--- title: "Jev (TypeSafe)" description: "Jev (TypeSafe) is a Marketplace integration for Home Assistant. It has 65 installs sharing analytics, needs the vendor's cloud and was last updated October 26, 2026." canonical: "https://vulpo.app/integrations/marketplace/abovecolin-ha-jev" url: "https://vulpo.app/integrations/marketplace/abovecolin-ha-jev.md" last-updated: "2026-10-22" version: "1.4" --- # Jev (TypeSafe) > Jev (TypeSafe) is a Marketplace integration for Home Assistant. It has 65 installs sharing analytics, needs the vendor's cloud and was last updated October 26, 2026. Jev (TypeSafe) is an integration that turns questions about the house into typed answers from a decision model. Each answer comes back as a probability, a chosen option or a score, and appears as a sensor or as the response of an action in an automation. Further actions ask several questions at once, locate something in a camera picture, and find the best threshold from recorded history. It also adds a conversation agent for Assist and a weekly house check that opens Repairs cards for entities unavailable for a week, low batteries and states that look like mistakes. Sensors track daily calls, input tokens and estimated cost against a configurable daily token budget. It needs the vendor's cloud or another provider, and is set up in the UI with a provider choice and an API key. Automatic summary of the README and code of AboveColin/HA-Jev. Something wrong, or are you the maintainer and want your own text here? [Get in touch](https://vulpo.app/contact) [Open in My Home Assistant](https://my.home-assistant.io/redirect/hacs_repository/?owner=AboveColin&repository=HA-Jev&category=integration) On Home Assistant 2026.11 or newer? [Open it in the Marketplace](https://my.home-assistant.io/redirect/marketplace_repository/?owner=AboveColin&repository=HA-Jev&category=integration) ## Supports - **Uses the vendor's cloud**: Needs an internet connection and an account with the vendor. - **From the Marketplace**: Installed through the Marketplace, formerly HACS. - **Actively maintained**: Its author released or pushed a change in the last six months. ## Setting it up Install it from the Marketplace and restart Home Assistant, then add it under Settings, Devices and services. ### Connect a decision model - **Provider**: Jev (TypeSafe), SubQ Snap and OpenRouter are hosted. A local server is one you run yourself and name in Advanced. - **API key**: The key the requests are sent with. Leave it empty only for a local server that needs no key. - **API address**: Leave it empty to use the provider's own address. A local server needs one here. The key is sent as a bearer header, so an http address puts it on the wire in clear. - **Model**: The model id sent with every request. Leave it empty to use the provider's default. ### Re-enter your API key - **API key**: Requests go to the API address already stored. ### Change the provider, key or address - **Provider**: Jev (TypeSafe), SubQ Snap and OpenRouter are hosted. A local server is one you run yourself and name in Advanced. - **API key**: Leave it empty to keep the current key while the address stays the same. A new address gets no key unless you type one. - **API address**: Leave it empty to use the provider's own address. A local server needs one here. The key is sent as a bearer header, so an http address puts it on the wire in clear. - **Model**: The model id sent with every request. Leave it empty to use the provider's default. ### After setup - **Daily input token budget (0 = no limit)**: Counted from midnight in the Home Assistant time zone. 0 means no limit. - **Price per million input tokens, in USD**: Sets the estimated cost sensor and the cost in the question preview. Nothing that is sent depends on it. - **Fall back to this agent**: Receives the whole sentence when Jev is not sure what to do. - **Act only above this confidence (0 to 1)**: Below this, Jev does nothing and passes the command to the fallback agent. - **Allow commands that name no room or device**: Turning everything off is always allowed, with or without this. - **Offer Jev as a tool to other LLM agents**: Adds two tools to the Assist API, one for yes/no questions and one for picking an option. Each LLM agent that uses Assist gets their descriptions in every prompt, which costs tokens at that agent's provider. Each call also counts against the budget above. - **Check the house once a week**: At 10:00, once 7 days have passed since the last check, Jev looks at every entity the Jev conversation agent can control in one request and opens a Repairs card for each state that looks like a mistake. The checks for unavailable entities and low batteries cost nothing and run with it. Texts from the code of [AboveColin/HA-Jev](https://github.com/AboveColin/HA-Jev). ## Entities - **Binary sensors**: Daily budget exceeded - **Sensors**: Calls today, Input tokens today, Estimated cost today, Latency, Payload ## Actions - **Ask a yes/no question**: Ask whether something is true and get the probability of yes, from 0 