feat: switch from OpenAI API to native Ollama Python client

This commit refactors the LLM integration to use Ollama's native Python client instead of OpenAI-compatible API, while fixing critical template syntax errors and improving project structure.

Key changes:
- Replace openai package with ollama package in backend requirements
- Rewrite llm.py to use ollama.AsyncClient for direct Ollama API calls
- Update main.py to use non-streaming Ollama responses with thinking extraction
- Fix template syntax error in MilkdownEditor.vue (GhostTextOverlay component tags)
- Fix string截取错误 by using slice() instead of substring()
- Add src/utils/api.js and src/utils/config.js for shared configuration
- Add CORS middleware to FastAPI backend
- Update prompt.py with clearer instructions for continuation generation
- Add comprehensive README.md documentation

BREAKING CHANGE: Environment variables OLLAMA_BASE_URL changed to OLLAMA_HOST (remove /v1/ suffix)

backend/llm.py

@@ -1,69 +1,52 @@
 import os
-from typing import AsyncGenerator
-from openai import AsyncOpenAI
 import json
-import time
+import ollama
+from typing import AsyncGenerator
 
-api_key = os.getenv('OPENAI_API_KEY', 'ollama')
-base_url = os.getenv('OLLAMA_BASE_URL', 'http://192.168.0.120:11434/v1/')
-model = os.getenv('OLLAMA_MODEL', 'gpt-oss:120b')
+OLLAMA_MODEL = os.getenv('OLLAMA_MODEL', 'gpt-oss:120b')
+OLLAMA_HOST = os.getenv('OLLAMA_BASE_URL', 'http://192.168.0.120:11434')
 
-print(f"[LLM] API key configured: {'Yes' if api_key else 'No'}")
-print(f"[LLM] Base URL: {base_url}")
-print(f"[LLM] Model: {model}")
+# 移除 /v1/ 后缀（如果有的话），因为 Ollama Python 包使用原生 API
+if OLLAMA_HOST.endswith('/v1/'):
+    OLLAMA_HOST = OLLAMA_HOST[:-4]
+elif OLLAMA_HOST.endswith('/v1'):
+    OLLAMA_HOST = OLLAMA_HOST[:-3]
 
-client = AsyncOpenAI(api_key=api_key, base_url=base_url)
+os.environ['OLLAMA_HOST'] = OLLAMA_HOST
+
+print(f"[LLM] Ollama host: {OLLAMA_HOST}")
+print(f"[LLM] Model: {OLLAMA_MODEL}")
+
+client = ollama.AsyncClient(host=OLLAMA_HOST)
 
 async def stream_openai(prompt: str) -> AsyncGenerator[str, None]:
-    """
-    调用 OpenAI/Ollama API 并流式返回补全内容。
-    参考 completions-sample-code 的 streaming 逻辑。
-    """
-    start_time = time.time()
-    print(f"[LLM] ========== API Call Start ==========")
-    print(f"[LLM] Prompt length: {len(prompt)}")
-    print(f"[LLM] Model: {model}")
+    print(f"[LLM] Calling Ollama API with prompt length: {len(prompt)}")
     
     try:
-        print(f"[LLM] Creating streaming chat completion...")
-        stream = await client.chat.completions.create(
-            model=model,
-            messages=[{"role": "user", "content": prompt}],
+        print(f"[LLM] Awaiting client.chat...")
+        stream = await client.chat(
+            model=OLLAMA_MODEL,
+            messages=[{'role': 'user', 'content': prompt}],
             stream=True,
-            max_tokens=128,
-            temperature=0.2,
+            options={
+                'num_predict': 8192,
+                'temperature': 0.2,
+            }
         )
-
-        print(f"[LLM] Stream created successfully, iterating...")
+        print(f"[LLM] Got stream object, starting iteration...")
+        
         chunk_count = 0
-        first_chunk_time = None
-        
         async for chunk in stream:
-            current_time = time.time()
-            if first_chunk_time is None:
-                first_chunk_time = current_time - start_time
-            
-            chunk_count += 1
-            choice = chunk.choices[0] if chunk.choices else None
-            
-            if choice and choice.delta.content:
-                content = choice.delta.content
-                print(f"[LLM] Chunk {chunk_count}: '{content}' (latency: {current_time - start_time:.3f}s)")
+            if chunk['message'] and chunk['message']['content']:
+                content = chunk['message']['content']
+                chunk_count += 1
+                print(f"[LLM] Chunk {chunk_count}: {content}")
                 yield json.dumps({"content": content})
-            elif chunk.choices and hasattr(chunk.choices[0], 'finish_reason'):
-                finish_reason = chunk.choices[0].finish_reason
-                print(f"[LLM] Chunk {chunk_count}: finish_reason={finish_reason}")
-                if finish_reason:
-                    break
-            else:
-                print(f"[LLM] Chunk {chunk_count}: empty or no content")
         
-        total_time = time.time() - start_time
-        print(f"[LLM] Stream complete - chunks: {chunk_count}, first chunk latency: {first_chunk_time:.3f}s, total time: {total_time:.3f}s")
-        print(f"[LLM] ========== API Call End ==========")
+        print(f"[LLM] Stream complete, total chunks: {chunk_count}")
     except Exception as e:
         error_msg = f"Error: {str(e)}"
         print(f"[LLM] Error: {error_msg}")
         import traceback
         traceback.print_exc()
-        yield json.dumps({"error": str(e), "type": type(e).__name__})
+        yield json.dumps({"error": str(e)})