mirror of
https://github.com/qingchencloud/clawpanel.git
synced 2026-05-27 11:20:04 +08:00
fix(assistant): 模型测试按钮改用流式累积 + 增强诊断信息
- 测试请求切换到 stream: true + SSE 累积，绕开部分兼容网关 non-streaming 分支对某些模型返回 200 + 空 body 的已知 bug， 行为与真实对话路径一致 - 后端 test_model_verbose 显式设置 Accept-Encoding: identity， 避免压缩协商带来的解码风险 - 用 resp.bytes() + 严格 UTF-8 decode，失败时 fallback 到 lossy 字符串 + 前 200 字节 hex dump，方便定位非 UTF-8 响应 - 展开 reqwest error source 链，响应头与字节数原样返回前端 - 前端结果面板突出显示完整模型回复、固定 prompt 标注、 响应头与 raw bytes hex，方便用户自查上游问题 - scripts/dev-api.js 同步 Rust 后端行为，保证 Web/桌面两侧诊断一致
This commit is contained in:
晴天
@@ -5246,11 +5246,132 @@ pub async fn test_model(
|
||||
Ok(reply)
|
||||
}
|
||||
|
||||
/// 从 SSE 流文本中累积 OpenAI 风格的 delta.content / delta.reasoning_content
|
||||
/// 格式示例：
|
||||
/// data: {"choices":[{"delta":{"content":"你好"}}]}
|
||||
/// data: {"choices":[{"delta":{"content":"，"}}]}
|
||||
/// data: [DONE]
|
||||
fn extract_sse_reply(text: &str) -> String {
|
||||
let mut content = String::new();
|
||||
let mut reasoning = String::new();
|
||||
let mut saw_data_line = false;
|
||||
for line in text.lines() {
|
||||
let data = if let Some(rest) = line.strip_prefix("data: ") {
|
||||
rest
|
||||
} else if let Some(rest) = line.strip_prefix("data:") {
|
||||
rest
|
||||
} else {
|
||||
continue;
|
||||
};
|
||||
saw_data_line = true;
|
||||
let data = data.trim();
|
||||
if data.is_empty() || data == "[DONE]" {
|
||||
continue;
|
||||
}
|
||||
if let Ok(v) = serde_json::from_str::<serde_json::Value>(data) {
|
||||
// OpenAI / 兼容后端：choices[0].delta.content
|
||||
let delta = v
|
||||
.get("choices")
|
||||
.and_then(|c| c.get(0))
|
||||
.and_then(|c| c.get("delta"));
|
||||
if let Some(d) = delta {
|
||||
if let Some(c) = d.get("content").and_then(|c| c.as_str()) {
|
||||
content.push_str(c);
|
||||
}
|
||||
if let Some(rc) = d.get("reasoning_content").and_then(|c| c.as_str()) {
|
||||
reasoning.push_str(rc);
|
||||
}
|
||||
}
|
||||
// Anthropic streaming: {"type":"content_block_delta","delta":{"type":"text_delta","text":"..."}}
|
||||
if v.get("type").and_then(|t| t.as_str()) == Some("content_block_delta") {
|
||||
if let Some(c) = v
|
||||
.get("delta")
|
||||
.and_then(|d| d.get("text"))
|
||||
.and_then(|t| t.as_str())
|
||||
{
|
||||
content.push_str(c);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
if !saw_data_line {
|
||||
return String::new();
|
||||
}
|
||||
if !content.is_empty() {
|
||||
content
|
||||
} else if !reasoning.is_empty() {
|
||||
format!("[reasoning] {reasoning}")
|
||||
} else {
|
||||
String::new()
|
||||
}
|
||||
}
|
||||
|
||||
/// 从单个 JSON 响应中提取 reply（兼容 OpenAI / Anthropic / Gemini / DashScope 非流式）
|
||||
fn extract_single_json_reply(text: &str) -> String {
|
||||
