The Issue: The Agent executes commands successfully, but they run in the Base/System environment instead of the project’s virtual environment. It ignores the active .venv shown in the editor.
I have so far failed in figuring out a correct fix for this. Any advice is welcome.
Hello,
Thank you for reporting this.
To ensure the Agent consistently uses your project’s virtual environment, you should define the environment path explicitly in your Workspace Rules.
- Create or update a rule file (e.g., .agent/rules/env.md) with the following instruction:
# Environment Configuration
When executing Python commands, ALWAYS use the interpreter located at:
`./.venv/bin/python` (or `.\.venv\Scripts\python.exe` on Windows).
Do not run `python` directly from the global path.
- Alternatively, you can force the agent to activate the environment at the start of its task sequence by adding this to your prompt: "Activate the virtual environment in .venv before running any scripts.