增加注入提示词功能

2026-03-13 17:22:28 +08:00 · 2026-03-13 17:22:28 +08:00 · ae059ffcb8
commit ae059ffcb8
parent f193c48ce1
10 changed files with 160 additions and 20 deletions
--- a/adapters/responses_cc_adapter.py
+++ b/adapters/responses_cc_adapter.py
@ -543,14 +543,17 @@ class ResponsesStreamConverter:
    def _rewrite_top_level_model(self, payload: JsonDict) -> JsonDict:
        """在保持上游事件结构不变的前提下回填展示模型名。

-        Responses 原生事件经常会在顶层带一个 `model` 字段；这里不重写内部结构，只把这个
-        顶层字段替换为当前映射的 Cursor 模型名，避免界面显示上游原始模型名。
+        原生 Responses 事件中 model 可能在顶层或嵌套在 response 子对象中，
+        两处都需要改写，避免界面显示上游原始模型名。
        """
        if not isinstance(payload, dict):
            return payload
        copied = dict(payload)
        if copied.get('model'):
            copied['model'] = self.model or copied['model']
+        if isinstance(copied.get('response'), dict) and copied['response'].get('model'):
+            copied['response'] = dict(copied['response'])
+            copied['response']['model'] = self.model or copied['response']['model']
        return copied


@ -597,12 +600,13 @@ class ResponsesToCCStreamConverter:
    def _handle_output_item_added(self, event_data: JsonDict) -> list[JsonDict]:
        """处理 Responses 的 output_item.added 事件。

-        这里主要关心 function_call，因为文本和 reasoning 的真正内容会分别在后续 delta
-        事件里补全；function_call 则需要先创建一个空参数的 tool_call 槽位。
+        上游事件结构为 {type, item: {type, call_id, name, ...}, output_index}，
+        实际的 function_call 信息在 item 子对象中。
        """
-        if event_data.get('type') != 'function_call':
+        item = event_data.get('item') or event_data
+        if item.get('type') != 'function_call':
            return []
-        call_id = event_data.get('call_id') or gen_id('call_')
+        call_id = item.get('call_id') or gen_id('call_')
        index = self._tool_slots.setdefault(call_id, self._tool_index)
        if index == self._tool_index:
            self._tool_index += 1
@ -612,7 +616,7 @@ class ResponsesToCCStreamConverter:
                'id': call_id,
                'type': 'function',
                'function': {
-                    'name': event_data.get('name', ''),
+                    'name': item.get('name', ''),
                    'arguments': '',
                },
            }]
@ -633,11 +637,16 @@ class ResponsesToCCStreamConverter:
        })]

    def _handle_completed(self, event_data: JsonDict) -> list[JsonDict]:
-        """处理 response.completed，补出聊天补全流的最终收尾 chunk。"""
-        self._usage = event_data.get('usage', {}) or {}
+        """处理 response.completed，补出聊天补全流的最终收尾 chunk。
+
+        上游事件结构为 {type, response: {output, usage, ...}}，
+        实际的 output/usage 在 response 子对象中。
+        """
+        resp = event_data.get('response') or event_data
+        self._usage = resp.get('usage', {}) or {}
        finish_reason = 'tool_calls' if any(
            isinstance(item, dict) and item.get('type') == 'function_call'
-            for item in event_data.get('output', [])
+            for item in resp.get('output', [])
        ) else 'stop'
        chunk = self._make_chunk(delta={}, finish_reason=finish_reason)
        chunk['usage'] = {
@ -718,7 +727,7 @@ def _append_responses_input_item(
    item: JsonDict = {
        'type': 'message',
        'role': role or 'user',
-        'content': _content_to_responses_parts(content),
+        'content': _content_to_responses_parts(content, role),
    }
    input_items.append(item)

@ -1013,13 +1022,19 @@ def _content_to_text(content: Any) -> str:
    return str(content) if content is not None else ''