serde_json::from_str::<serde_json::Value>(text)
|
||||
.ok()
|
||||
.and_then(|v| {
|
||||
if let Some(arr) = v.get("content").and_then(|c| c.as_array()) {
|
||||
let text = arr
|
||||
.iter()
|
||||
.filter(|b| b.get("type").and_then(|t| t.as_str()) == Some("text"))
|
||||
.filter_map(|b| b.get("text").and_then(|t| t.as_str()))
|
||||
.collect::<Vec<_>>()
|
||||
.join("");
|
||||
if !text.is_empty() {
|
||||
return Some(text);
|
||||
}
|
||||
}
|
||||
if let Some(t) = v
|
||||
.get("candidates")
|
||||
.and_then(|c| c.get(0))
|
||||
.and_then(|c| c.get("content"))
|
||||
.and_then(|c| c.get("parts"))
|
||||
.and_then(|p| p.get(0))
|
||||
.and_then(|p| p.get("text"))
|
||||
.and_then(|t| t.as_str())
|
||||
.filter(|s| !s.is_empty())
|
||||
{
|
||||
return Some(t.to_string());
|
||||
}
|
||||
if let Some(msg) = v
|
||||
.get("choices")
|
||||
.and_then(|c| c.get(0))
|
||||
.and_then(|c| c.get("message"))
|
||||
{
|
||||
let content = msg.get("content").and_then(|c| c.as_str()).unwrap_or("");
|
||||
if !content.is_empty() {
|
||||
return Some(content.to_string());
|
||||
}
|
||||
if let Some(rc) = msg
|
||||
.get("reasoning_content")
|
||||
.and_then(|c| c.as_str())
|
||||
.filter(|s| !s.is_empty())
|
||||
{
|
||||
return Some(format!("[reasoning] {rc}"));
|
||||
}
|
||||
}
|
||||
if let Some(t) = v
|
||||
.get("output")
|
||||
.and_then(|o| o.get("text"))
|
||||
.and_then(|t| t.as_str())
|
||||
.filter(|s| !s.is_empty())
|
||||
{
|
||||
return Some(t.to_string());
|
||||
}
|
||||
None
|
||||
})
|
||||
.unwrap_or_default()
|
||||
}
|
||||
|
||||
/// 测试模型（详细版 #Compat-1）：返回完整 req/resp 信息，供前端 debug 面板展示
|
||||
///
|
||||
/// 相比 test_model：
|
||||
/// - 不会因 400/422/429 等吞掉错误返回"连接正常"，一律如实回传 status + body
|
||||
/// - 返回结构化 JSON：success/status/req_url/req_body/resp_body/reply/error/elapsed_ms/used_api
|
||||
/// - 前端拿到后可以直接渲染 debug 面板，无需在 webview 里走外部 fetch（规避 status 0）
|
||||
/// - OpenAI 兼容路径使用 stream:true（绕开某些 new-api 后端的 non-streaming bug，
|
||||
/// 并与真实对话行为一致）
|
||||
#[tauri::command]
|
||||
pub async fn test_model_verbose(
|
||||
base_url: String,
|
||||
@@ -5307,20 +5428,25 @@ pub async fn test_model_verbose(
|
||||
}
|
||||
_ => {
|
||||
let url = format!("{}/chat/completions", base);
|
||||
// 关键：测试请求用 stream: true 而非 stream: false
|
||||
// 理由：部分兼容网关的 non-streaming 分支对某些模型会返回 200 + 空 body，
|
||||
// 而 streaming 分支是真实对话路径，所有 provider 都稳定支持。
|
||||
// 测试走 stream: true + SSE 累积，行为与真实对话一致。
|
||||
let body = json!({
|
||||
"model": model_id,
|
||||
"messages": [{"role": "user", "content": "你好，请用一句话回复"}],
|
||||
"max_tokens": 200,
|
||||
"stream": false
|
||||
"stream": true
|
||||
});
|
||||
let mut req = client
|
||||
.post(&url)
|
||||
.header("Accept-Encoding", "identity")
|
||||
.header("Accept", "text/event-stream")
