# AI use cases for Adelie (lmliheng/Adelie) Repo: [lmliheng/Adelie](https://github.com/lmliheng/Adelie) | Use cases: 4 | Page: https://dedupedai.com/use-cases/lmliheng-adelie/ Adelie is a local-first, open-source multi-agent app development platform, forked from PenguinHarness, that builds, evaluates, optimises and deploys agent applications on your own machine or server. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Turn a docs URL into a cited Q&A app: https://dedupedai.com/use-cases/lmliheng-adelie/one-sentence-rag-docs-app/ - Improve an agent version by version with benchmarks: https://dedupedai.com/use-cases/lmliheng-adelie/self-evolve-agent-with-benchmark/ - Have an agent pull data from a logged-in website: https://dedupedai.com/use-cases/lmliheng-adelie/browser-agent-data-extraction/ - Set up the data-analysis plugin for dataset questions: https://dedupedai.com/use-cases/lmliheng-adelie/data-analysis-plugin-setup/ ## How to turn a docs URL into a cited Q&A app with Adelie Repo: [lmliheng/Adelie](https://github.com/lmliheng/Adelie) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Describe a docs-expert Q&A app in one sentence and get a runnable agent app. **Who it is for:** A developer or team lead who needs an internal help bot over a public documentation repo and does not want to scaffold a RAG app by hand. **The problem:** Building a retrieval-augmented Q&A app from scratch means collecting sources, wiring retrieval, adding citations and writing run instructions. **The outcome:** Adelie produces the whole agent application (scaffold, code and run instructions) in one go, with retrieval-based answers that cite the source files. ### Steps 1. Install the command line with `curl -fsSL https://github.com/lmliheng/Adelie/releases/latest/download/install.sh | sh` on Linux or macOS, or `npm install -g @adelie/cli` on any platform with Node 24 or newer. 2. Run `penguin web` and open the first-login link the server prints, then set the password for the built-in admin account. 3. On the Models page, add an API key for one model provider. 4. Describe the app in one sentence, naming the docs URL and the expert persona, and ask for a RAG Q&A app with cited sources. 5. Open the generated app, check that answers link back to the original files, and run it with the instructions Adelie returns. ### Prompt to paste into your agent ```text Using the Adelie repo (lmliheng/Adelie), ask the local Adelie instance to build a RAG Q&A app over that answers as and cites source files. Expected output: a runnable app folder with run instructions. Verify that a sample answer includes a citation that opens the original file. ``` ### Caveat The Windows installer is not code-signed yet, so SmartScreen shows a warning on first run, and macOS has no build yet. Risk, remote script: The Linux and macOS one-line installer is a curl pipe to sh, so read the script at its release URL first if your policy requires it. 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/lmliheng-adelie/one-sentence-rag-docs-app/ ## How to improve an agent version by version with benchmarks with Adelie Repo: [lmliheng/Adelie](https://github.com/lmliheng/Adelie) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Run a benchmark, find the drop in score, and ship version N+1 with replayable traces. **Who it is for:** An AI engineer whose agent scores lower after a change and needs to find which step caused the regression. **The problem:** Agent quality drifts between versions and it is hard to see which request or round caused the score to fall. **The outcome:** The agent is evaluated on a Benchmark, the weak area is identified, and a new version is produced with a snapshot taken before each round and every request replayable in the trace viewer. ### Steps 1. Install and start Adelie with `npm install -g @adelie/cli` and then `penguin web`. 2. Open the agent you want to improve and run a Benchmark from the interface. 3. Open the trace viewer and find the requests where the score dropped. 4. Note the snapshot taken before the round, then produce version N+1 of the agent through the self-evolution engine. 5. Run the Benchmark again on version N+1 and compare the results with the previous round. ### Prompt to paste into your agent ```text Using the Adelie repo (lmliheng/Adelie), run a Benchmark on the agent in the local web interface, locate the requests whose score fell in the trace viewer, and produce version N+1. Expected output: a new agent version with a Benchmark score for each round. Verify that the snapshot from before the round can be restored. ``` ### Caveat The README shows the benchmark chart from upstream PenguinHarness measurements and the upstream roadmap lists the benchmark suite as not yet released, so check what the Benchmark feature covers in your build before relying on it. 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/lmliheng-adelie/self-evolve-agent-with-benchmark/ ## How to have an agent pull data from a logged-in website with Adelie Repo: [lmliheng/Adelie](https://github.com/lmliheng/Adelie) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Let an agent read pages, click and type, and extract data such as order history. **Who it is for:** An operations or finance user who needs data from a website they already sign in to, such as order records, and wants an agent to collect it. **The problem:** Copying records from a web account by hand is slow, and the site has no export the user can use. **The outcome:** The agent uses the built-in browser to read pages, click and type, and return extracted data, signed in with credentials imported from the user's own browser. ### Steps 1. Install Adelie with `npm install -g @adelie/cli` or use the desktop installer for your platform. 2. Start the interface with `penguin web` and open the plugin market to confirm the `browser-automation` plugin is installed. 3. Import the signed-in credentials from your own browser for the target site. 4. Ask the agent to open the site through `penguin browser` and extract the records you need into a file. 5. Check a sample of the extracted rows against the site before using the full export. ### Prompt to paste into your agent ```text Using the Adelie repo (lmliheng/Adelie), use the browser-automation plugin and `penguin browser` to open , extract into a CSV file, and report the row count. Expected output: a CSV of extracted records. Verify three rows against the live site before accepting the export. ``` ### Caveat The README only names Amazon orders as an example of extraction, so confirm that other sites work before relying on them. Risk, credentials: The agent uses sign-in credentials imported from your browser, so limit the scope to accounts you own and review what is stored in the data root at ~/.adelie/data. 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/lmliheng-adelie/browser-agent-data-extraction/ ## How to set up the data-analysis plugin for dataset questions with Adelie Repo: [lmliheng/Adelie](https://github.com/lmliheng/Adelie) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Install the data-analysis plugin and ask an agent questions about a dataset. **Who it is for:** An analyst who needs an agent to answer questions about a local dataset without writing code. **The problem:** Setting up an agent with the right tools for data analysis takes time, and the tool set has to fit the task. **The outcome:** The data-analysis plugin is installed from the plugin market and the agent answers dataset questions through the web interface or terminal chat. ### Steps 1. Install Adelie with `npm install -g @adelie/cli` and start it with `penguin web`. 2. Open the plugin market page and install the `data-analysis` plugin in one click. 3. Open a new chat in the web interface or run `penguin chat` in the terminal. 4. Attach or point the agent at your dataset and ask a specific question about it. 5. Check the agent's answer against a value you compute by hand from the same file. ### Prompt to paste into your agent ```text Using the Adelie repo (lmliheng/Adelie), install the data-analysis plugin from the plugin market, then ask the agent in penguin chat to answer about . Expected output: a short answer with the figure used. Verify one figure by recomputing it from the file. ``` ### Caveat The README makes accuracy claims on a data-analysis set measured by the upstream authors, so verify results on your own data. 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/lmliheng-adelie/data-analysis-plugin-setup/