校验处理图区域并明确 multipart 字节上传

将处理图真实尺寸传入视觉绑定校验,提前拒绝零尺寸和越界区域。

保持非法区域直接失败,不转换为 NeedRework。

将 image 和 mask multipart 部分改为 Part::bytes。
This commit is contained in:
2026-09-11 20:11:11 +08:00
parent e7cd008d7c
commit e7af7cf4e4
7 changed files with 89 additions and 17 deletions
@@ -190,7 +190,38 @@ mod tests {
children: vec![],
rework_count: 0,
};
assert!(validate_binding_response(&BindingResp { decisions: vec![] }, &[&node]).is_err());
assert!(
validate_binding_response(&BindingResp { decisions: vec![] }, &[&node], (1, 1))
.is_err()
);
}
#[test]
fn binding_validation_rejects_area_outside_processed_image() {
let node = SeparationNode {
id: NodeId::new("image").unwrap(),
kind: SeparationNodeKind::ImageTarget,
global_pos_x_px: 0,
global_pos_y_px: 0,
width_px: 1,
height_px: 1,
note: SeparationNote::default(),
children: vec![],
rework_count: 0,
};
let response = BindingResp {
decisions: vec![BindingDecision::Ok {
to_node: node.id.clone(),
extracted_area: BindingArea {
global_pos_x_px: 1,
global_pos_y_px: 0,
width_px: 1,
height_px: 1,
},
}],
};
let error = validate_binding_response(&response, &[&node], (1, 1)).unwrap_err();
assert!(error.contains("超出处理图边界"));
}
#[test]
@@ -236,6 +267,7 @@ mod tests {
advice: note.to_string(),
}],
&paths,
(1, 1),
)
.unwrap();
}
@@ -275,7 +307,8 @@ mod tests {
&BindingResp {
decisions: vec![decision]
},
&[&node]
&[&node],
(1, 1)
)
.is_err());
}
@@ -310,7 +343,7 @@ mod tests {
},
}];
let paths = HashMap::from([(id.clone(), "ui/.sidecar/cut.png".to_string())]);
apply_batch_patch(&mut state, 0, &decisions, &paths).unwrap();
apply_batch_patch(&mut state, 0, &decisions, &paths, (1, 1)).unwrap();
assert_eq!(state.bound[0].node_id, id);
assert_eq!(state.trees[0].root.children.len(), 1);
}
@@ -146,6 +146,7 @@ pub fn construct_separation_state(state: &State) -> SeparationState {
pub fn validate_binding_response(
response: &BindingResp,
batch: &[&SeparationNode],
processed_dimensions: (u32, u32),
) -> Result<(), String> {
let expected = batch
.iter()
@@ -165,6 +166,11 @@ pub fn validate_binding_response(
if !seen.insert(node_id.clone()) {
return Err(format!("视觉绑定重复返回节点:{}", node_id.as_str()));
}
if let BindingDecision::Ok { extracted_area, .. } = decision {
extracted_area
.validate_in(processed_dimensions.0, processed_dimensions.1)
.map_err(|error| format!("节点 {} 的分离区域无效:{error}", node_id.as_str()))?;
}
if let BindingDecision::NeedRework { advice, .. } = decision {
if advice.trim().is_empty() {
return Err("NeedRework 必须包含问题描述".to_string());
@@ -19,9 +19,17 @@ pub(super) async fn visual_binding(
processed_url: String,
sidecar: PathBuf,
nodes: &[&SeparationNode],
processed_dimensions: (u32, u32),
) -> Result<BindingResp, String> {
let started = Instant::now();
let result = visual_binding_inner(source_url, processed_url, sidecar, nodes).await;
let result = visual_binding_inner(
source_url,
processed_url,
sidecar,
nodes,
processed_dimensions,
)
.await;
app_log!(
"ui_separation.visual_binding.timing outcome={} elapsed_ms={} nodes={}",
if result.is_ok() { "ok" } else { "error" },
@@ -36,6 +44,7 @@ async fn visual_binding_inner(
processed_url: String,
sidecar: PathBuf,
nodes: &[&SeparationNode],
processed_dimensions: (u32, u32),
) -> Result<BindingResp, String> {
app_log!(
"ui_separation.visual_binding.start nodes={} source_url_chars={} processed_url_chars={}",
@@ -137,7 +146,7 @@ async fn visual_binding_inner(
})
}
},
|value: &BindingResp| validate_binding_response(value, nodes),
|value: &BindingResp| validate_binding_response(value, nodes, processed_dimensions),
)
.await;
match &result {
@@ -1,9 +1,9 @@
use crate::platform_session::PlatformSessionSnapshot;
use base64::Engine as _;
use serde::Deserialize;
use std::fs;
use std::path::PathBuf;
use std::time::Instant;
use std::{fs, io::Cursor};
#[derive(Deserialize)]
struct RawEditResponse {
@@ -125,7 +125,7 @@ async fn raw_extract_inner(
pub(super) async fn write_processed_image(
processed_url: String,
target: PathBuf,
) -> Result<(), String> {
) -> Result<(u32, u32), String> {
