cc-viewer 1.2.7 → 1.2.9

package/interceptor.js

@@ -4,6 +4,7 @@ import { appendFileSync, mkdirSync, readdirSync, readFileSync, writeFileSync, st
 import { homedir } from 'node:os';
 import { fileURLToPath } from 'node:url';
 import { dirname, join, basename } from 'node:path';
+import { LOG_DIR } from './findcc.js';
 
 
 const __filename = fileURLToPath(import.meta.url);
@@ -30,7 +31,7 @@ function generateNewLogFilePath() {
   let cwd;
   try { cwd = process.cwd(); } catch { cwd = homedir(); }
   const projectName = basename(cwd).replace(/[^a-zA-Z0-9_\-\.]/g, '_');
-  const dir = join(homedir(), '.claude', 'cc-viewer', projectName);
+  const dir = join(LOG_DIR, projectName);
   try { mkdirSync(dir, { recursive: true }); } catch { }
   return { filePath: join(dir, `${projectName}_${ts}.jsonl`), dir, projectName };
 }
@@ -241,8 +242,6 @@ function isAnthropicApiPath(urlStr) {
   }
 }
 
-// 不再需要折叠函数，保存完整 JSON 供前端渲染
-
 // 组装流式消息为完整的 message 对象
 function assembleStreamMessage(events) {
   let message = null;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cc-viewer",
-  "version": "1.2.7",
+  "version": "1.2.9",
   "description": "Claude Code Logger visualization management tool",
   "license": "MIT",
   "main": "server.js",
@@ -47,6 +47,7 @@
     "proxy.js",
     "interceptor.js",
     "i18n.js",
+    "findcc.js",
     "locales/",
     "concepts/"
   ],

package/proxy.js

@@ -34,13 +34,7 @@ function getOriginalBaseUrl() {
 export function startProxy() {
   return new Promise((resolve, reject) => {
     const server = createServer(async (req, res) => {
-      // [Debug] Log incoming request
-      // console.error(`[CC-Viewer Proxy] Received request: ${req.method} ${req.url}`);
-
-      // Handle CORS preflight if needed (though claude cli probably doesn't send OPTIONS)
-
       const originalBaseUrl = getOriginalBaseUrl();
-      const targetUrl = new URL(req.url, originalBaseUrl);
 
       // Use the patched fetch (which logs to cc-viewer)
       try {
@@ -48,25 +42,12 @@ export function startProxy() {
         const headers = { ...req.headers };
         delete headers.host; // Let fetch set the host
 
-        // [Fix] Handle compressed body
-        // If content-encoding is set (e.g. gzip), and we read the raw stream into a buffer,
-        // we are passing the compressed buffer as body.
-        // fetch will automatically add content-length, but might not handle content-encoding correctly if we just pass the buffer?
-        // Actually, fetch should handle it fine if we pass the headers.
-
-        // However, if we are reading the body here to pass it to fetch, we are buffering it.
-        // For large uploads this might be bad, but for text prompts it's fine.
-
-        // We need to read the body if any
         const buffers = [];
         for await (const chunk of req) {
           buffers.push(chunk);
         }
         const body = Buffer.concat(buffers);
 
-        // [Debug] Log body size
-        // console.error(`[CC-Viewer Proxy] Request body size: ${body.length}`);
-
         const fetchOptions = {
           method: req.method,
           headers: headers,
@@ -80,94 +61,14 @@ export function startProxy() {
           fetchOptions.body = body;
         }
 
-        // [Crucial Fix]
-        // If originalBaseUrl is also a proxy or special endpoint, make sure we construct the full URL correctly.
-        // originalBaseUrl might end with /v1 or not.
-        // req.url from proxy server is usually just the path (e.g. /v1/messages) if client is configured with base_url=http://localhost:port
-        // So new URL(req.url, originalBaseUrl) should work.
-
-        // However, if originalBaseUrl already contains a path (e.g. /api/anthropic), and req.url is /v1/messages,
-        // new URL(req.url, originalBaseUrl) might treat req.url as absolute path if it starts with /, replacing the path in originalBaseUrl?
-        // Let's test: new URL('/v1/messages', 'https://example.com/api').toString() -> 'https://example.com/v1/messages' (path replaced!)
-
-        // This is why we get 404! The user's base URL is https://antchat.alipay.com/api/anthropic
-        // But our proxy constructs: https://antchat.alipay.com/v1/messages
-        // It lost the /api/anthropic part!
-
-        // We need to append req.url to the pathname of originalBaseUrl, carefully avoiding double slashes.
-
-        const originalUrlObj = new URL(originalBaseUrl);
-        // Ensure original pathname doesn't end with slash if we append
-        let basePath = originalUrlObj.pathname;
-        if (basePath.endsWith('/')) basePath = basePath.slice(0, -1);
-
-        // req.url starts with / usually
-        const reqPath = req.url.startsWith('/') ? req.url : '/' + req.url;
-
-        // Check if we should join them
-        // If req.url is /v1/messages, and base is /api/anthropic, we want /api/anthropic/v1/messages
-        originalUrlObj.pathname = basePath + reqPath;
-        // Search params are already in req.url? Yes, req.url includes query string in Node http server.
-        // Wait, new URL(req.url, base) parses query string correctly.
-        // But if we manually concat pathname, we need to handle query string separately.
-
-        // Let's do it simpler: use string concatenation for the full URL
-        // But we need to handle the origin correctly.
-
-        // Better approach:
-        // 1. Remove trailing slash from originalBaseUrl
+        // 拼接完整 URL，保留 originalBaseUrl 中的路径前缀
         const cleanBase = originalBaseUrl.endsWith('/') ? originalBaseUrl.slice(0, -1) : originalBaseUrl;
-        // 2. Remove leading slash from req.url
         const cleanReq = req.url.startsWith('/') ? req.url.slice(1) : req.url;
-        // 3. Join
         const fullUrl = `${cleanBase}/${cleanReq}`;
 
