# AI use cases for Radiant (templetongroup/radiant) Repo: [templetongroup/radiant](https://github.com/templetongroup/radiant) | Use cases: 5 | Page: https://dedupedai.com/use-cases/templetongroup-radiant/ Radiant is a free, MIT licensed local coding harness for Mac that runs chat with coding agents, a live activity feed, and a real terminal in one window. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Chat with a coding agent inside your project folder: https://dedupedai.com/use-cases/templetongroup-radiant/chat-with-agent-in-project-folder/ - Code with a local model that never leaves your Mac: https://dedupedai.com/use-cases/templetongroup-radiant/private-local-model-coding/ - Run a build loop that retries until a test command passes: https://dedupedai.com/use-cases/templetongroup-radiant/self-checking-loop-build/ - Run several coding jobs in parallel with a token budget: https://dedupedai.com/use-cases/templetongroup-radiant/parallel-graph-of-jobs/ - Review and commit only the changes you approve: https://dedupedai.com/use-cases/templetongroup-radiant/review-and-commit-agent-changes/ ## How to chat with a coding agent inside your project folder with Radiant Repo: [templetongroup/radiant](https://github.com/templetongroup/radiant) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README Install Radiant on a Mac and have an agent read, edit and run commands in one project folder. **Who it is for:** A developer on a Mac who wants a coding agent working on a real repo, with every action visible. **The problem:** Coding agents often act in a terminal or editor you cannot see, so it is hard to tell what they changed. **The outcome:** An agent that answers in one window, edits files in your chosen folder, and shows each tool call in the Activity panel. ### Steps 1. Download Radiant for Mac (Apple Silicon) from https://github.com/templetongroup/radiant/releases/latest/download/radiant.dmg. 2. Open the DMG and drag Radiant into Applications, then open it. 3. If Ollama or LM Studio is running, a local model is picked automatically. For cloud models, open ⚙ Settings and paste an API key next to a provider. 4. Click the path chip in the top bar and point the workspace folder at your project. 5. Ask the agent to fix or explain something. Approve each shell command when it asks, and check the Activity panel for file edits. ### Prompt to paste into your agent ```text Using the repo templetongroup/radiant, install the Radiant Mac app from its latest release DMG, set the workspace folder to , and ask the agent to explain the main entry file of that project. Expected output: a chat reply that names the files it read. Verify that the Activity panel lists each file read as a tool call. ``` ### Caveat Radiant only runs on a Mac, and the release DMG is for Apple Silicon, so Intel Macs need the build-from-source route. 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/templetongroup-radiant/chat-with-agent-in-project-folder/ ## How to code with a local model that never leaves your Mac with Radiant Repo: [templetongroup/radiant](https://github.com/templetongroup/radiant) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Use Ollama or LM Studio on your Mac so no key is needed and no code is sent to a cloud model. **Who it is for:** A developer working on private or unreleased code who cannot send it to a cloud provider. **The problem:** Cloud coding assistants send code off the machine, and a local model is hard to judge before a large download. **The outcome:** A chat session on a local model that Radiant has labelled for your Mac before you download it. ### Steps 1. Start Ollama or LM Studio on your Mac. 2. Open Radiant. Start a session and confirm the local model it picks automatically. 3. Check the label on each local model (Runs well, Runs tight or Won't run) before you download it. 4. Point the workspace folder at your private project and ask the agent a question about it. 5. Read the composer bar, which states the model answering, and confirm it is the local one. ### Prompt to paste into your agent ```text Using the repo templetongroup/radiant on a Mac with a local model running in Ollama or LM Studio, start a Radiant session on the folder and ask the model to list the top-level modules. Expected output: a list of modules from a local model with no API key set. Verify that the composer bar names the local model as the one answering. ``` ### Caveat Local model quality depends on your Mac, and Radiant's Runs well, Runs tight or Won't run label is its own measure of fit, not a benchmark. 