Skip to content
Recued
Menu
← Back to recipes

Mistral safety and classification brief

by recued-core v1

Read-only Mistral safety workflow. It runs moderation and classification on explicit text inputs, optionally creates embeddings for retrieval readiness, and summarizes scores through recued-core.mistral. It never uploads files, mutates agents or jobs, enables tools, uses streaming, or calls arbitrary Mistral API routes.

How it works 9 steps

Inspect the data fetches, transforms, gates, and output this recipe runs.

Process (9 steps)
moderation ?
classification ?
embeddings ?
skip: config.run_embeddings not_equal true
moderation_results default
Apply default
classification_results default
Apply default
summarize ?
moderation_table to_table
Format results as a data table
classification_table to_table
Format results as a data table
card to_summary
Format results as a summary card
Settings 8 configurable

Configurable at install. Defaults shown — change them anytime in Recued.

focus setting = Summarize safety and classification signals for the supplied texts. Compare moderation categories, classification results, embedding readiness metadata, and practical follow-up actions. Do not invent facts outside the Mistral responses.
inputs setting =
mistral setting =
max length setting = 900
run embeddings setting = true
embedding model setting = mistral-embed
moderation model setting = mistral-moderation-latest
classification model setting =

Trust & control

What installing this recipe would let it do. Recued grants these permissions at install — review them there before approving.

Where it runs

Device + server Runs on your device (browser) or your server.

Permissions it requires

Read your Mistral connection
Declared by the recipe — Recued grants these at install, where you review them before approving.

About

Tags

mistral mistralmoderationclassificationembeddingssafetyai-summarypack:mistral

Details

9 steps 8 configurable settings recipe_id: mistral-safety-classification-brief