Use case

Instant answers on Agribalyse + EF3.1

Ask an LLM environmental questions in natural language, while keeping the answer grounded in Volca through MCP instead of relying on a generic model alone.

Natural-language questions are easy to ask and hard to ground

A plain LLM can sound convincing without being tied to the actual dataset or method. Volca makes a different promise: let the model use a real environmental engine through MCP, so the path from question to answer stays connected to the underlying data.

What this use case shows

  • ask a concrete question in natural language
  • let an LLM call Volca through MCP
  • ground the answer in Agribalyse + EF3.1
  • keep a path back to inspectable environmental data instead of a black-box reply

Example questions

  • What is the difference between organic and conventional tomatoes in water consumption?
  • Which processes contribute most to this result?
  • What changes between these two variants?
  • Where does this result come from structurally?

Why it matters

For recurring environmental questions, people want the speed of chat without losing seriousness. MCP gives the LLM a way to use Volca as the computation and data layer instead of improvising from generic training data.

How the workflow works

  • the user asks a question in natural language
  • the LLM uses Volca tools exposed through MCP
  • Volca queries the dataset and method context
  • the answer can stay tied to a real environmental workflow

What Volca contributes

  • database access and environmental computation
  • structured outputs the model can use
  • inspectability beyond the surface-level answer
  • a reusable engine for agent and assistant workflows

What this is not

This is not just a nicer chat wrapper. The point is to plug an LLM into a real environmental engine so answers can be generated faster without severing the connection to datasets, methods, and inspectable structure.

Current scope

This use case currently focuses on LLM-assisted instant answers through Volca MCP, with Agribalyse + EF3.1 as the narrow, credible dataset-and-method scope.

If you want to see the browsing and inspection side of Agribalyse in the web UI, see the dedicated page for that workflow.

Want to test this on real environmental questions?