Upload 4 files

README.md

@@ -10,4 +10,4 @@ license: mit
 ---
 
 # ZAI API — chat.z.ai
-OpenAI-compatible API powered by Z.ai (GLM-5.1 / GLM-5-Turbo) via Playwright.
+OpenAI-compatible API: GLM-5.1 / GLM-5-Turbo / GLM-5V-Turbo via Playwright.

main.py

@@ -6,13 +6,12 @@ import threading
 import json
 import re
 from typing import Optional
-from fastapi import FastAPI, Header, HTTPException, Request
+from fastapi import FastAPI, Request
 from fastapi.responses import JSONResponse
 
 
-API_SECRET_KEY = os.getenv("API_SECRET_KEY", "2026-2026")
+API_SECRET_KEY = os.getenv("API_SECRET_KEY", "change-secret-key-2026")
 
-# الموديلات المتاحة في chat.z.ai
 ZAI_MODELS = ["GLM-5.1", "GLM-5-Turbo", "GLM-5V-Turbo"]
 
 
@@ -55,9 +54,9 @@ class AsyncBrowserThread(threading.Thread):
             user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36",
             viewport={'width': 1920, 'height': 1080}
         )
-
-        await context.add_init_script("Object.defineProperty(navigator, 'webdriver', {get: () => undefined})")
-
+        await context.add_init_script(
+            "Object.defineProperty(navigator, 'webdriver', {get: () => undefined})"
+        )
         page = await context.new_page()
 
         try:
@@ -65,30 +64,45 @@ class AsyncBrowserThread(threading.Thread):
             await page.goto("https://chat.z.ai/", wait_until="domcontentloaded")
             await asyncio.sleep(3)
 
-            # Input: textarea#chat-input (confirmed from DOM)
+            # ── Input: confirmed from DOM ─────────────────────────────
             await page.wait_for_selector("textarea#chat-input", timeout=60000)
             await page.fill("textarea#chat-input", prompt)
             await asyncio.sleep(0.5)
             await page.press("textarea#chat-input", "Enter")
+            print(f"[ZAI-SERVER] Sent ({len(prompt)} chars)")
 
-            # Response: #response-content-container (confirmed from DOM)
+            # ── Wait for response container ───────────────────────────
             await asyncio.sleep(2)
             await page.wait_for_selector("#response-content-container", timeout=120000)
 
+            # ── Wait until response stabilizes ────────────────────────
             last_text = ""
             unchanged_count = 0
             while unchanged_count < 5:
-                containers = await page.query_selector_all("#response-content-container")
-                if containers:
-                    current_text = await containers[-1].inner_text()
-                    if current_text == last_text and current_text.strip() != "":
-                        unchanged_count += 1
-                    else:
-                        last_text = current_text
-                        unchanged_count = 0
+                current_text = await page.evaluate("""
+                    () => {
+                        const prose = document.querySelector(
+                            '#response-content-container .markdown-prose'
+                        );
+                        if (!prose) return '';
+                        const clone = prose.cloneNode(true);
+
+                        // ✅ شيل كل direct div children (thinking blocks)
+                        // الرد الحقيقي كله في <p> مش <div>
+                        clone.querySelectorAll(':scope > div').forEach(el => el.remove());
+
+                        return clone.innerText.trim();
+                    }
+                """)
+                if current_text == last_text and current_text.strip():
+                    unchanged_count += 1
+                else:
+                    last_text = current_text
+                    unchanged_count = 0
                 await asyncio.sleep(1.0)
 
-            return _clean_zai_response(last_text)
+            print(f"[ZAI-SERVER] Response: {len(last_text)} chars")
+            return last_text.strip()
 
         except Exception as e:
             print(f"[ZAI-SERVER] Error: {e}")
@@ -100,30 +114,10 @@ class AsyncBrowserThread(threading.Thread):
     def process_request(self, prompt: str):
         if not self.ready_event.wait(timeout=60):
             raise Exception("Error From Browser")
-
         future = asyncio.run_coroutine_threadsafe(self._talk_to_zai(prompt), self.loop)
         return future.result(timeout=120)
 
