# AI use cases for Agav (prapaa-ai/agav) Repo: [prapaa-ai/agav](https://github.com/prapaa-ai/agav) | Use cases: 4 | Page: https://dedupedai.com/use-cases/prapaa-ai-agav/ Agav is a terminal-native AI coding agent that inspects code, edits files, runs tests, and verifies changes in real repositories. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Inspect, edit and test a repo from the terminal: https://dedupedai.com/use-cases/prapaa-ai-agav/inspect-edit-test-repo-from-terminal/ - Run a scripted code review in CI: https://dedupedai.com/use-cases/prapaa-ai-agav/non-interactive-code-review-in-ci/ - Explain an error log by piping it to the agent: https://dedupedai.com/use-cases/prapaa-ai-agav/explain-error-log-from-stdin/ - Run the agent against a local Ollama model: https://dedupedai.com/use-cases/prapaa-ai-agav/run-agent-with-local-ollama-model/ ## How to inspect, edit and test a repo from the terminal with Agav Repo: [prapaa-ai/agav](https://github.com/prapaa-ai/agav) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Ask an agent to read code, change files, run tests and verify the result inside your own project. **Who it is for:** A developer working in a real repository who wants help fixing or changing code without leaving the terminal. **The problem:** Switching between an editor, a chat window and a test runner slows down small code changes and makes it hard to confirm the result. **The outcome:** The agent reads the relevant files, proposes and applies edits, runs the tests, and you can revert an edit with /undo. ### Steps 1. Install Agav with `curl -fsSL https://agav.dev/install.sh | bash`. 2. Open a terminal in your repository folder. 3. Start the agent with `agav`, or choose a provider with `agav --provider openai --model gpt-4o`. 4. Ask for a specific change, such as fixing one failing test, and let the agent inspect the files and run the tests. 5. Approve or reject each destructive action when prompted, and use `/undo` to revert an edit you do not want. ### Prompt to paste into your agent ```text Using the Agav repo at prapaa-ai/agav, run `agav` inside and ask it to fix . Expected output: a list of edited files and the test run result. Verify that the test command the agent ran passes after the edit. ``` ### Caveat Agav is an experimental beta, so the README warns of breaking changes and rough edges. Risk, remote script: The install step pipes a script from Agav.dev directly into bash, so review the script first if your policy requires it. Risk, credentials: Cloud providers need your own API key, and the README does not say how keys are entered, so check the docs before pasting one. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. 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/prapaa-ai-agav/inspect-edit-test-repo-from-terminal/ ## How to run a scripted code review in CI with Agav Repo: [prapaa-ai/agav](https://github.com/prapaa-ai/agav) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README Call Agav run from a pipeline to review a folder and produce output a script can read. **Who it is for:** A CI maintainer who wants an automatic review step on code in a project folder for each build. **The problem:** Manual review of every change does not scale, and interactive tools do not fit a pipeline. **The outcome:** The pipeline prints a review of the target folder, and the result can be captured for later steps. ### Steps 1. Install Agav in the CI job with `curl -fsSL https://agav.dev/install.sh | bash`. 2. Make sure the provider credentials your model needs are available to the job. 3. Run `agav run "review the code in src/"` from the repository root. 4. Enable the JSON Schema output option described in the CLI reference at docs.agav.dev/reference/cli so a later step can parse the result. (inferred, check the README) 5. Fail or pass the job based on the review output your team defines. ### Prompt to paste into your agent ```text Using the Agav repo at prapaa-ai/agav, run `agav run "review the code in "` from in a non-interactive shell. Expected output: a review text or structured output for that folder. Verify that the command exits and the output names at least one file inside . ``` ### Caveat The README names the JSON Schema output option but does not show its flag, so confirm the exact syntax in the CLI reference before wiring it into a pipeline. Risk, credentials: The CI job needs model provider credentials stored as secrets, and the README does not describe how they are passed to the binary. Risk, remote script: The install step downloads and runs a script from Agav.dev inside your CI environment. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. 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/prapaa-ai-agav/non-interactive-code-review-in-ci/ ## How to explain an error log by piping it to the agent with Agav Repo: [prapaa-ai/agav](https://github.com/prapaa-ai/agav) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Pipe a log file into Agav with a question to get a plain explanation of the error. **Who it is for:** A developer or operator who has a build or runtime error log and needs to understand it quickly. **The problem:** Long error logs are hard to read under time pressure, and searching them by hand takes time. **The outcome:** A plain-language explanation of the error printed to the terminal, based on the piped log. ### Steps 1. Install Agav with `curl -fsSL https://agav.dev/install.sh | bash`. 2. Save the failing output to a file, for example by copying it into error.log. 3. Run `cat error.log | agav -P "explain this error"`. 4. Read the explanation and check it against the lines in the log that it cites. ### Prompt to paste into your agent ```text Using the Agav repo at prapaa-ai/agav, run `cat | agav -P "explain this error"` in . Expected output: a short explanation of the error in the log. Verify that every cause the explanation names appears in the log file. ``` ### Caveat The README shows the piped command but does not say what happens to log content on each provider, so check your provider's data terms. Risk, remote script: The install step downloads and runs a script from Agav.dev directly into bash. Risk, credentials: The log content is sent to the chosen model provider, so remove secrets from the log before piping it if the provider is a cloud service. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. 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/prapaa-ai-agav/explain-error-log-from-stdin/ ## How to run the agent against a local Ollama model with Agav Repo: [prapaa-ai/agav](https://github.com/prapaa-ai/agav) | Difficulty: Intermediate | Time: 60 minutes | Basis: From the README Point Agav at Ollama so code and prompts stay on a model you run yourself. **Who it is for:** A developer with private code who does not want to send it to a hosted model provider. **The problem:** Cloud model providers control where prompts and code go, and some projects cannot send code to them. **The outcome:** The agent answers questions and works on your repository using a locally hosted Ollama model. ### Steps 1. Install Agav with `curl -fsSL https://agav.dev/install.sh | bash`. 2. Set up Ollama on your machine and pull the model you want to use, such as llama3.2. 3. Open your repository folder and run `agav --provider ollama --model llama3.2`. 4. Ask a read-only question first, such as `what does this project do?`, to confirm the local model responds. 5. Switch to a different model during the session with `/model` if you want to compare results. ### Prompt to paste into your agent ```text Using the Agav repo at prapaa-ai/agav, run `agav --provider ollama --model ` inside and ask it what the project does. Expected output: a short description grounded in the repository files. Verify that the response came from the local Ollama server and that the named files exist in . ``` ### Caveat The README lists Ollama as a provider but does not explain how to install Ollama, so follow Ollama's own setup steps first. Risk, remote script: The Agav install step downloads and runs a script from Agav.dev directly into bash. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. 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/prapaa-ai-agav/run-agent-with-local-ollama-model/