The AI prompt field lets you generate text, summaries, classifications, and insights directly in your Baserow tables using dynamic prompts that reference other fields. It supports multiple AI providers and can process documents (PDF, DOCX, TXT, MD) for automated analysis.
The AI prompt field is available on Premium plans and higher. Upgrade your subscription to access this feature.
The AI prompt field transforms your Baserow tables into intelligent automation engines. You can reference data from other fields to create dynamic prompts, classify content, summarize documents, or generate creative text.
These prompts can reference other fields in your table, making responses contextual and dynamic. Configure the model, temperature, and output format, then click generate in any row to produce AI-powered results.
AI prompt fields use large language models (LLMs) to process prompts you create. Unlike simple text inputs, these prompts function like formulas. You can combine static text, field references, and data manipulation functions to guide the AI.
Each cell includes a generate button that processes your prompt on demand. You can use formulas to construct precise inputs and automatically regenerate AI content when your source data changes, keeping your database always up to date.
Common use cases:

Ensure AI is configured in your workspace:
+ to add a new field.| Setting | Purpose |
|---|---|
| AI type & model | Choose your provider (Baserow, OpenAI, Anthropic, etc.) and model (e.g., GPT-4o, Claude 3.5) |
| Temperature | Control creativity (0-2). Use 0–0.3 for factual tasks; 0.7+ for creative writing |
| File field | The first compatible file in the field |
| Output type | Format the response. Use “Text” for open answers or “Choices” for strict classification |
| Regenerate | Toggle “Regenerate when referenced fields change” to keep AI results in sync with your data automatically |
| Prompt | The instruction sent to the AI. This supports the formula builder for complex logic |
The prompt editor is a powerful formula builder. You are not limited to simple text; you can use inputs, data, operators, and functions to craft the perfect context.
concat('Classify this review: ', field('Review Text')))You have two ways to generate content:

The AI prompt field can analyze text-based documents directly. This is ideal for extracting invoice numbers, summarizing contracts, or standardizing resumes.
| Format | Extension | Common Use |
|---|---|---|
| Plain text | .txt |
Notes, logs, transcripts |
| Markdown | .md |
Documentation, articles |
.pdf |
Invoices, reports, contracts | |
| Word | .docx |
Documents, proposals |
The AI will process the first compatible file found in that cell.

Returns free-form AI-generated text. Best for summaries, descriptions, or creative content where responses vary widely.
Example: “Write a professional email response to this customer inquiry: {Support Ticket}”
Constrains the AI to select from predefined options. This transforms qualitative data (text) into quantitative data (categories).
Setup:
Fields > Review Text”The AI will analyze your prompt and return only one of the specific options you defined.
Temperature controls response randomness:
| Range | Behavior | Best For |
|---|---|---|
| 0 - 0.3 | Focused, deterministic | Factual summaries, data extraction, classification |
| 0.4 - 0.6 | Balanced | General business writing, Q&A |
| 0.7 - 1.0 | Creative, varied | Marketing copy, brainstorming, creative content |
| 1.1 - 2.0 | Highly experimental | Rarely recommended; very unpredictable |
Pro tip: Start at 0.5 and adjust based on results. Lower if responses are too random, raise if they’re too repetitive.

The self-hosted version of Baserow has no AI prompt
You need to configure the AI prompt field to enable it in self-hosted instances. The settings are located in your environment configuration in the developer documentation.
For example, to use OpenAI, you must set BASEROW_OPENAI_API_KEY and BASEROW_OPENAI_MODELS
Available models depend on your generative AI configuration. Cloud users have access to OpenAI models by default. Self-hosted users can configure OpenAI, Anthropic, Ollama, OpenRouter or Mistral models.
Yes, if you enable it. In the field settings, toggle “Regenerate when referenced fields change”. When this is enabled, any change to the source data (e.g., editing the text in a “Notes” field) will automatically trigger the AI to re-analyze and update the result.
Baserow Premium plans include the use of AI prompts in the field. Monitor your usage through your provider’s dashboard.
Yes. The prompt input supports the full Baserow formula language. You can use functions like upper(), trim(), or if() to clean or format your data before sending it to the AI.
Yes. Reference unlimited fields and select from your table’s fields. Example: Compare {Fields > Product A} and {Fields > Product B} and explain which is better for {Fields > Use Case}.
Common causes include unsupported file formats, corrupted files, or files exceeding the model’s token limit. Ensure files are in supported formats (.txt, .md, .pdf, .docx) and under 10MB. Check error messages in the cell for specific guidance.
Yes. You can generate all AI values in one go by clicking the Generate all AI values button located in the field header. This is useful when importing large datasets that need immediate processing.
Need help? Visit the Baserow community or contact support for assistance with your account.