User repository on GitHub
Home Assistant Assist conversation agent powered by TypeSafe's Jev (System One) model
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Integrations & Adapters
TypeSafe Assist
A Home Assistant Assist conversation agent that sends each command to Jev as one request with a speculative fan-out of questions, then hands the typed answers to Home Assistant's own intents to do the device control.
- Stars
- 3
- Language
- Python
- Published by
- Community
- Author
- JanOstrowka
- Last verified
- Oct 2026
About this project
A conversation agent for Home Assistant Assist that uses TypeSafe's Jev to interpret commands.
Quoted from the project's README. Read the full README.
- Languages
- Python
Highlights
- Each command goes out as a single Jev request carrying a fan-out of questions: intent, target, device, room, brightness and the rest.
- Answers map onto built-in intents such as HassTurnOn, HassLightSet and HassGetState, and Home Assistant does the control while templates speak the reply.
- Turn off the fan with two fans is resolved without the fallback agent: it asks which one, then settles the reply with a second, tiny Jev request.
- Installs through HACS as a custom integration, or by copying custom_components/typesafe_conversation; requires Home Assistant 2025.6 or newer.
Watch out
This is an independent, unofficial community integration, not affiliated with, endorsed by or supported by TypeSafe, and it runs against your own TypeSafe account and API key under TypeSafe's terms and pricing. Data exposure is the thing to check before enabling it: per the README's notes, only entities exposed to Assist are sent to TypeSafe, so narrow that list deliberately rather than by accident. Spoken responses are English only. Intents that need free text, such as a shopping list or a broadcast, are routed to the fallback agent because Jev cannot extract free text, and the README recommends configuring one, whether OpenAI, Anthropic or Ollama. Keep Prefer handling commands locally enabled so exact sentence matches stay on-device. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
Home Assistant. Jev is a System One model that does not generate text; it answers typed questions about the command and the exposed entities and returns choices with calibrated probabilities. Those answers are mapped onto Home Assistant's built-in intents, HassTurnOn, HassLightSet, HassGetState and the rest, and those do the device control. Spoken confirmations come from templates rather than from the model.
The Assist command text plus the entities exposed to Assist, and per the README's notes only those. Before enabling it, review that exposure list, since anything Assist can see is a candidate for the request. The Assist debug view shows each request with Jev's latency, the top intent probabilities, the confidence and the slots sent, which is the place to confirm what was actually transmitted.
Without falling back to a chat model. Unavailable devices are skipped, a command from a satellite prefers the speaker's room, plurals target every device of that kind, and otherwise the agent asks something like Which fan: A or B? and resolves the reply with a second, tiny Jev request. Setup also lets you set confidence thresholds if the defaults are too eager.
Free text is the gap: intents like a shopping list or a broadcast are handed to the fallback agent because Jev cannot extract free text, and a fallback agent is recommended anyway for conversational or multi-action commands. Installation is HACS or a manual copy of custom_components/typesafe_conversation, Home Assistant 2025.6 or newer, a TypeSafe API key, then set the pipeline's conversation agent in Voice assistants. Spoken replies are English.
Through HACS as the integration TypeSafe Conversation (Jev), or by copying custom_components/typesafe_conversation. Then add the integration, enter the TypeSafe API key, optionally set a fallback agent and confidence thresholds, and in Voice assistants set the pipeline's conversation agent to the new Jev entity.
Each request appears with Jev's latency, the top intent probabilities, the confidence, and the slots sent to Home Assistant, which is the place to confirm what was actually transmitted.
The README recommends it so exact sentence matches stay on-device rather than being sent to Jev and the fallback agent. Only entities exposed to Assist are sent to TypeSafe, and spoken responses are English only.
More Jev resources like this
25GitHub stars
Agent runtime with a terminal UI and a messaging gateway that keeps memory, self-built skills, session search, scheduling, and subagents across Telegram, Discord, Slack, WhatsApp, Signal, and CLI.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
wundercorp/lokiPython
wundercorp/lokiPython51GitHub stars
GraphRAG layer that runs a 4-phase ingest-to-answer pipeline over Neo4j, Memgraph, AGE or Kuzu using Laya or Jev decisions instead of per-hop LLM calls.
Jev Integrations & Adapters#community#jevbest#catalog-import
5GitHub stars
Community Agent Zero plugin that exposes one typesafe_query tool for combined Choice, Noul, and Score questions over the same evidence, with probability cards in chat.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
3clyp50/a0-typesafe-aiPython
3clyp50/a0-typesafe-aiPython69GitHub stars
Provider-neutral TypeScript runtime for System One decisions: write choice, noul, and score code once, then run it on TypeSafe Jev, local Reflex, any compatible /v1/systemone endpoint, or a custom provider.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
iamaamir/system-oneTypeScript
iamaamir/system-oneTypeScript23GitHub stars
Memory-maintenance library that checks every stored memory against each new event with Jev yes-or-no votes, then marks stale facts superseded or contradicted instead of editing them.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
chopratejas/invalidatePython
chopratejas/invalidatePython6GitHub stars
MCP layer that lets an agent harness ask Jev for a bounded choice from a candidate set it already allows, while permissions and execution stay with the host.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
typakon4/jev-layerJavaScript
typakon4/jev-layerJavaScriptFrom the Jev community
Posts from builders shipping with Jev right now.