{"node":{"id":"23ed6294-2f43-4e48-aa8c-feb963ce217b","handle":"23ed","type":"note","authorName":"Vincent Yeung","intent":"truth","title":"Laya, a free decision model on your own machine: where it matches Jev and where it does not","aiSummary":"Laya, from Convai Innovations, is a free decision model that runs on your own machine and does the job Jev does in the cloud: you give it a question and the answers, and it returns a probability for each, with no charge per question and no text leaving your machine. It matches Jev on questions one message can answer, like whether a message holds a link or which kind it is, and gets wrong the questions that need context, like which thread a short reply belongs to. How to install it, how to split the work between Laya and Jev, and how to train it for the questions it misses.","content":"Run Laya when you want a decision model on your own machine. You give it a question and the answers to choose from, and it gives a probability for each, the same job Jev does as TypeSafe's paid cloud service. Laya is free to download from Convai Innovations and free to ask. Because it runs on your machine, your text never leaves it.\n\nLaya is built on ModernBERT-large, a public model that reads text. Its Apache 2.0 licence lets you use it in paid work.\n\n**Setting it up.** Install a PyTorch build that matches your graphics card, or the processor-only build if you have no graphics card. Then install Laya with `pip install laya`. The first question downloads the model, and later runs load it from your disk. On a graphics card, one decision takes tens of milliseconds.\n\n**Use Laya for questions one message can answer.** Ask whether a message contains a web link, or which of a few kinds a short message is, and Laya gives the same answer as Jev with about the same confidence. That makes it the right tool for checks you run on every message. Which language a message is in is another one.\n\n**Use Jev for questions that need context.** Laya gets these wrong, because the answer depends on how one piece of text relates to another:\n\n- which of several conversation threads a short reply like \"go\" belongs to\n- whether a reply starts a new thread\n- sorting real messages six ways, such as question, instruction or idea\n- whether a short reply needs a language model to understand it\n\nSend these to Jev, or to a language model running on your own machine.\n\n**Write the split yourself.** Laya's `Router` picks which Laya model answers, by the language of the text. It never hands a question on to a bigger model, so sending the hard questions elsewhere is code you write.\n\n**Train it for the questions it misses.** The base Laya models need further training before they do well on mixed decision questions. The Laya repository includes a notebook that trains it on your own labelled examples. So a question it gets wrong out of the box, such as which thread a reply belongs to, can be trained from your own conversations.","capturedAt":"2026-09-21T02:58:19.416Z","createdAt":"2026-09-21T02:58:19.417Z","updatedAt":"2026-10-28T02:49:22.422Z","spanId":null,"contributor":null,"audiences":[{"kind":"public","createdAt":"2026-09-21T03:00:49.120Z"}],"fingerprint":"f65c8d300ae4aff3968c5c5815ea288bacf15d08bde597b9df3c36dc4f7f0fb6","revisionNo":"4","structureHash":"5d72c65a4ec88a36e8340ee7f6e10d5c0339dc1b0cbc0d2b5765a12bd2490f3b","page":"https://lattris.com/by/b/memories/23ed6294-2f43-4e48-aa8c-feb963ce217b"},"links":[{"direction":"out","verb":"references","reason":"only if you send the questions that need context to Jev and have not set it up yet","hash":"ab0cfca92d1adaa511f8d5e3a65d528d477c90cbb19cb85fdb6fd7935a51cc93","node":{"id":"8c85109d-51c6-43bc-9ab4-98a5f4b02d00","handle":"8c85","type":"note","title":"How to set up Jev","aiSummary":"Get a TypeSafe key and keep it out of your shell history, install the typesafe-sdk package and TypeSafe's Claude Code skill, learn the call shape from the installed package's own help rather than from any page, the four things to find there before your first call, and the smallest call that proves you are set