-def _content_to_responses_parts(content: Any) -> list[JsonDict]:
-    """将普通消息内容转换为 Responses `input_text` 数组。"""
+def _content_to_responses_parts(content: Any, role: str = 'user') -> list[JsonDict]:
+    """将普通消息内容转换为 Responses 内容块数组。
+
+    assistant 消息使用 output_text，其他角色使用 input_text。
+    """
    if isinstance(content, list):
        text = _extract_text(content)
    else:
        text = _content_to_text(content)
-    return [{'type': 'input_text', 'text': text}] if text else []
+    if not text:
+        return []
+    part_type = 'output_text' if role == 'assistant' else 'input_text'
+    return [{'type': part_type, 'text': text}]


 def _stringify_output(content: Any) -> str:
--- a/routes/admin.py
+++ b/routes/admin.py
@ -135,6 +135,8 @@ def add_mapping():
        'backend': data.get('backend', 'auto'),
        'target_url': data.get('target_url', ''),
        'api_key': data.get('api_key', ''),
+        'custom_instructions': data.get('custom_instructions', ''),
+        'instructions_position': data.get('instructions_position', 'prepend'),
    }
    return _save_and_respond(s, f'映射已添加: {name}')

@ -157,6 +159,8 @@ def update_mapping(name):
        'backend': data.get('backend', 'auto'),
        'target_url': data.get('target_url', ''),
        'api_key': data.get('api_key', ''),
+        'custom_instructions': data.get('custom_instructions', ''),
+        'instructions_position': data.get('instructions_position', 'prepend'),
    }
    if new_name != name:
        del mappings[name]
--- a/routes/chat.py
+++ b/routes/chat.py
@ -33,6 +33,9 @@ from routes.common import (
    build_responses_target,
    build_route_context,
    chat_error_chunk,
+    inject_instructions_anthropic,
+    inject_instructions_cc,
+    inject_instructions_responses,
    log_route_context,
    log_usage,
    sse_data_message,
@ -108,6 +111,7 @@ def _handle_openai_backend(ctx: RouteContext, payload: dict[str, Any]):
    )

    payload = normalize_request(payload, ctx.upstream_model)
+    payload = inject_instructions_cc(payload, ctx.custom_instructions, ctx.instructions_position)
    _dbg(
        f'标准化完成：模型={payload.get("model")} '
        f'工具数={len(payload.get("tools", []))}'
@ -194,6 +198,7 @@ def _handle_responses_backend(ctx: RouteContext, payload: dict[str, Any]):
    """
    responses_payload = cc_to_responses_request(payload)
    responses_payload['model'] = ctx.upstream_model
+    responses_payload = inject_instructions_responses(responses_payload, ctx.custom_instructions, ctx.instructions_position)
    _dbg(
        '已转换为 Responses 请求：字段=' + str(list(responses_payload.keys()))
        + f' 输入项数={len(responses_payload.get("input", []))}'
@ -270,6 +275,7 @@ def _handle_anthropic_backend(ctx: RouteContext, payload: dict[str, Any]):
    """处理走 Anthropic Messages 后端的聊天补全请求。"""
    payload['model'] = ctx.upstream_model
    anthropic_payload = cc_to_messages_request(payload)
+    anthropic_payload = inject_instructions_anthropic(anthropic_payload, ctx.custom_instructions, ctx.instructions_position)
    _dbg(
        '已转换为 Messages 请求：字段=' + str(list(anthropic_payload.keys()))
        + f' 消息数={len(anthropic_payload.get("messages", []))}'
--- a/routes/common.py
+++ b/routes/common.py
@ -22,7 +22,8 @@ class RouteContext:
    """数据面路由使用的标准请求上下文。

    路由层会先根据客户端模型名解析出统一上下文，后续处理函数只需要关心
-    上游模型、后端类型、目标地址、鉴权信息和流式标记，而不必重复访问配置层。
+    上游模型、后端类型、目标地址、鉴权信息、流式标记和自定义指令，
+    而不必重复访问配置层。
    """

    client_model: str
@ -31,6 +32,8 @@ class RouteContext:
    target_url: str
    api_key: str
    is_stream: bool
+    custom_instructions: str
+    instructions_position: str


 def build_route_context(client_model: str, is_stream: bool) -> RouteContext:
@ -43,6 +46,8 @@ def build_route_context(client_model: str, is_stream: bool) -> RouteContext:
        target_url=mapping['target_url'],
        api_key=mapping['api_key'],
        is_stream=is_stream,
+        custom_instructions=mapping.get('custom_instructions', ''),
+        instructions_position=mapping.get('instructions_position', 'prepend'),
    )


