48 lines
1.7 KiB
Python
48 lines
1.7 KiB
Python
"""上下文预算守卫：历史过长时折叠中间消息，保护 prompt 预算。"""
|
|
|
|
from app.ai_provider.types.messages import BaseMessage
|
|
from app.application.agent.extensions import (
|
|
AgentContext,
|
|
AgentRunContext,
|
|
ExtensionRegistry,
|
|
)
|
|
from app.application.agent.python_skills import PythonSkillDefinition
|
|
|
|
# messages 总字符数软上限；超出则保留首条与末尾 _TAIL_KEEP 条，中间折叠。
|
|
_MAX_CONTEXT_CHARS = 12000
|
|
_TAIL_KEEP = 8
|
|
|
|
|
|
async def _trim_history(context: AgentContext, _run: AgentRunContext) -> AgentContext:
|
|
messages = context.messages
|
|
total = sum(len(message.content) for message in messages)
|
|
if total <= _MAX_CONTEXT_CHARS or len(messages) <= _TAIL_KEEP + 1:
|
|
return context
|
|
first = messages[0]
|
|
tail = messages[-_TAIL_KEEP:]
|
|
dropped = len(messages) - 1 - _TAIL_KEEP
|
|
# 在首条追加折叠提示（新建对象避免污染原 message）；不在历史中间插入新消息，
|
|
# 以免破坏 provider 对 system 位置或多轮 tool 消息配对的预期。
|
|
folded = BaseMessage(
|
|
content=f"{first.content}\n\n[已省略 {dropped} 条较早历史]",
|
|
role=first.role,
|
|
)
|
|
return AgentContext(
|
|
messages=(folded, *tail),
|
|
tools=context.tools,
|
|
model_hint=context.model_hint,
|
|
)
|
|
|
|
|
|
def _setup(registry: ExtensionRegistry) -> None:
|
|
registry.add_context_transformer(_trim_history)
|
|
|
|
|
|
SKILL = PythonSkillDefinition(
|
|
name="context_budget",
|
|
description="对话历史过长时折叠中间消息，控制 prompt 预算。",
|
|
triggers=("长对话", "context", "历史"),
|
|
instructions="历史较长时优先依赖近期消息，不必回溯很早的内容。",
|
|
setup=_setup,
|
|
)
|