Hi team,
I wanted to propose adding support for a local Ollama-based NLP backend to enable natural-language computed fields.
Context
Currently, computed fields require manual SQL expressions. This proposal introduces an optional NLP layer that allows users to define computed fields using natural language, processed locally via Ollama.
Why this is useful
- Enables non-technical users to create computed fields easily
- Keeps everything local (no external API dependency)
- Aligns with the growing use of LLM-assisted data exploration
What’s implemented
I’ve already created a working implementation here:
#491
Key aspects:
- Local Ollama integration
- Translation of natural language → computed field expressions
- Designed to be optional and non-intrusive to existing workflows
Questions
- Would this align with the current roadmap?
- Would you prefer this as a plugin instead of core integration?
- Any preferred abstraction/interface for NLP backends?
- Happy to make any changes or adjustments based on feedback.
Thanks!
Hi team,
I wanted to propose adding support for a local Ollama-based NLP backend to enable natural-language computed fields.
Context
Currently, computed fields require manual SQL expressions. This proposal introduces an optional NLP layer that allows users to define computed fields using natural language, processed locally via Ollama.
Why this is useful
What’s implemented
I’ve already created a working implementation here:
#491
Key aspects:
Questions
Thanks!