 
-def _clean_zai_response(text: str) -> str:
-    """
-    Z.ai بيعرض thinking block فوق الرد الحقيقي.
-    نشيل كل حاجة قبل أول سطر فارغ (الـ thinking)
-    ونرجع الرد الفعلي فقط.
-    """
-    lines = text.splitlines()
-    result_lines = []
-    found_blank = False
-    for i, line in enumerate(lines):
-        if not found_blank and line.strip() == "" and i > 0:
-            found_blank = True
-            continue
-        if found_blank:
-            result_lines.append(line)
-    cleaned = "\n".join(result_lines).strip()
-    return cleaned if cleaned else text.strip()
-
-
 browser_engine = AsyncBrowserThread()
 browser_engine.start()
 
@@ -173,7 +167,9 @@ def format_prompt(messages, tools=None):
                 tc_descriptions = []
                 for tc in tool_calls_in_msg:
                     func = tc.get("function", {})
-                    tc_descriptions.append(f"Called '{func.get('name', '?')}' with: {func.get('arguments', '{}')}")
+                    tc_descriptions.append(
+                        f"Called '{func.get('name', '?')}' with: {func.get('arguments', '{}')}"
+                    )
                 assistant_content += "\n[Previous tool calls: " + "; ".join(tc_descriptions) + "]"
             if assistant_content.strip():
                 parts.append(f"[Assistant]: {assistant_content}")
@@ -219,15 +215,12 @@ def format_tools_instruction(tools, user_question=""):
     instruction += "You MUST use one of the tools below to answer this question.\n"
     instruction += "Do NOT answer directly. Do NOT say you don't have information.\n"
     instruction += "You MUST respond with ONLY a JSON object to call the tool.\n\n"
-
     instruction += "RESPONSE FORMAT - respond with ONLY this JSON, nothing else:\n"
     instruction += '{"tool_calls": [{"name": "TOOL_NAME", "arguments": {"param": "value"}}]}\n\n'
-
     instruction += "RULES:\n"
     instruction += "- Your ENTIRE response must be valid JSON only\n"
     instruction += "- No markdown, no code blocks, no explanation\n"
     instruction += "- No text before or after the JSON\n\n"
-
     instruction += "Available tools:\n\n"
 
     for tool in tools:
@@ -235,10 +228,7 @@ def format_tools_instruction(tools, user_question=""):
         name = func.get("name", "unknown")
         desc = func.get("description", "No description")
         params = func.get("parameters", {})
-
-        instruction += f"Tool: {name}\n"
-        instruction += f"Description: {desc}\n"
-
+        instruction += f"Tool: {name}\nDescription: {desc}\n"
         if params.get("properties"):
             instruction += "Parameters:\n"
             required_params = params.get("required", [])
@@ -250,14 +240,11 @@ def format_tools_instruction(tools, user_question=""):
         instruction += "\n"
 
     instruction += "=== END OF TOOLS ===\n\n"
-
     first_tool = tools[0] if tools else {}
     first_func = first_tool.get("function", first_tool)
     first_name = first_func.get("name", "tool")
-
     instruction += f'EXAMPLE: If the user asks a question, respond with:\n'
     instruction += '{"tool_calls": [{"name": "' + first_name + '", "arguments": {"input": "the user question here"}}]}\n\n'
-
     instruction += "Now respond with the JSON to call the appropriate tool:\n\n"
     return instruction
 
@@ -308,6 +295,11 @@ def parse_tool_calls(response_text):
 app = FastAPI(title="zai_api for n8n")
 
 
+def _auth(request: Request) -> bool:
+    auth = request.headers.get("authorization", "")
+    return auth.replace("Bearer ", "").strip() == API_SECRET_KEY
+
+
 @app.post("/v1/chat/completions")
 async def chat_completions(request: Request):
     try:
@@ -315,8 +307,7 @@ async def chat_completions(request: Request):
     except Exception:
         return JSONResponse(status_code=400, content={"error": {"message": "Invalid JSON payload"}})
 
