# AI use cases for Jev Chat (w3cj/jev-chat) Repo: [w3cj/jev-chat](https://github.com/w3cj/jev-chat) | Use cases: 4 | Page: https://dedupedai.com/use-cases/w3cj-jev-chat/ A tool-calling chat bar that runs real MCP tools for weather, units, Wikipedia, recipes, search, tasks and smart home, with no LLM writing the replies. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Ask weather, unit and recipe questions without an LLM: https://dedupedai.com/use-cases/w3cj-jev-chat/ask-everyday-questions-without-an-llm/ - Add Todoist reminders and control Home Assistant by chat: https://dedupedai.com/use-cases/w3cj-jev-chat/manage-tasks-and-home-lights-by-chat/ - Answer factual questions from Wikipedia and web search: https://dedupedai.com/use-cases/w3cj-jev-chat/answer-facts-from-wikipedia-and-web-search/ - Trace why the chat bar picked a tool: https://dedupedai.com/use-cases/w3cj-jev-chat/trace-why-a-tool-was-picked/ ## How to ask weather, unit and recipe questions without an LLM with Jev Chat Repo: [w3cj/jev-chat](https://github.com/w3cj/jev-chat) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Run the chat bar locally and get weather, unit conversions and recipes from tools, not model text. **Who it is for:** A person who wants quick everyday answers from a chat bar and does not want a language model to make up numbers or facts. **The problem:** Chatbots can state a temperature, conversion or recipe step that no source returned. Users need answers where every value comes from a tool. **The outcome:** A running local chat app that answers weather, unit and recipe questions, with every value on screen typed by the user or returned by a tool. ### Steps 1. Run `corepack enable` to pick up the pinned pnpm version. 2. Run `pnpm install` in the repo folder. 3. Run `cp .env.example .env` and set TYPESAFE_API_KEY in the new file. 4. Run `pnpm dev`, then open http://localhost:5173 in a browser on the same machine. 5. Type "350F in celsius" and "What's the weather in Denver tomorrow?" and check that each reply cites tool data. ### Prompt to paste into your agent ```text In the jev-chat repo, copy .env.example to .env, set TYPESAFE_API_KEY to , and run pnpm dev. Send the message "350F in celsius" to the chat at http://localhost:5173 and report the reply. Verify that the reply value came from the units tool and not from model text. ``` ### Caveat The app only works in English, and every message gets an error reply until a paid TypeSafe key is set. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/w3cj-jev-chat/ask-everyday-questions-without-an-llm/ ## How to add Todoist reminders and control Home Assistant by chat with Jev Chat Repo: [w3cj/jev-chat](https://github.com/w3cj/jev-chat) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README Connect Todoist and Home Assistant so one chat bar can add reminders and switch devices. **Who it is for:** A person who runs Home Assistant and uses Todoist, and wants to add a reminder or switch a light from one typed message. **The problem:** Switching between task apps and home control panels breaks flow, and voice assistants are not always at hand. **The outcome:** A chat bar that adds a Todoist task from a typed reminder and changes a Home Assistant device after a confirm step. ### Steps 1. Run `cp .env.example .env` and set TYPESAFE_API_KEY, TODOIST_API_KEY, HASS_URL and HASS_TOKEN in the new file. 2. Enable the MCP Server integration in Home Assistant, which the README places at /api/mcp. 3. Run `pnpm tools:list` to confirm the Todoist and Home Assistant servers expose their tools. 4. Run `pnpm dev` and open http://localhost:5173. 5. Type "Remind me to renew my passport tomorrow", then "Turn off the kitchen lights", and answer the confirm card before the light changes. ### Prompt to paste into your agent ```text In the jev-chat repo, set TYPESAFE_API_KEY, TODOIST_API_KEY, HASS_URL and HASS_TOKEN in .env to values from , then run pnpm tools:list. Expected output is the tools each connected server exposes. Verify that the Home Assistant server is listed before sending a light command. ``` ### Caveat Servers without a key show as "no key" in the sidebar, and a Home Assistant command acts on real devices only after the confirm step. Risk, credentials: The Todoist key and Home Assistant token give the app access to your task list and home devices, so keep the .env file private. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/w3cj-jev-chat/manage-tasks-and-home-lights-by-chat/ ## How to answer factual questions from Wikipedia and web search with Jev Chat Repo: [w3cj/jev-chat](https://github.com/w3cj/jev-chat) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Get sourced facts from Wikipedia and Brave Search, with no model writing the answer. **Who it is for:** A researcher or student who wants short factual answers pulled from a reference source and does not want a summary invented by a model. **The problem:** General chatbots blend memory and guesses into answers, so users cannot tell which facts came from a source. **The outcome:** Factual answers such as a mountain height or a podcast host, each built from Wikipedia or search tool data. ### Steps 1. Run `pnpm install` in the repo folder. 2. Run `cp .env.example .env`, set TYPESAFE_API_KEY, and add BRAVE_API_KEY if you want web search. 3. Run `pnpm dev` and open http://localhost:5173. 4. Type "How tall is Mount Rainier?" and "Who hosts the Syntax podcast?". 5. Open the inspector pane for a reply to see which tool was called and what it returned. ### Prompt to paste into your agent ```text In the jev-chat repo, run pnpm dev with TYPESAFE_API_KEY set in .env, then send "How tall is Mount Rainier?" to the chat at http://localhost:5173. Report the reply and the tool named in the inspector. Verify that the height value appears in the tool result shown in the inspector. ``` ### Caveat Without BRAVE_API_KEY, only the no-key servers such as Wikipedia answer, so web search questions will not return results. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/w3cj-jev-chat/answer-facts-from-wikipedia-and-web-search/ ## How to trace why the chat bar picked a tool with Jev Chat Repo: [w3cj/jev-chat](https://github.com/w3cj/jev-chat) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Use the inspector to read every classifier answer and probability behind a reply. **Who it is for:** A developer tuning or debugging tool routing who needs to know why a message went to one tool and not another. **The problem:** When a routed reply is wrong, it is hard to tell whether the classifier, the spell check or the policy caused it. **The outcome:** A clear record for any reply showing the request, each question, each probability and the action the code took. ### Steps 1. Run `pnpm dev` and open http://localhost:5173. 2. Send an ambiguous message, such as a short request that could fit two tools. 3. Open the inspector pane for that reply. 4. Check the original message, any spell or follow-up changes, and the probabilities for each tool choice. 5. Run `pnpm test` to confirm the unit tests still pass after any change. ### Prompt to paste into your agent ```text In the jev-chat repo, run pnpm dev, send the message to the chat at http://localhost:5173, and open the inspector for the reply. Report the tool choice probabilities and the policy action. Verify that the inspector shows the original message and the answer to each question. ``` ### Caveat The inspector shows what Jev answered and what the code did with it, but it does not change how Jev classifies a message. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/w3cj-jev-chat/trace-why-a-tool-was-picked/