thalovant/thalovant-m2v-intents

thalovant/thalovant-m2v-intents

A static-embedding intent classifier for the Thalovant hub. It maps an utterance to one of 168 labels of the form <skill_id>:<intent>, exactly as ovos-m2v-pipeline registers them, and it knows only the Thalovant skills: every label comes from the fleet corpus and nothing else.

Base: Jarbas/m2v-256-distiluse-base-multilingual-cased-v2, fine-tuned with model2vec's classifier trainer. Held-out accuracy 0.9014 over 15481 rows across 68 languages.

Use

from model2vec.inference import StaticModelPipeline

model = StaticModelPipeline.from_pretrained("thalovant/thalovant-m2v-intents")
model.predict(["will it rain tomorrow"])

Configure the pipeline with "mode": "classifier" and this model's path; labels.json beside the weights lists the labels the hub may route to.

index.json names every sentence in the training corpus by a digest of its language and text, with the labels that publish it. thalovant-skillkit check uses it to fail a skill that publishes a sentence another skill already owns, and asks this classifier what it makes of the rest -- without the fleet's sentences leaving their private repositories.

Held-out accuracy per language

What the classifier confuses

Each line is a sentence whose true intent the classifier read as another intent. Two skills on one line is a collision the fleet should resolve.

• —thalovant-skill-news.thalovant:global.news read as thalovant-skill-news.thalovant:news (17): "gioca alle ultime notizie del mondo" (it-IT)
• —thalovant-skill-weather.thalovant:condition.weather read as thalovant-skill-weather.thalovant:current.weather (16): "hva er værforholdene" (nb-NO)
• —thalovant-skill-date-time.thalovant:holiday.when read as thalovant-skill-date-time.thalovant:days.until (13): "カナダの日はいつですか?" (ja-JP)
• —thalovant-skill-interaction-modes.thalovant:party.mode.enable read as thalovant-skill-interaction-modes.thalovant:party.mode.disable (13): "sprememba na način zabave" (sl-SI)
• —thalovant-skill-joke-garden.thalovant:party.host read as thalovant-skill-joke-garden.thalovant:joke.garden (13): "závěrečný kolo" (cs-CZ)
• —thalovant-skill-news.thalovant:news read as thalovant-skill-news.thalovant:global.news (13): "игра something световни новини" (bg-BG)
• —thalovant-skill-weather.thalovant:sunrise.weather read as thalovant-skill-weather.thalovant:sunset.weather (10): "सूरज कब उगता है" (hi-IN)
• —thalovant-skill-date-time.thalovant:days.until read as thalovant-skill-date-time.thalovant:holiday.when (9): "kdy je kanada day?" (cs-CZ)
• —thalovant-skill-guide.thalovant:ThalovantValues read as thalovant-skill-guide.thalovant:WhatIsThalovant (9): "apa itu filsafat thalovant" (id-ID)
• —thalovant-skill-source-scout.thalovant:source.scout read as thalovant-skill-safety-guide.thalovant:safety.guide (9): "kur aš turėčiau patikrinti šiandieną orą" (lt-LT)
• —thalovant-skill-interaction-modes.thalovant:party.mode.disable read as thalovant-skill-interaction-modes.thalovant:party.mode.enable (8): "menghalang mod parti" (ms-MY)
• —thalovant-skill-joke-garden.thalovant:party.host read as thalovant-skill-rewrite-forge.thalovant:rewrite.forge (8): "hårdare" (sv-SE)
• —thalovant-skill-reminder.thalovant:reminder.details read as thalovant-skill-reminder.thalovant:reminder.list (8): "erakutsi gogorarazpenaren xehetasunak." (eu-ES)
• —thalovant-skill-weather.thalovant:alerts.weather read as thalovant-skill-weather.thalovant:current.weather (8): "ali je tu hudo vreme?" (sl-SI)
• —thalovant-skill-date-time.thalovant:date.for read as thalovant-skill-date-time.thalovant:date.current (7): "dün ne zaman oldu?" (tr-TR)

Trained from the corpus built 2026-09-27T04:36:26Z covering 23 skills.