-    authorization = request.headers.get("authorization", "")
-    if not authorization or authorization.replace("Bearer ", "").strip() != API_SECRET_KEY:
+    if not _auth(request):
         return JSONResponse(status_code=401, content={"error": {"message": "Invalid API Key"}})
 
     messages = data.get("messages", [])
@@ -331,9 +322,7 @@ async def chat_completions(request: Request):
         response_text = browser_engine.process_request(prompt)
         p_tokens = len(prompt.split())
         c_tokens = len(response_text.split())
-        tool_calls = None
-        if tools:
-            tool_calls = parse_tool_calls(response_text)
+        tool_calls = parse_tool_calls(response_text) if tools else None
 
         if tool_calls:
             return {
@@ -346,17 +335,16 @@ async def chat_completions(request: Request):
                 "usage": {"prompt_tokens": p_tokens, "completion_tokens": c_tokens,
                     "total_tokens": p_tokens + c_tokens}
             }
-        else:
-            return {
-                "id": f"chatcmpl-{uuid.uuid4().hex[:29]}",
-                "object": "chat.completion",
-                "created": int(start_time),
-                "model": data.get("model", "GLM-5.1"),
-                "choices": [{"index": 0, "message": {"role": "assistant", "content": response_text},
-                    "finish_reason": "stop"}],
-                "usage": {"prompt_tokens": p_tokens, "completion_tokens": c_tokens,
-                    "total_tokens": p_tokens + c_tokens}
-            }
+        return {
+            "id": f"chatcmpl-{uuid.uuid4().hex[:29]}",
+            "object": "chat.completion",
+            "created": int(start_time),
+            "model": data.get("model", "GLM-5.1"),
+            "choices": [{"index": 0, "message": {"role": "assistant", "content": response_text},
+                "finish_reason": "stop"}],
+            "usage": {"prompt_tokens": p_tokens, "completion_tokens": c_tokens,
+                "total_tokens": p_tokens + c_tokens}
+        }
     except Exception as e:
         return JSONResponse(status_code=500, content={"error": str(e)})
 
@@ -368,8 +356,7 @@ async def responses(request: Request):
     except Exception:
         return JSONResponse(status_code=400, content={"error": {"message": "Invalid JSON payload"}})
 
-    authorization = request.headers.get("authorization", "")
-    if not authorization or authorization.replace("Bearer ", "").strip() != API_SECRET_KEY:
+    if not _auth(request):
         return JSONResponse(status_code=401, content={"error": {"message": "Invalid API Key"}})
 
     input_data = data.get("input", "")
@@ -393,9 +380,7 @@ async def responses(request: Request):
         response_text = browser_engine.process_request(prompt)
         p_tokens = len(prompt.split())
         c_tokens = len(response_text.split())
-        tool_calls = None
-        if tools:
-            tool_calls = parse_tool_calls(response_text)
+        tool_calls = parse_tool_calls(response_text) if tools else None
 
         if tool_calls:
             output_items = []
@@ -418,18 +403,17 @@ async def responses(request: Request):
                 "usage": {"input_tokens": p_tokens, "output_tokens": c_tokens,
                     "total_tokens": p_tokens + c_tokens}
             }
-        else:
-            return {
-                "id": f"resp-{uuid.uuid4().hex[:29]}",
-                "object": "response",
-                "created_at": int(start_time),
-                "model": data.get("model", "GLM-5.1"),
-                "status": "completed",
-                "output": [{"type": "message", "role": "assistant",
-                    "content": [{"type": "output_text", "text": response_text}]}],
-                "usage": {"input_tokens": p_tokens, "output_tokens": c_tokens,
-                    "total_tokens": p_tokens + c_tokens}
-            }
+        return {
+            "id": f"resp-{uuid.uuid4().hex[:29]}",
+            "object": "response",
+            "created_at": int(start_time),
+            "model": data.get("model", "GLM-5.1"),
+            "status": "completed",
+            "output": [{"type": "message", "role": "assistant",
+                "content": [{"type": "output_text", "text": response_text}]}],
+            "usage": {"input_tokens": p_tokens, "output_tokens": c_tokens,
+                "total_tokens": p_tokens + c_tokens}
+        }
     except Exception as e:
         return JSONResponse(status_code=500, content={"error": str(e)})