-        // [Debug] Proxying to
-        // console.error(`[CC-Viewer Proxy] Forwarding to: ${fullUrl}`);
-
         const response = await fetch(fullUrl, fetchOptions);
 
-        // [Crucial Fix]
-        // Handle decompression manually if needed.
-        // Node's fetch automatically decompresses if 'compress: true' is default?
-        // Actually, fetch handles gzip/deflate by default.
-        // But if we pipe the response body to the client response (res), we need to be careful.
-        // The issue is likely that we are trying to read the body as text/json for logging, 
-        // but it might be compressed or binary.
-
-        // Let's modify how we handle the response body.
-        // We need to:
-        // 1. Pipe the response to the client (res) so Claude Code gets the data.
-        // 2. Clone the response to read it for logging? fetch response.clone() might not work with streaming body easily.
-
-        // Better approach: intercept the stream.
-        // Or simply: don't log the response body for now to avoid breaking the stream.
-        // User just wants to see the request in the viewer.
-        // If we want to log response, we need to handle it carefully.
-
-        // Let's check where ZlibError comes from. It likely comes from `response.text()` or `response.json()` 
-        // if the content-encoding header is set but fetch didn't decompress it automatically?
-        // Or maybe we are double decompressing?
-
-        // Wait, if we use `response.body.pipe(res)`, that's fine.
-        // But do we read the body elsewhere?
-
-        // Let's look at how we log the response.
-        // We are not logging response body in this proxy.js currently.
-        // Wait, line 105: `response.body.pipe(res);`
-
-        // If the error happens, it might be because `fetch` failed to decompress?
-        // Or maybe `response.body` is already decompressed stream, but we are piping it to `res` which expects raw?
-        // No, `res` (http.ServerResponse) expects raw data.
-
-        // If `fetch` decompresses automatically, then `response.body` yields decompressed chunks.
-        // But `res` writes those chunks to the client.
-        // The client (Claude Code) expects compressed data if it sent `Accept-Encoding: gzip`.
-        // If we send decompressed data but keep `Content-Encoding: gzip` header, client will try to decompress again -> ZlibError!
-
-        // FIX: Remove content-encoding header from response headers before piping to client.
-        // This tells the client "the data I'm sending you is NOT compressed" (because fetch already decompressed it).
-
+        // fetch 自动解压，需移除编码相关 header 避免客户端重复解压
         const responseHeaders = {};
         for (const [key, value] of response.headers.entries()) {
           // Skip Content-Encoding and Transfer-Encoding to let Node/Client handle it
@@ -178,20 +79,11 @@ export function startProxy() {
 
         res.writeHead(response.status, responseHeaders);
 
-        // Also log that we are piping
-        // console.error(`[CC-Viewer Proxy] Response status: ${response.status}`);
-
         if (response.body) {
-          // We need to convert Web Stream (response.body) to Node Stream for piping to res
-          // Node 18+ fetch returns a Web ReadableStream.
-          // We can use Readable.fromWeb(response.body)
           const { Readable } = await import('node:stream');
           // @ts-ignore
           const nodeStream = Readable.fromWeb(response.body);
           nodeStream.pipe(res);
-
-          // Optional: Log response body for debugging (careful with streams)
-          // For now, let's just ensure reliability.
         } else {
           res.end();
         }

package/server.js

@@ -5,9 +5,9 @@ import { dirname, join, extname, basename } from 'node:path';
 import { homedir, userInfo, platform } from 'node:os';
 import { execSync } from 'node:child_process';
 import { LOG_FILE, _initPromise, _resumeState, resolveResumeChoice, _projectName, _cachedApiKey, _cachedAuthHeader, _cachedHaikuModel } from './interceptor.js';
+import { LOG_DIR } from './findcc.js';
 import { t, detectLanguage } from './i18n.js';
 
-const LOG_DIR = join(homedir(), '.claude', 'cc-viewer');
 const PREFS_FILE = join(LOG_DIR, 'preferences.json');