@ -120,3 +125,71 @@ def chat_error_chunk(message: str, error_type: str = 'upstream_error') -> str:
 def responses_error_event(message: str) -> str:
    """构造 Responses 流式接口使用的错误事件。"""
    return sse_event_message('error', {'error': message})
+
+
+# ─── 自定义指令注入 ──────────────────────────────
+
+
+def _merge_text(custom: str, existing: str, position: str) -> str:
+    """根据 position 决定自定义指令与原有内容的拼接顺序。"""
+    if not existing:
+        return custom
+    if position == 'append':
+        return existing + '\n\n' + custom
+    return custom + '\n\n' + existing
+
+
+def inject_instructions_cc(payload: dict[str, Any], instructions: str, position: str = 'prepend') -> dict[str, Any]:
+    """向 Chat Completions 请求注入自定义指令。
+
+    position='prepend' 时放在 system 消息开头，'append' 时放在末尾。
+    """
+    if not instructions:
+        return payload
+
+    messages = payload.get('messages', [])
+    if messages and messages[0].get('role') == 'system':
+        first = messages[0]
+        original = first.get('content') or ''
+        first['content'] = _merge_text(instructions, original, position)
+    else:
+        messages.insert(0, {'role': 'system', 'content': instructions})
+        payload['messages'] = messages
+
+    logger.info('已注入自定义指令到 CC system 消息 (%d 字符, %s)', len(instructions), position)
+    return payload
+
+
+def inject_instructions_responses(payload: dict[str, Any], instructions: str, position: str = 'prepend') -> dict[str, Any]:
+    """向 Responses 请求注入自定义指令（写入 instructions 字段）。
+
+    position='prepend' 时放在 instructions 开头，'append' 时放在末尾。
+    """
+    if not instructions:
+        return payload
+
+    existing = payload.get('instructions') or ''
+    payload['instructions'] = _merge_text(instructions, existing, position)
+
+    logger.info('已注入自定义指令到 Responses instructions (%d 字符, %s)', len(instructions), position)
+    return payload
+
+
+def inject_instructions_anthropic(payload: dict[str, Any], instructions: str, position: str = 'prepend') -> dict[str, Any]:
+    """向 Anthropic Messages 请求注入自定义指令（写入 system 字段）。
+
+    position='prepend' 时放在 system 开头，'append' 时放在末尾。
+    """
+    if not instructions:
+        return payload
+
+    existing = payload.get('system') or ''
+    if isinstance(existing, list):
+        existing = '\n'.join(
+            block.get('text', '') for block in existing
+            if isinstance(block, dict) and block.get('type') == 'text'
+        )
+    payload['system'] = _merge_text(instructions, existing, position)
+
+    logger.info('已注入自定义指令到 Anthropic system (%d 字符, %s)', len(instructions), position)
+    return payload
--- a/routes/messages.py
+++ b/routes/messages.py
@ -13,6 +13,7 @@ from flask import Blueprint, request, jsonify

 import settings
 from config import Config
+from routes.common import inject_instructions_anthropic
 from utils.http import build_anthropic_headers, forward_request, sse_response

 logger = logging.getLogger(__name__)
@ -29,11 +30,16 @@ def messages_passthrough():

    logger.info(f'[透传] model={model} 流式={is_stream}')

-    url_base = settings.get_url()
-    api_key = settings.get_key()
+    mapping = settings.resolve_model(model)
+    url_base = mapping['target_url']
+    api_key = mapping['api_key']
+    custom_instructions = mapping.get('custom_instructions', '')
+    instructions_position = mapping.get('instructions_position', 'prepend')
    headers = build_anthropic_headers(api_key)
    url = f'{url_base.rstrip("/")}/v1/messages'