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/templetongroup-radiant/private-local-model-coding/ ## How to run a build loop that retries until a test command passes with Radiant Repo: [templetongroup/radiant](https://github.com/templetongroup/radiant) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Give each step a shell check, and let failed steps retry with the reason until the check exits 0. **Who it is for:** A developer with a test or lint command who wants an agent to keep fixing a change until it passes. **The problem:** Agents often stop after one attempt, and you have to rerun and re-explain each failure yourself. **The outcome:** A loop that repeats each failed step with the failure reason until the check passes, or stops when a check reports blocked or the budget runs out. ### Steps 1. Open a session with the workspace folder set to your project. 2. Create a loop with one step per change. Give each step a check: a shell command that must exit 0, such as your own test command. 3. Set a spend budget. Long builds have no round cap, so the budget is the ceiling. 4. Start the loop and watch the Activity panel as each failed step goes round again with its reason. 5. If a check answers blocked, the loop ends at once. Review the reason before you restart it. ### Prompt to paste into your agent ```text Using the repo templetongroup/radiant, set up a loop in a Radiant session on where each step must pass a shell test command that exits 0, and run it. Expected output: a loop report showing the rounds each step took and whether each check passed. Verify that the report lists the spend against the budget you set. ``` ### Caveat Loop checks only prove what your command tests, so a weak test command will let a wrong change pass. 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/templetongroup-radiant/self-checking-loop-build/ ## How to run several coding jobs in parallel with a token budget with Radiant Repo: [templetongroup/radiant](https://github.com/templetongroup/radiant) | Difficulty: Advanced | Time: 45 minutes | Basis: From the README Build a graph of jobs where only real dependencies wait, and stop a repeating graph at a token budget. **Who it is for:** A developer splitting a feature into independent stages who wants them done faster and within a cost limit. **The problem:** Running coding jobs one after another wastes time when many of them do not depend on each other. **The outcome:** A graph where independent steps run at the same time, with tokens and dollar cost shown for each step and round. ### Steps 1. Create a graph with one step for each job, such as a docs update, a test addition and a refactor. 2. Tick only the steps that a job actually reads. Everything left unticked runs at the same time. 3. For a repeating graph, set a token budget so it can stop before it overspends. 4. Start the graph and open the run view to see tokens and dollars at list price for each step and round. 5. Check the output of each step in the Activity panel before you merge the results. ### Prompt to paste into your agent ```text Using the repo templetongroup/radiant, build a graph in a Radiant session on with three independent jobs and run it. Expected output: a run view listing the tokens and dollar cost for each job. Verify that the independent jobs started at the same time, not one after another. ``` ### Caveat Dollar figures in the run view are estimates at list price, not your actual provider bill. 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/templetongroup-radiant/parallel-graph-of-jobs/ ## How to review and commit only the changes you approve with Radiant Repo: [templetongroup/radiant](https://github.com/templetongroup/radiant) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README Check each file an agent changed with diffs, commit the ticked files, and undo if needed. **Who it is for:** A developer who lets an agent edit a repo and wants to review and commit only the parts they accept. **The problem:** Agent edits can be spread across many files, and it is easy to commit something you did not check. **The outcome:** A commit made from only the files you ticked, with a one-hour Restore available for undo. ### Steps 1. Open the Changes panel in the right-hand tab for the chat's folder. 2. Check the M, A, D, R and U letters and the +/- line counts for each file. 3. Open the inline diff for each file you want to review. 4. Tick the files you accept and write a message in the commit box. 5. Use Undo to roll the files back if a change is wrong. Nothing is pushed, so push the commit yourself. ### Prompt to paste into your agent ```text Using the repo templetongroup/radiant, open the Changes panel for a Radiant session on , tick the files that pass review, and commit them with a message you write. Expected output: a local commit containing only the ticked files. Verify that the unticked files still show as changed in the Changes panel. ``` ### Caveat Radiant commits locally only and never pushes, so you must push the commit yourself if you want it on a remote. 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/templetongroup-radiant/review-and-commit-agent-changes/