mirror of
https://github.com/CherryHQ/cherry-studio.git
synced 2025-12-25 19:30:17 +08:00
84e78560f4
* refactor: streamline system prompt handling and introduce built-in tools - Removed the static SYSTEM_PROMPT_THRESHOLD from BaseApiClient and replaced it with a constant in constant.ts. - Updated API clients (AnthropicAPIClient, GeminiAPIClient, OpenAIApiClient, OpenAIResponseAPIClient) to simplify system prompt logic by eliminating unnecessary checks and using new utility functions for prompt building. - Introduced built-in tools functionality, including a new 'think' tool, to enhance the tool usage experience. - Refactored ApiService to integrate built-in tools and adjust system prompt modifications accordingly. - Added utility functions for managing built-in tools in mcp-tools.ts and created a new tools index for better organization. * refactor(tests): update prompt tests to use new buildSystemPromptWithTools function - Renamed the function used in prompt tests from buildSystemPrompt to buildSystemPromptWithTools to reflect recent changes in prompt handling. - Adjusted test cases to ensure compatibility with the updated function, maintaining the integrity of user prompt handling. * refactor(ApiService, mcp-tools, prompt): enhance tool usage and prompt handling - Updated ApiService to improve system prompt construction based on tool usage mode, ensuring clearer logic for tool integration. - Enhanced mcp-tools with a new response structure for the 'think' tool, allowing for better handling of tool responses. - Expanded prompt utility functions to include detailed instructions for using the 'think' tool, improving user guidance. - Refactored tests to validate new prompt building logic and tool integration, ensuring robust functionality across scenarios. * fix: enhance prompt * feat(McpToolChunkMiddleware): enhance tool call handling with built-in tool support - Added support for built-in tools in the parseAndCallTools function, allowing for conditional tool invocation based on tool type. - Implemented a check to return early if the tool call response is null, improving error handling.
42 lines
817 B
HTML
42 lines
817 B
HTML
<!doctype html>
|
|
<html lang="zh-CN">
|
|
<head>
|
|
<meta charset="UTF-8" />
|
|
<style>
|
|
html,
|
|
body {
|
|
height: 100%;
|
|
margin: 0;
|
|
padding: 0;
|
|
}
|
|
body {
|
|
min-height: 100vh;
|
|
display: flex;
|
|
flex-direction: column;
|
|
}
|
|
#root {
|
|
flex: 1 1 auto;
|
|
background: var(--trace-bg);
|
|
}
|
|
footer {
|
|
height: 36px;
|
|
text-align: center;
|
|
font-size: 12px;
|
|
color: grey;
|
|
background: var(--trace-header-bg);
|
|
flex-shrink: 0;
|
|
line-height: 36px;
|
|
position: relative;
|
|
z-index: 1;
|
|
}
|
|
footer p {
|
|
margin: 0;
|
|
}
|
|
</style>
|
|
</head>
|
|
<body>
|
|
<div id="root"></div>
|
|
<script type="module" src="/src/trace/traceWindow.tsx"></script>
|
|
</body>
|
|
</html>
|