+    payload = inject_instructions_anthropic(payload, custom_instructions, instructions_position)
+
    if not is_stream:
        resp, err = forward_request(url, headers, payload)
        if err:
--- a/routes/responses.py
+++ b/routes/responses.py
@ -22,6 +22,9 @@ from routes.common import (
    build_openai_target,
    build_responses_target,
    build_route_context,
+    inject_instructions_anthropic,
+    inject_instructions_cc,
+    inject_instructions_responses,
    log_route_context,
    log_usage,
    responses_error_event,
@ -73,6 +76,7 @@ def _build_cc_payload(payload: dict[str, Any], ctx: RouteContext) -> dict[str, A
    """
    cc_payload = responses_to_cc(payload)
    cc_payload['model'] = ctx.upstream_model
+    cc_payload = inject_instructions_cc(cc_payload, ctx.custom_instructions, ctx.instructions_position)
    _dbg(
        '已转换为聊天补全中间表示：字段=' + str(list(cc_payload.keys()))
        + f' 消息数={len(cc_payload.get("messages", []))}'
@ -171,6 +175,7 @@ def _handle_responses_backend(ctx: RouteContext, payload: dict[str, Any]):
    """
    payload = dict(payload)
    payload['model'] = ctx.upstream_model
+    payload = inject_instructions_responses(payload, ctx.custom_instructions, ctx.instructions_position)
    url, headers = build_responses_target(ctx)

    if ctx.is_stream:
--- a/settings.py
+++ b/settings.py
@ -2,7 +2,7 @@

 使用 data/settings.json 存储可通过管理面板修改的设置：
  - proxy_target_url / proxy_api_key: 可覆盖环境变量的全局配置
-  - model_mappings: Cursor 模型名 → {upstream_model, backend, target_url, api_key}
+  - model_mappings: Cursor 模型名 → {upstream_model, backend, target_url, api_key, custom_instructions}
 """

 import json
@ -90,6 +90,8 @@ def resolve_model(model_name):
            'backend': backend,
            'target_url': m.get('target_url') or base_url,
            'api_key': m.get('api_key') or base_key,
+            'custom_instructions': m.get('custom_instructions') or '',
+            'instructions_position': m.get('instructions_position') or 'prepend',
        }

    return {
@ -97,6 +99,8 @@ def resolve_model(model_name):
        'backend': _auto_detect(model_name),
        'target_url': base_url,
        'api_key': base_key,
+        'custom_instructions': '',
+        'instructions_position': 'prepend',
    }


--- a/static/admin.css
+++ b/static/admin.css
@ -19,8 +19,8 @@ code{background:var(--input);padding:1px 5px;border-radius:4px;font-size:12px;fo
 .field{margin-bottom:16px}
 .field label{display:block;font-size:13px;color:var(--muted);margin-bottom:6px;font-weight:500}
 .input-wrap{position:relative}
-.input-wrap input,.input-wrap select{width:100%;background:var(--input);border:1px solid var(--border);border-radius:8px;padding:10px 14px;color:var(--text);font-size:14px;outline:none;transition:border-color .2s}
-.input-wrap input:focus,.input-wrap select:focus{border-color:var(--primary)}
+.input-wrap input,.input-wrap select,.input-wrap textarea{width:100%;background:var(--input);border:1px solid var(--border);border-radius:8px;padding:10px 14px;color:var(--text);font-size:14px;outline:none;transition:border-color .2s}
+.input-wrap input:focus,.input-wrap select:focus,.input-wrap textarea:focus{border-color:var(--primary)}
 .input-wrap .eye{position:absolute;right:10px;top:50%;transform:translateY(-50%);background:none;border:none;color:var(--muted);cursor:pointer;padding:4px;font-size:16px}
 select{cursor:pointer;appearance:none;background-image:url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='12' height='12' fill='%238899ab'%3E%3Cpath d='M6 8L1 3h10z'/%3E%3C/svg%3E");background-repeat:no-repeat;background-position:right 12px center}

