height, width and depth from a text such as sheet 34.5 × 47 cm | The instance’s chat model | AI account |
| AI suggestions | Collects all suggestions for field values, to apply or dismiss | No model of its own | A run or an assistant has suggested something |
Moliri also finds duplicates and similar spellings without a model. Which check finds what is explained under Finding duplicates .
Working with an assistant
| Function | What it does | Where it runs | Requirement |
|---|---|---|---|
| Access tokens | Connect an AI assistant such as Claude or Cursor to Moliri via MCP. It reads with your rights and submits suggestions | At the assistant of your choice | A token in the profile |
| Field catalogue for AI | Puts the structure of a project’s fields on the clipboard, without object data | Moliri sends nothing | None |
| Prompts to copy | Ready-made requests for a language model of your choice: for formulas of virtual fields , Word templates , new fields and value lists , event rules and the mapping to standards | Moliri sends nothing | None |
With a copied prompt, you paste the answer back yourself. Moliri then checks it like any other input in the editor concerned.
Further reading
- AI overview : which of these functions are set up in your instance and in your browser
- Where the AI runs : what runs where and where data goes
- AI jobs and nightly runs : how the functions work on the server