|
||||
.json(&body);
|
||||
if !api_key.is_empty() {
|
||||
req = req.header("Authorization", format!("Bearer {api_key}"));
|
||||
}
|
||||
("Chat Completions", url, body, req)
|
||||
("Chat Completions (SSE)", url, body, req)
|
||||
}
|
||||
};
|
||||
|
||||
@@ -5353,79 +5479,104 @@ pub async fn test_model_verbose(
|
||||
|
||||
let status = resp.status();
|
||||
let status_code = status.as_u16();
|
||||
// 读取响应体：若失败（如 gzip/brotli 解码异常、非法 UTF-8）直接返回错误，不静默吞成空串
|
||||
let text = match resp.text().await {
|
||||
Ok(t) => t,
|
||||
|
||||
// 先抓取响应头（text() 会消耗 resp）—— 这是关键诊断信息：
|
||||
// Content-Encoding 告诉我们是否压缩、是 br/gzip/zstd 还是啥
|
||||
// Content-Type 告诉我们是否是 JSON / text
|
||||
// Content-Length 告诉我们服务器声明的响应体大小
|
||||
let resp_headers = {
|
||||
let mut map = serde_json::Map::new();
|
||||
for (k, v) in resp.headers().iter() {
|
||||
map.insert(
|
||||
k.to_string(),
|
||||
serde_json::Value::String(v.to_str().unwrap_or("<non-utf8>").to_string()),
|
||||
);
|
||||
}
|
||||
serde_json::Value::Object(map)
|
||||
};
|
||||
|
||||
// 读取响应体：改用 bytes() 拿原始字节（reqwest 会按 Content-Encoding 自动解压），
|
||||
// 然后自己做 UTF-8 decode。这样：
|
||||
// 1. 失败时能给出更精确的错误分类（网络错误 vs 解压错误 vs UTF-8 错误）
|
||||
// 2. UTF-8 失败时能 fallback 到 hex dump + lossy string，方便诊断
|
||||
let bytes = match resp.bytes().await {
|
||||
Ok(b) => b,
|
||||
Err(e) => {
|
||||
let mut err_chain = format!("{e}");
|
||||
let mut src: Option<&dyn std::error::Error> = std::error::Error::source(&e);
|
||||
while let Some(s) = src {
|
||||
err_chain.push_str(&format!(" → {s}"));
|
||||
src = std::error::Error::source(s);
|
||||
}
|
||||
return Ok(json!({
|
||||
"success": false,
|
||||
"status": status_code,
|
||||
"reqUrl": req_url,
|
||||
"reqBody": req_body_json,
|
||||
"respHeaders": resp_headers,
|
||||
"respBody": "",
|
||||
"respRawHex": "",
|
||||
"respByteCount": 0,
|
||||
"reply": "",
|
||||
"error": format!("读取响应体失败: {e} (可能是压缩编码未支持或非 UTF-8 响应)"),
|
||||
"error": format!("读取响应字节失败: {err_chain}"),
|
||||
"elapsedMs": elapsed_ms,
|
||||
"usedApi": used_api,
|
||||
}));
|
||||
}
|
||||
};
|
||||
let byte_count = bytes.len();
|
||||
|
||||
// 前 200 字节的 hex dump（无论成功失败都附上，方便调试）
|
||||
let hex_preview = bytes
|
||||
.iter()
|
||||
.take(200)
|
||||
.map(|b| format!("{b:02x}"))
|
||||
.collect::<Vec<_>>()
|
||||
.join(" ");
|
||||
|
||||
// 尝试严格 UTF-8 decode；失败时 fallback 到 lossy 并在 error 里带上诊断
|
||||
let text = match std::str::from_utf8(&bytes) {
|
||||
Ok(s) => s.to_string(),
|
||||
Err(e) => {
|
||||
let lossy = String::from_utf8_lossy(&bytes).into_owned();
|
||||
let ascii_preview: String = bytes
|
||||
.iter()
|
||||
.take(80)
|
||||
.map(|&b| {
|
||||
if (0x20..=0x7e).contains(&b) {
|
||||
b as char
|
||||
} else {
|
||||
'.'
