# [How C3] Jan - An Open Source ChatGPT [<img src="https://hackmd.io/_uploads/SJPg16QSxe.png"/>](https://hackmd.io/_uploads/SJPg16QSxe.png) The open source software [Jan.ai](https://jan.ai/) is a simple chat interface for Linux, Mac and Windows that allows you to bring your own open source llms. Once downloaded and installed it greets you with the welcome screen. You can either setup a local model which is gonna run on your local machine or you decide to setup a remote provider. On most devices you can only use smaller models and even if you install on a gaming PC with a 4090 or 5090 you won't be able to run a 70b model which is why we focus on using comput3 as a remote provider. ## Connect your comput3 account Click "Set up remote provider" and lets get started with the setup. [<img src="https://hackmd.io/_uploads/HynebTXBex.png"/>](https://hackmd.io/_uploads/HynebTXBex.png) In the bottom of the list click "Add Provider" just call it comput3 and fill the following two fields in the UI: - Base URL: https://api.comput3.ai/v1 - API Key: your-comput3-api-key (scroll to the end of this post to find it) Once you did that you can click the Refresh button. If the 2 parameters are correct it will display the models that are available for your api key. By default this base url grants you access to llama3:70b and if you hold 500 ai16z tokens you'll also unlock hermes3:70b. [<img src="https://hackmd.io/_uploads/SkmEZa7Hgx.png"/>](https://hackmd.io/_uploads/SkmEZa7Hgx.png) ## Start Chatting with Llama3.3 70B Now that your models are configured click "New Chat" in the bottom left of the UI/UX and select the model of your choice and you can start chatting with the selected llm. Chatlogs are stored on your local computer. [<img src="https://hackmd.io/_uploads/S1cS-TQSgg.png"/>](https://hackmd.io/_uploads/S1cS-TQSgg.png) [<img src="https://hackmd.io/_uploads/SJVLWpXree.png"/>](https://hackmd.io/_uploads/SJVLWpXree.png) ## Custom Assistants The default assistant "Jan" is configured to use various tools to which he might refer to in his responses. This which might confuse it for regular chatgpt-like dialogues. To solve this issue vavigate to Assistants in the top left (or click Jan in the top bar) to create a new one. Lets call him Alfred with the instructions "You are a helpful AI assistant." and save him. [<img src="https://hackmd.io/_uploads/SJhXP6mSxg.png"/>](https://hackmd.io/_uploads/SJhXP6mSxg.png) You can switch between assistants in the top left - next to the search bar. You might need to create a new Chat first before being able to change the assistant. Generally its recommended to always create new chats as you switch to different topics which helps the language model not getting confused by completely unrelated stuff from earlier. [<img src="https://hackmd.io/_uploads/HkN2I6Xrxl.png"/>](https://hackmd.io/_uploads/HkN2I6Xrxl.png) As an open source project Jan is under constant development by a small team for quite some time. However if you are looking for a closed source alternative that might be a bit more capable check out [LM Studio](https://lmstudio.ai/) and put in your base url + api key to connect comput3. # Setup your comput3 Account After you've [acquired](https://dexscreener.com/solana/34qhkhrhyningwruftqjnb2vfv8oqyqd5tvdwwde1man) and [staked](https://app.streamflow.finance/staking/solana/mainnet/DGZebyog1twdFaKwN2KZkBLxDkzX4rh7L61S6TAty8et) some $com token you can [login to your dashboard](https://launch.comput3.ai/) with your solana wallet and grab your API key at the bottom page. You can do way more things in this dashboard like spin up comfyui instances or launch dedicated ollama instances with broader model selection but for standard 70b llm inference via the default comput3 base url (https://api.comput3.ai/v1) this is not needed. In this tutorial we keep it simple. [<img src="https://hackmd.io/_uploads/BJGO16mreg.png"/>](https://hackmd.io/_uploads/BJGO16mreg.png) If you run into issues join the [official comput3 Discord server](https://discord.gg/EFZPGm9qUu) and ask the community.