to 1. Use one of these per label when several labels can be true at once. Fields: State, Images, Question, Yes means, No means, Threshold, API key, Model, Include all attributes, Background - **Pick one option**: Pick one option from your list, and get the probability of every option plus a confidence value. Fields: State, Images, Question, Options, Option descriptions, API key, Model, Include all attributes, Background - **Rate on a scale**: Rate the text against levels you describe, and get a number that can land between two levels. Fields: State, Images, Question, Levels, API key, Model, Include all attributes, Background - **Ask several questions at once**: Ask any mix of question types about one piece of text, in a single request. Every question is judged independently against the same text and they run in parallel, so twenty questions take about as long as one. Fields: State, Images, Questions, API key, Model, Include all attributes - **Find where something is**: Ask where something is in the current picture of a camera or an image entity. The answer is a box: its left, top, right and bottom edges as fractions of the picture's width and height, from the top left corner, with a confidence. Needs Jev 1.14. Fields: Image, What to find, Background, API key, Model - **Find the best threshold**: Compare the recorded history of a yes/no probability with an entity that shows what was really true, and get the threshold that would have been right most often. Reads only the recorder, so it costs no tokens. Fields: Probability, What was really true, State that means yes, Days of history - **Check the house**: Look for entities that have been unavailable for a week, batteries below 10 % and, if Ask Jev is on, states that look like a mistake. Opens a Repairs card for each one and returns the list. Fields: Ask Jev, API key - **Undo the house check**: Put back what the house check turned off, with the state it had before. Fields: Entities, API key ## Troubleshooting - **Diagnostics**: Downloads a diagnostics file to attach to a bug report. - **System health**: Adds its status to Home Assistant's system information. - **Repair notices**: Notices it can show in Settings, Repairs: Jev stopped evaluating: daily budget reached, Has been unavailable for a week, The weekly house check did not reach Jev ## From the README > Ask TypeSafe Jev questions about your house and get numbers back. Jev is a decision model, not a chat model. It answers a typed question with a probability, a choice or a score, and this integration turns each answer into an entity you can automate on. > > Not affiliated with TypeSafe. The API client is > > - Questions become sensors: a probability, one of your options with its distribution, or a number on a scale. Add them in the UI, in YAML, or both. > - Five actions, jev.noul, jev.choice, jev.score, jev.ask and jev.box, answer inside an automation and return a response variable. > - An AI Task entity answers ai\_task.generate\_data when it is called. A boolean, select or number field becomes the matching question. > - Actions and the AI Task can look at cameras: list camera or image entities, or attach a snapshot, and the picture goes in beside the state, scaled to what the API takes. jev.box finds where… > - A conversation agent for Assist routes spoken commands through the same model. > - A house check opens a Repairs card for entities unavailable for a week, low batteries and, in one request, states that look like a mistake. A light left on can be turned off from its card and put… > - It reports what it spends: calls, input tokens and estimated cost per day, against a daily token budget. > - 25 blueprints to import with one click. Seven start from your own question, and 18 handle one situation each, such as a forgotten washing machine or a window open while the heating runs. From the README of [AboveColin/HA-Jev](https://github.com/AboveColin/HA-Jev#readme), [MIT License](https://spdx.org/licenses/MIT.html). Shortened. ## At a glance - Kind: Marketplace integration - Category: [Sensors](https://vulpo.app/integrations/category/sensors.md) - Number: 65 installs sharing analytics; 79 GitHub stars - Connection: Needs the vendor's cloud - Upkeep: Updated in the last 6 months - Needs: Needs the Marketplace (HACS on 2026.10 and older) - Latest version: 1.22.3, last updated October 26, 2026 - Domain: `jev` - Source: - Issues: - Screenshot: https://vulpo.app/integrations-img/e2eafacc58049036.webp (from AboveColin/HA-Jev's README) In Vulpo: Chart card, Apple Watch Independent directory, made by the developer of Vulpo. Not affiliated with Home Assistant or the Open Home Foundation. [About these numbers](https://vulpo.app/integrations/about) Last updated 2026-10-22. Source: https://vulpo.app/integrations/marketplace/abovecolin-ha-jev --- - HTML version of this page: https://vulpo.app/integrations/marketplace/abovecolin-ha-jev - Site index for agents: https://vulpo.app/llms.txt - Everything in one document: https://vulpo.app/llms-full.txt - All pages: https://vulpo.app/sitemap.xml