LM: todo!() panics on AssistantContent::Image

[darinkishore]

Problem

lm/mod.rs has two todo!() panics for AssistantContent::Image:

• Line 393 (inside execute_tool_loop)
• Line 501 (inside call)

If a model returns an image response, this panics at runtime instead of returning a proper error.

Fix

Replace with:

AssistantContent::Image(_) => {
    return Err(anyhow::anyhow!("Image responses are not yet supported"));
}

Or, if image support is planned, return a structured PredictError variant.

Related

• proper error handling #67 (proper error handling)

[darinkishore]

Confirmed — exact locations

crates/dspy-rs/src/core/lm/mod.rs:

• Line 393: AssistantContent::Image(_image) => todo!(), — inside execute_tool_loop, in the iteration match arm
• Line 501: AssistantContent::Image(_image) => todo!(), — inside call, in the initial response match arm

Both are in match arms handling the LM response type. If any model returns an image (e.g., DALL-E via tool, or multimodal output), this panics the entire runtime.

Minimal fix: replace both with return Err(anyhow::anyhow!("Image responses are not yet supported")).

Better fix (if #67 typed errors land first): return Err(LmCallError::UnsupportedContent).