|
||||
}
|
||||
})
|
||||
.collect();
|
||||
return Ok(json!({
|
||||
"success": false,
|
||||
"status": status_code,
|
||||
"reqUrl": req_url,
|
||||
"reqBody": req_body_json,
|
||||
"respHeaders": resp_headers,
|
||||
"respBody": lossy,
|
||||
"respRawHex": hex_preview,
|
||||
"respByteCount": byte_count,
|
||||
"reply": "",
|
||||
"error": format!("响应体 UTF-8 解码失败: {e} | 字节数={byte_count} | 前 80 字节 ASCII='{ascii_preview}'"),
|
||||
"elapsedMs": elapsed_ms,
|
||||
"usedApi": used_api,
|
||||
}));
|
||||
}
|
||||
};
|
||||
|
||||
// 提取 reply 文本（兼容 OpenAI / Anthropic / Gemini / DashScope）
|
||||
let reply = serde_json::from_str::<serde_json::Value>(&text)
|
||||
.ok()
|
||||
.and_then(|v| {
|
||||
if let Some(arr) = v.get("content").and_then(|c| c.as_array()) {
|
||||
let text = arr
|
||||
.iter()
|
||||
.filter(|b| b.get("type").and_then(|t| t.as_str()) == Some("text"))
|
||||
.filter_map(|b| b.get("text").and_then(|t| t.as_str()))
|
||||
.collect::<Vec<_>>()
|
||||
.join("");
|
||||
if !text.is_empty() {
|
||||
return Some(text);
|
||||
}
|
||||
}
|
||||
if let Some(t) = v
|
||||
.get("candidates")
|
||||
.and_then(|c| c.get(0))
|
||||
.and_then(|c| c.get("content"))
|
||||
.and_then(|c| c.get("parts"))
|
||||
.and_then(|p| p.get(0))
|
||||
.and_then(|p| p.get("text"))
|
||||
.and_then(|t| t.as_str())
|
||||
.filter(|s| !s.is_empty())
|
||||
{
|
||||
return Some(t.to_string());
|
||||
}
|
||||
if let Some(msg) = v
|
||||
.get("choices")
|
||||
.and_then(|c| c.get(0))
|
||||
.and_then(|c| c.get("message"))
|
||||
{
|
||||
let content = msg.get("content").and_then(|c| c.as_str()).unwrap_or("");
|
||||
if !content.is_empty() {
|
||||
return Some(content.to_string());
|
||||
}
|
||||
if let Some(rc) = msg
|
||||
.get("reasoning_content")
|
||||
.and_then(|c| c.as_str())
|
||||
.filter(|s| !s.is_empty())
|
||||
{
|
||||
return Some(format!("[reasoning] {rc}"));
|
||||
}
|
||||
}
|
||||
if let Some(t) = v
|
||||
.get("output")
|
||||
.and_then(|o| o.get("text"))
|
||||
.and_then(|t| t.as_str())
|
||||
.filter(|s| !s.is_empty())
|
||||
{
|
||||
return Some(t.to_string());
|
||||
}
|
||||
None
|
||||
})
|
||||
.unwrap_or_default();
|
||||
// 提取 reply 文本：同时兼容 SSE 流（stream:true）和单次 JSON（stream:false）
|
||||
// 优先尝试 SSE 解析（OpenAI 兼容路径现在用 stream:true），失败再回退到单 JSON
|
||||
let reply = {
|
||||
let sse_reply = extract_sse_reply(&text);
|
||||
if !sse_reply.is_empty() {
|
||||
sse_reply
|
||||
} else {
|
||||
extract_single_json_reply(&text)
|
||||
}
|
||||
};
|
||||
|
||||
let success = status.is_success() && !reply.is_empty();
|
||||
let error = if !status.is_success() {
|
||||
@@ -5441,7 +5592,10 @@ pub async fn test_model_verbose(
|
||||
"status": status_code,
|
||||
"reqUrl": req_url,
|
||||
"reqBody": req_body_json,
|
||||
"respHeaders": resp_headers,
|
||||
"respBody": text,
|
||||
"respRawHex": hex_preview,
|
||||
"respByteCount": byte_count,
|
||||
"reply": reply,
|
||||
"error": error,
|
||||
"elapsedMs": elapsed_ms,
|
||||
|
||||
Reference in New Issue
Block a user