up.","erased":false,"href":"https://lattris.com/api/lattris/by/b/memories/8c85109d-51c6-43bc-9ab4-98a5f4b02d00?links=1","fingerprint":"e58b9f9428d50daee5533ce4832baa9f33342660ff5ac6a6b147558f412407be","revisionNo":"7","structureHash":"e598ec0fb6d1a55e11dfa63658c14c60f18b0a630b0a9797f1944afed800abc2","intent":"recipe","authorName":"Vincent Yeung","page":"https://lattris.com/by/b/memories/8c85109d-51c6-43bc-9ab4-98a5f4b02d00"}},{"direction":"out","verb":"references","reason":"only if you have not used Jev and need to see what it does before you compare","hash":"383ac0a4fd69822d648b1eba369d099bafd960b966c8494e6e48bd243488b589","node":{"id":"f0f5decc-32a9-49de-ad3d-0ea1871263fc","handle":"f0f5","type":"note","title":"Jev picks from answers you write: when to use it and how to word the questions","aiSummary":"Jev, TypeSafe's decision model, picks from answers you write and gives each one a probability. Use it when the same judgment is needed many times and you can write every answer down first. Read the spread of probabilities to find the unsure answers. Word each answer tightly, and ask first whether the question applies, because Jev picks a wrong answer when the right one is missing. A person still decides what matters, because some picks are wrong and Jev does not learn from its mistakes.","erased":false,"href":"https://lattris.com/api/lattris/by/b/memories/f0f5decc-32a9-49de-ad3d-0ea1871263fc?links=1","fingerprint":"43a30188c8bed6cb34e1c429c68b09aff611314c63650658756f1563927cdb67","revisionNo":"13","structureHash":"0df22d1c0fbb9aa7e1ba50bd4a7eab4f3b9a833a10eec51ab1aea1289522cef3","intent":"truth","authorName":"Vincent Yeung","page":"https://lattris.com/by/b/memories/f0f5decc-32a9-49de-ad3d-0ea1871263fc"}},{"direction":"in","verb":"references","reason":"only if you need a decision without a cloud call","hash":"6bdb221ef474818b6e5a86c68e6db1a0f1f3af746e7de5e094b0b17852b0f8ec","node":{"id":"f0f5decc-32a9-49de-ad3d-0ea1871263fc","handle":"f0f5","type":"note","title":"Jev picks from answers you write: when to use it and how to word the questions","aiSummary":"Jev, TypeSafe's decision model, picks from answers you write and gives each one a probability. Use it when the same judgment is needed many times and you can write every answer down first. Read the spread of probabilities to find the unsure answers. Word each answer tightly, and ask first whether the question applies, because Jev picks a wrong answer when the right one is missing. A person still decides what matters, because some picks are wrong and Jev does not learn from its mistakes.","erased":false,"href":"https://lattris.com/api/lattris/by/b/memories/f0f5decc-32a9-49de-ad3d-0ea1871263fc?links=1","fingerprint":"43a30188c8bed6cb34e1c429c68b09aff611314c63650658756f1563927cdb67","revisionNo":"13","structureHash":"0df22d1c0fbb9aa7e1ba50bd4a7eab4f3b9a833a10eec51ab1aea1289522cef3","intent":"truth","authorName":"Vincent Yeung","page":"https://lattris.com/by/b/memories/f0f5decc-32a9-49de-ad3d-0ea1871263fc"}}],"dictionary":{"lattris":"A personal knowledge base — how a person captures what they know as linked blocks they own and control. To understand how this knowledge base works: https://lattris.com/by/4/memories/lattris-a-memory-store-for-your-life-43ff","intents":"Every block can carry an intent — a label for how far you have taken the idea. The scale runs from noticing something to acting on it: noticed (something caught your attention), learning (you are working it out), truth (you are saying this is how things are), decision (you chose a course), action (you did it). Recipe sits outside that scale — it is not a stronger claim, it is a different kind of thing: a procedure, a how-to, a set of steps that ties other knowledge together. Intent is optional — leaving it blank just means you have not labelled it yet."},"instructions_to_ai":"This is knowledge from a personal knowledge base. Read it and consider how it's relevant to what the person you're talking to needs. Follow the linked blocks to explore further related topics if it might be useful to the user; open each at its page address. Start by letting the user know what this is about and how it may be relevant to them."}