From dd592d918d56b122fcca2a273c509de8063c130f Mon Sep 17 00:00:00 2001
From: rob thijssen <grenade@rob.tn>
Date: Thu, 4 Jun 2026 13:57:43 +0300
Subject: [PATCH] =?UTF-8?q?test(neuron):=20C2=20=E2=80=94=20guard=20Respon?=
 =?UTF-8?q?ses=E2=86=92chat=20image=20translation=20contract?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The Responses request translator already emits the chat `image_url`
Parts array Stage B5's vision path consumes, and the non-streaming
(`chat_completion`) and streaming (`responses_stream` → `inference_stream`,
Stage C1) Responses paths both route image content to the vision-aware
prefill — so vision works end-to-end through `/v1/responses` with no
translator change required.

Add a multi-image test asserting order preservation and that the
`detail` hint is tolerated (and dropped, since chat image_url has no
analogue), locking the translator's output to the exact
`image_url.url` shape `extract_images_from_request` walks.

Closes part of #16 (Stage C2).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---
 crates/neuron/src/wire/openai_responses.rs | 48 ++++++++++++++++++++++
 1 file changed, 48 insertions(+)

diff --git a/crates/neuron/src/wire/openai_responses.rs b/crates/neuron/src/wire/openai_responses.rs
index 8821910..5cec8ac 100644
--- a/crates/neuron/src/wire/openai_responses.rs
+++ b/crates/neuron/src/wire/openai_responses.rs
@@ -646,6 +646,54 @@ mod tests {
         assert_eq!(parts[1]["image_url"]["url"], "data:image/png;base64,AAA=");
     }
 
+    #[test]
+    fn multiple_images_translate_in_order_and_tolerate_detail() {
+        // C2: a Responses request carrying several InputImage parts
+        // (with `detail` set) must translate to a chat Parts array that
+        // preserves image order and the `image_url.url` shape the chat
+        // vision path (`extract_images_from_request`) walks. The
+        // `detail` hint has no chat-completions analogue we forward, so
+        // it's dropped — but it must not break translation.
+        let req = ResponsesRequest {
+            model: "m".into(),
+            input: ResponsesInput::Items(vec![ResponsesInputItem::Message {
+                role: "user".into(),
+                content: ResponsesMessageContent::Parts(vec![
+                    ResponsesContentPart::InputText {
+                        text: "compare these".into(),
+                    },
+                    ResponsesContentPart::InputImage {
+                        image_url: "data:image/png;base64,FIRST".into(),
+                        detail: Some("high".into()),
+                    },
+                    ResponsesContentPart::InputImage {
+                        image_url: "data:image/png;base64,SECOND".into(),
+                        detail: None,
+                    },
+                ]),
+            }]),
+            instructions: None,
+            stream: false,
+            max_output_tokens: None,
+            temperature: None,
+            top_p: None,
+            previous_response_id: None,
+            extra: Value::Object(Default::default()),
+        };
+        let chat = request_to_chat(req).unwrap();
+        let parts = match &chat.messages[0].content {
+            MessageContent::Parts(p) => p.clone(),
+            other => panic!("expected Parts, got {other:?}"),
+        };
+        // text + two images, in input order.
+        assert_eq!(parts.len(), 3);
+        assert_eq!(parts[0]["type"], "text");
+        assert_eq!(parts[1]["image_url"]["url"], "data:image/png;base64,FIRST");
+        assert_eq!(parts[2]["image_url"]["url"], "data:image/png;base64,SECOND");
+        // `detail` is not forwarded into the chat image_url object.
+        assert!(parts[1]["image_url"].get("detail").is_none());
+    }
+
     #[test]
     fn text_only_parts_collapse_to_string() {
         let req = ResponsesRequest {