fix: Ollama model validation, TUI CSS, WebUI overhaul
1. Ollama: validate model is configured before API call, show helpful error 2. TUI: remove unsupported scroll-behavior CSS property 3. WebUI server: add full tool calling loop (FC + JSON action) 4. WebUI HTML: complete redesign - Modern dark theme with better typography - Markdown rendering via marked.js - Tool call/result visualization with colored cards - Dynamic sidebar (engines + tools from server) - Better input area with gradient send button - Responsive layout
This commit is contained in:
xiaoli-cli-bot committed
1 parent
baa2cbc49a
commit
7a72a5c69e
4 files changed
+415
-965
No files matched your search
+110
-28
@@ -97,36 +97,103 @@ class AIEngineBridge:
|
||||
}
|
||||
|
||||
def chat(self, message: str, callback=None):
|
||||
"""发送消息到 AI 并获取回复"""
|
||||
"""发送消息到 AI 并获取回复（支持工具调用循环）"""
|
||||
if not self.cli:
|
||||
return "❌ 小狸核心未初始化，请检查配置"
|
||||
return {"type": "error", "message": "小狸核心未初始化，请检查配置"}
|
||||
|
||||
results = [] # 收集所有事件
|
||||
|
||||
try:
|
||||
# 使用小狸的核心对话处理
|
||||
import io
|
||||
import contextlib
|
||||
# 添加用户消息到历史
|
||||
self.cli.shared_conversation_history.append({
|
||||
"role": "user", "content": message
|
||||
})
|
||||
|
||||
# 捕获输出
|
||||
captured = io.StringIO()
|
||||
with contextlib.redirect_stdout(captured):
|
||||
# 构建系统提示词
|
||||
liugin_prompts = self.cli.get_liugin_usage_prompts()
|
||||
system_prompt = self.cli._build_system_prompt(liugin_prompts)
|
||||
liugin_prompts = self.cli.get_liugin_usage_prompts()
|
||||
system_prompt = self.cli._build_system_prompt(liugin_prompts)
|
||||
|
||||
# 生成响应
|
||||
if self.cli.current_engine:
|
||||
response = self.cli.current_engine.generate_response(
|
||||
message,
|
||||
system_prompt=system_prompt
|
||||
)
|
||||
else:
|
||||
response = "❌ 没有可用的 AI 引擎"
|
||||
loop_count = 0
|
||||
max_loops = 15
|
||||
current_input = message
|
||||
|
||||
output = captured.getvalue()
|
||||
return response or "(无响应)"
|
||||
while loop_count < max_loops:
|
||||
# 生成 AI 响应
|
||||
if not self.cli.current_engine:
|
||||
return {"type": "error", "message": "没有可用的 AI 引擎"}
|
||||
|
||||
response = self.cli.current_engine.generate_response(
|
||||
current_input, system_prompt=system_prompt
|
||||
)
|
||||
|
||||
if not response:
|
||||
return {"type": "reply", "message": "(无响应)"}
|
||||
|
||||
# 检查 FC 响应（Ollama Function Calling）
|
||||
fc_result = self.cli._handle_fc_response(response, message)
|
||||
if fc_result:
|
||||
# FC 工具已执行，结果作为下一轮输入
|
||||
results.append({"type": "fc_tool_done", "message": fc_result[:200]})
|
||||
current_input = fc_result
|
||||
loop_count += 1
|
||||
continue
|
||||
|
||||
# 解析混合响应（文本 + JSON 工具调用）
|
||||
try:
|
||||
text_content, json_data = self.cli._parse_mixed_response(response)
|
||||
except Exception:
|
||||
text_content, json_data = response, None
|
||||
|
||||
if json_data and isinstance(json_data, dict):
|
||||
if json_data.get('action') == 'use_tool':
|
||||
tool_name = json_data.get('tool', '')
|
||||
tool_args = json_data.get('args', '')
|
||||
results.append({"type": "tool_call", "name": tool_name, "args": tool_args[:200]})
|
||||
|
||||
# 执行工具
|
||||
try:
|
||||
tool_result = self.cli._execute_tool_by_name(tool_name, tool_args)
|
||||
except Exception as e:
|
||||
tool_result = f"工具执行错误: {e}"
|
||||
|
||||
results.append({"type": "tool_result", "name": tool_name, "result": str(tool_result)[:500]})
|
||||
|
||||
# 把工具结果加入历史并继续循环
|
||||
self.cli.shared_conversation_history.append({
|
||||
"role": "tool", "content": str(tool_result),
|
||||
"name": tool_name
|
||||
})
|
||||
current_input = f"工具 {tool_name} 的执行结果:\n{tool_result}"
|
||||
loop_count += 1
|
||||
continue
|
||||
|
||||
elif json_data.get('continue') or json_data.get('need_continue'):
|
||||
msg = json_data.get('message', text_content or '')
|
||||
if msg:
|
||||
results.append({"type": "partial_reply", "message": msg})
|
||||
self.cli.shared_conversation_history.append({"role": "assistant", "content": msg})
|
||||
current_input = "继续"
|
||||
loop_count += 1
|
||||
continue
|
||||
|
||||
elif json_data.get('message'):
|
||||
final = json_data.get('message')
|
||||
self.cli.shared_conversation_history.append({"role": "assistant", "content": final})
|
||||
results.append({"type": "reply", "message": final})
|
||||
break
|
||||
|
||||
# 纯文本响应
|
||||
display = text_content if text_content else response
|
||||
self.cli.shared_conversation_history.append({"role": "assistant", "content": display})
|
||||
results.append({"type": "reply", "message": display})
|
||||
break
|
||||
|
||||
if not results:
|
||||
results.append({"type": "reply", "message": response or "(无响应)"})
|
||||
|
||||
return {"type": "multi", "events": results}
|
||||
|
||||
except Exception as e:
|
||||
return f"❌ 错误: {e}"
|
||||
return {"type": "error", "message": f"错误: {e}"}
|
||||
|
||||
def execute_tool(self, tool_name: str, tool_args: str):
|
||||
"""直接执行工具"""
|
||||
@@ -195,15 +262,30 @@ async def ws_handler(websocket):
|
||||
|
||||
# 在线程中执行 AI 调用（避免阻塞事件循环）
|
||||
loop = asyncio.get_event_loop()
|
||||
response = await loop.run_in_executor(
|
||||
result = await loop.run_in_executor(
|
||||
None, bridge.chat, user_msg
|
||||
)
|
||||
|
||||
# 发送回复
|
||||
await websocket.send(json.dumps({
|
||||
"type": "reply",
|
||||
"data": {"message": response},
|
||||
}, ensure_ascii=False))
|
||||
# 处理返回结果
|
||||
if isinstance(result, dict):
|
||||
if result.get("type") == "multi":
|
||||
# 多事件：逐个发送
|
||||
for event in result.get("events", []):
|
||||
await websocket.send(json.dumps({
|
||||
"type": event.get("type", "reply"),
|
||||
"data": event,
|
||||
}, ensure_ascii=False))
|
||||
else:
|
||||
await websocket.send(json.dumps({
|
||||
"type": result.get("type", "reply"),
|
||||
"data": result,
|
||||
}, ensure_ascii=False))
|
||||
else:
|
||||
# 兼容旧格式
|
||||
await websocket.send(json.dumps({
|
||||
"type": "reply",
|
||||
"data": {"message": str(result)},
|
||||
}, ensure_ascii=False))
|
||||
|
||||
elif msg_type == "tool":
|
||||
# 直接工具调用
|
||||
|
||||
Reference in new issue
Block a user