app-server: expose image-context tri-state

codex-rs/protocol/src/models.rs

@@ -72,6 +72,12 @@ pub enum ContentItem {
     OutputText { text: String },
 }
 
+impl ContentItem {
+    pub fn is_input_image(&self) -> bool {
+        matches!(self, ContentItem::InputImage { .. })
+    }
+}
+
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, JsonSchema, TS)]
 #[serde(rename_all = "snake_case")]
 /// Classifies an assistant message as interim commentary or final answer text.
@@ -195,6 +201,24 @@ pub enum ResponseItem {
     Other,
 }
 
+impl ResponseItem {
+    pub fn has_input_image(&self) -> bool {
+        match self {
+            ResponseItem::Message { content, .. } => {
+                content.iter().any(ContentItem::is_input_image)
+            }
+            ResponseItem::FunctionCallOutput { output, .. } => {
+                output.content_items().is_some_and(|items| {
+                    items
+                        .iter()
+                        .any(FunctionCallOutputContentItem::is_input_image)
+                })
+            }
+            _ => false,
+        }
+    }
+}
+
 pub const BASE_INSTRUCTIONS_DEFAULT: &str = include_str!("prompts/base_instructions/default.md");
 
 /// Base instructions for the model in a thread. Corresponds to the `instructions` field in the ResponsesAPI.
@@ -795,6 +819,12 @@ pub enum FunctionCallOutputContentItem {
     InputImage { image_url: String },
 }
 
+impl FunctionCallOutputContentItem {
+    pub fn is_input_image(&self) -> bool {
+        matches!(self, FunctionCallOutputContentItem::InputImage { .. })
+    }
+}
+
 /// Converts structured function-call output content into plain text for
 /// human-readable surfaces.
 ///

codex-rs/protocol/src/openai_models.rs

@@ -72,6 +72,35 @@ pub enum InputModality {
     Image,
 }
 
+pub const INPUT_MODALITY_TEXT_MASK: i64 = 1;
+pub const INPUT_MODALITY_IMAGE_MASK: i64 = 1 << 1;
+
+pub fn input_modality_to_mask(modality: InputModality) -> i64 {
+    match modality {
+        InputModality::Text => INPUT_MODALITY_TEXT_MASK,
+        InputModality::Image => INPUT_MODALITY_IMAGE_MASK,
+    }
+}
+
+pub fn input_modalities_to_mask(modalities: &[InputModality]) -> i64 {
+    modalities
+        .iter()
+        .copied()
+        .map(input_modality_to_mask)
+        .fold(0, |mask, bit| mask | bit)
+}
+
+pub fn input_modalities_from_mask(mask: i64) -> Vec<InputModality> {
+    let mut modalities = Vec::new();
+    if mask & INPUT_MODALITY_TEXT_MASK != 0 {
+        modalities.push(InputModality::Text);
+    }
+    if mask & INPUT_MODALITY_IMAGE_MASK != 0 {
+        modalities.push(InputModality::Image);
+    }
+    modalities
+}
+
 /// Backward-compatible default when `input_modalities` is omitted on the wire.
 ///
 /// Legacy payloads predate modality metadata, so we conservatively assume both text and images are