上下文压缩

s01 > s02 > s03 > s04 > s05 > [ s06 ] | s07 > s08 > s09 > s10 > s11 > s12

"上下文总会满, 要有办法腾地方" -- 三层压缩策略, 换来无限会话。

Harness 层: 压缩 -- 干净的记忆, 无限的会话。

问题

上下文窗口是有限的。读一个 1000 行的文件就吃掉 ~4000 token; 读 30 个文件、跑 20 条命令, 轻松突破 100k token。不压缩, 智能体根本没法在大项目里干活。

解决方案

三层压缩, 激进程度递增:

Every turn:
+------------------+
| Tool call result |
+------------------+
        |
        v
[Layer 1: micro_compact]        (silent, every turn)
  Replace tool_result > 3 turns old
  with "[Previous: used {tool_name}]"
        |
        v
[Check: tokens > 50000?]
   |               |
   no              yes
   |               |
   v               v
continue    [Layer 2: auto_compact]
              Save transcript to .transcripts/
              LLM summarizes conversation.
              Replace all messages with [summary].
                    |
                    v
            [Layer 3: compact tool]
              Model calls compact explicitly.
              Same summarization as auto_compact.

工作原理

1. 第一层 -- micro_compact: 每次 LLM 调用前, 将旧的 tool result 替换为占位符。

function microCompact(messages: Message[]): Message[] {
  const toolResults: ToolResultBlock[] = [];

  for (const message of messages) {
    if (message.role !== "user" || !Array.isArray(message.content)) continue;
    for (const part of message.content) {
      if (isToolResultBlock(part)) {
        toolResults.push(part);
      }
    }
  }

  if (toolResults.length <= KEEP_RECENT) {
    return messages;
  }

  for (const result of toolResults.slice(0, -KEEP_RECENT)) {
    result.content = `[Previous: used ${toolName}]`;
  }

  return messages;
}

1. 第二层 -- auto_compact: token 超过阈值时, 保存完整对话到磁盘, 让 LLM 做摘要。

async function autoCompact(messages: Message[]): Promise<Message[]> {
  const transcriptPath = resolve(TRANSCRIPT_DIR, `transcript_${Date.now()}.jsonl`);
  for (const message of messages) {
    appendFileSync(transcriptPath, `${JSON.stringify(message)}\n`, "utf8");
  }

  const response = await client.messages.create({
    model: MODEL,
    messages: [{
      role: "user",
      content: "Summarize this conversation for continuity...\n\n" +
        JSON.stringify(messages).slice(0, 80_000),
    }],
    max_tokens: 2000,
  });

  return [
    { role: "user", content: `[Conversation compressed. Transcript: ${transcriptPath}]` },
    { role: "assistant", content: "Understood. I have the context from the summary. Continuing." },
  ];
}

1. 第三层 -- manual compact: compact 工具按需触发同样的摘要机制。

2. 循环整合三层:

export async function agentLoop(messages: Message[]) {
  while (true) {
    microCompact(messages);

    if (estimateTokens(messages) > THRESHOLD) {
      messages.splice(0, messages.length, ...(await autoCompact(messages)));
    }

    const response = await client.messages.create(...);
    // ... tool execution ...

    if (manualCompact) {
      messages.splice(0, messages.length, ...(await autoCompact(messages)));
    }
  }
}

完整历史通过 transcript 保存在磁盘上。信息没有真正丢失, 只是移出了活跃上下文。

相对 s05 的变更

组件	之前 (s05)	之后 (s06)
Tools	5	5 (基础 + compact)
上下文管理	无	三层压缩
Micro-compact	无	旧结果 -> 占位符
Auto-compact	无	token 阈值触发
Transcripts	无	保存到 .transcripts/

试一试

cd learn-claude-code

试试这些 prompt:

cd agents-ts
npm install
npm run s06

1. Read every TypeScript file in the agents-ts directory one by one (观察 micro-compact 替换旧结果)
2. Keep reading files until compression triggers automatically
3. Use the compact tool to manually compress the conversation