@ -61,6 +61,7 @@ main{padding:28px 0 60px}
 .tag-responses{background:rgba(59,130,246,.15);color:#60a5fa}
 .tag-auto{background:rgba(139,92,246,.15);color:#a78bfa}
 .tag-override{background:rgba(59,130,246,.1);color:var(--primary)}
+.tag-instructions{background:rgba(234,179,8,.15);color:var(--yellow)}
 .mapping-actions{margin-left:auto;display:flex;gap:6px}
 .empty{text-align:center;padding:40px;color:var(--muted)}

--- a/static/admin.html
+++ b/static/admin.html
@ -115,6 +115,24 @@
        <button class="eye" onclick="togglePwd('mKey')">&#128065;</button>
      </div>
    </div>
+    <div class="field">
+      <label>自定义指令 <span style="color:var(--muted)">（可选）</span></label>
+      <div class="input-wrap"><textarea id="mInstructions" rows="5" placeholder="注入到请求中的自定义系统指令&#10;&#10;Responses 格式 → 写入 instructions 字段（developer 级别）&#10;Chat Completions 格式 → 注入为首条 system 消息&#10;Anthropic 格式 → 写入 system 字段" style="resize:vertical;min-height:80px"></textarea></div>
+      <div class="hint">
+        每个模型可单独配置。指令会被注入到发往上游的请求中，模型会无条件信任该内容。<br>
+        适用场景：自定义人格、解除限制、注入 CTF 环境假设等。
+      </div>
+    </div>
+    <div class="field">
+      <label>注入位置</label>
+      <div class="input-wrap">
+        <select id="mInsPosition">
+          <option value="prepend">前置（推荐，优先级更高）</option>
+          <option value="append">后置</option>
+        </select>
+      </div>
+      <div class="hint">前置：自定义指令放在系统提示词最前面，模型优先看到；后置：放在末尾</div>
+    </div>
    <div class="modal-footer">
      <button class="btn btn-ghost" onclick="closeModal()">取消</button>
      <button class="btn btn-primary" id="modalSaveBtn" onclick="saveMapping()">保存</button>
--- a/static/admin.js
+++ b/static/admin.js
@ -137,6 +137,7 @@ async function loadMappings() {
        ? 'responses'
        : backend;
    const hasOverride = m.target_url || m.api_key;
+    const hasInstructions = !!m.custom_instructions;
    return `<div class="mapping-item">
      <div class="mapping-top">
        <span class="mapping-name">${esc(name)}</span>
@ -145,6 +146,7 @@ async function loadMappings() {
        <div class="mapping-meta">
          <span class="tag ${tagClass}">${tagLabel}</span>
          ${hasOverride ? '<span class="tag tag-override">自定义地址</span>' : ''}
+          ${hasInstructions ? '<span class="tag tag-instructions">自定义指令</span>' : ''}
        </div>
        <div class="mapping-actions">
          <button class="btn btn-ghost btn-sm" onclick="openEditModal('${esc(name)}')">编辑</button>
@ -167,6 +169,8 @@ function openAddModal() {
  document.getElementById('mBackend').value = 'auto';
  document.getElementById('mUrl').value = '';
  document.getElementById('mKey').value = '';
+  document.getElementById('mInstructions').value = '';
+  document.getElementById('mInsPosition').value = 'prepend';
  document.getElementById('modal').classList.add('active');
 }

@ -183,6 +187,8 @@ async function openEditModal(name) {
    document.getElementById('mBackend').value = m.backend || 'auto';
    document.getElementById('mUrl').value = m.target_url || '';
    document.getElementById('mKey').value = m.api_key || '';
+    document.getElementById('mInstructions').value = m.custom_instructions || '';
+    document.getElementById('mInsPosition').value = m.instructions_position || 'prepend';
    document.getElementById('modal').classList.add('active');
  } catch (e) {
    toast('错误: ' + e.message, false);
@ -206,6 +212,8 @@ async function saveMapping() {
    backend: document.getElementById('mBackend').value,
    target_url: document.getElementById('mUrl').value.trim(),
    api_key: document.getElementById('mKey').value.trim(),
+    custom_instructions: document.getElementById('mInstructions').value,
+    instructions_position: document.getElementById('mInsPosition').value,
  };

  try {