Skip to content

EXAMPLE.md

Latest commit

 

History

History
140 lines (94 loc) · 3.45 KB

EXAMPLE.md

File metadata and controls

140 lines (94 loc) · 3.45 KB

 

Interactive Example

This is a hands-on example workflow for using mini-coding-agent with Ollama on a small Python project.

The flow is:

  1. create a fresh repo
  2. launch the agent
  3. implement binary_search.py
  4. edit the implementation
  5. add pytest tests
  6. run tests
  7. fix anything that fails

This example assumes:

  • ollama serve is already running
  • you already pulled a model such as qwen3.5:4b (e.g., via ollama pull qwen3.5:4b )
  • you already cloned or forked rasbt/mini-coding-agent
  • you already ran uv sync in your local mini-coding-agent folder

If you have sufficient memory, consider using a larger Qwen 3.5 model instead:

 

1. Create a fresh repo

cd mini-coding-agent
mkdir -p ./tmp/binary-search-repo
cd ./tmp/binary-search-repo
git init

At this point the repo is basically empty:

ls -la

 

2. Launch the agent

Open the agent from your mini-coding-agent clone, but point it at the new repo:

cd mini-coding-agent
uv run mini-coding-agent \
  --cwd ./tmp/binary-search-repo \
  --model "qwen3.5:4b"

 

3. Ask it to implement binary search

At the mini-coding-agent> prompt, paste:

Inspect this repository and create a minimal binary_search.py file. Implement an iterative binary_search(nums, target) function for a sorted list of integers. Return the index if the target exists and -1 if it does not. Keep the code very small.

After the agent finishes, inspect the result in another terminal or code editor. The contents are shown below:

 

4. Ask it to edit the implementation

Now make a small follow-up change. Back in the agent REPL, paste:

Update binary_search.py so it raises ValueError if the input list is not sorted in ascending order. Keep the implementation simple.

Check the file again:

 

5. Ask it to add unit tests

Back in the REPL, paste:

Create test_binary_search.py with pytest tests for found, missing, empty list, first element, last element, and unsorted input raising ValueError. Keep the tests small and readable.

Inspect the new test file:

 

6. Ask it to run the tests

Back in the REPL, paste:

Run pytest for this repo. If any test fails, fix the code or tests and rerun until everything passes.

You can also verify manually in a different terminal window:

uv run pytest tmp/binary-search-repo

 

7. Inspect the final repo state

Check what changed:

cd mini-coding-agent
cd ./tmp/binary-search-repo
git status --short

You should now have:

  • README.md
  • binary_search.py
  • test_binary_search.py

 

8. Useful interactive commands

While the agent is running, these commands are available:

  • /help shows the available slash commands and what each one does.
  • /memory prints the agent's distilled working memory for the current session.
  • /session shows the path to the saved session JSON file on disk.
  • /reset clears the current conversation history and working memory.
  • /exit leaves the interactive agent.