let started = Instant::now();
let result = write_processed_image_inner(processed_url, target).await;
app_log!(
@@ -136,7 +136,10 @@ pub(super) async fn write_processed_image(
result
}
async fn write_processed_image_inner(processed_url: String, target: PathBuf) -> Result<(), String> {
async fn write_processed_image_inner(
processed_url: String,
target: PathBuf,
) -> Result<(u32, u32), String> {
app_log!(
"ui_separation.processed_image.write.start target_file={} data_url_chars={}",
target
@@ -153,6 +156,11 @@ async fn write_processed_image_inner(processed_url: String, target: PathBuf) ->
let processed_bytes = base64::engine::general_purpose::STANDARD
.decode(encoded)
.map_err(|error| format!("解析处理图失败:{error}"))?;
let dimensions = image::ImageReader::new(Cursor::new(processed_bytes.as_slice()))
.with_guessed_format()
.map_err(|error| format!("识别处理图格式失败:{error}"))?
.into_dimensions()
.map_err(|error| format!("读取处理图尺寸失败:{error}"))?;
let byte_len = processed_bytes.len();
fs::write(&target, processed_bytes)
.map_err(|error| format!("写入处理图失败:{}: {error}", target.display()))
@@ -165,6 +173,7 @@ async fn write_processed_image_inner(processed_url: String, target: PathBuf) ->
.unwrap_or("<unknown>"),
byte_len
);
dimensions
})
})
.await
@@ -153,18 +153,25 @@ pub(crate) async fn separate_ui_impl(
}
};
let processed_path = sidecar.join(format!("processed-{}.png", separation.bound.len()));
if let Err(error) =
extract::write_processed_image(processed_url.clone(), processed_path.clone()).await
let processed_dimensions = match extract::write_processed_image(
processed_url.clone(),
processed_path.clone(),
)
.await
{
app_log!("ui_separation.error stage=processed_image_write tree_index={} batch_index={} error={error}", tree_index, batch_index);
write_separation_state(state_path.clone(), separation.clone()).await?;
return Err(error);
}
Ok(dimensions) => dimensions,
Err(error) => {
app_log!("ui_separation.error stage=processed_image_write tree_index={} batch_index={} error={error}", tree_index, batch_index);
write_separation_state(state_path.clone(), separation.clone()).await?;
return Err(error);
}
};
let binding = match binding::visual_binding(
source_url.clone(),
processed_url,
sidecar.clone(),
&batch,
processed_dimensions,
)
.await
{
@@ -225,7 +232,13 @@ pub(crate) async fn separate_ui_impl(
write_separation_state(state_path.clone(), separation.clone()).await?;
return Err(error);
}
patch::apply_batch_patch(&mut separation, tree_index, &binding.decisions, &cut_paths)?;
patch::apply_batch_patch(
&mut separation,
tree_index,
&binding.decisions,
&cut_paths,
processed_dimensions,
)?;
write_separation_state(state_path.clone(), separation.clone()).await?;
app_log!(
"ui_separation.batch_completed tree_index={} batch_index={} cuts={} bound={} problematic={} elapsed_ms={}",
@@ -11,6 +11,7 @@ pub fn apply_batch_patch(
tree_index: usize,
decisions: &[BindingDecision],
cut_paths: &HashMap<NodeId, String>,
processed_dimensions: (u32, u32),
) -> Result<(), String> {
app_log!(
"ui_separation.batch_patch.start tree_index={} decisions={} cut_paths={}",
@@ -34,6 +35,7 @@ pub fn apply_batch_patch(
decisions: decisions.to_vec(),
},
&batch,
processed_dimensions,
)?;
let rework_counts = batch_nodes
.iter()
@@ -139,7 +139,7 @@ pub async fn create_vector_engine_raw_image_edit(
// GPT-Image-2 原生只返回 b64_json;不要添加 URL 下载或 response_format 兼容分支。
.part(
"image",
Part::stream_with_length(image_bytes.clone(), image_bytes.len() as u64)
Part::bytes(image_bytes.to_vec())
.file_name(image_file_name)
.mime_str(image_mime_type.as_str())
.map_err(|error| invalid_request(failure_context, error.to_string()))?,
@@ -156,7 +156,7 @@ pub async fn create_vector_engine_raw_image_edit(
if let Some(mask) = options.mask {
form = form.part(
"mask",
Part::stream_with_length(mask.bytes.clone(), mask.bytes.len() as u64)
Part::bytes(mask.bytes.to_vec())
.file_name(mask.file_name)
.mime_str(mask.mime_type.as_str())
.map_err(|error| invalid_request(failure_context, error.to_string()))?,