Good catch. We'd have to integrate with jinja2 (or similar) and tokenzize as we format the context, so that we know which spans are instructions and which spans

Good catch. We'd have to integrate with jinja2 (or similar) and tokenzize as we format the context, so that we know which spans are instructions and which spans are data. Which makes it more complex but still very achievable.