移除 runtime 的 max_loops 限制并同步 compact 语义

[wynxing]

背景

当前测试分支需要去掉 runtime 的 max_loops 限制，避免 ReAct 主循环在固定轮数后被 runtime 强制中断。

这次改动同时清理了围绕 max_loops 和 loop_limit continuation checkpoint 的实现、配置、测试与文档，避免仓库中继续保留已经失效的配置项和语义说明。

变更

• 移除 runtime.Run() 中的 max_loops 终止判断
• 移除 loop_limit compact mode 及相关 continuation checkpoint 逻辑
• 移除 config / YAML 持久化中的 max_loops 配置项
• 更新相关测试：
  • 删除依赖 max_loops/loop_limit 的旧测试
  • 补充“多轮工具调用仍可继续直到最终 assistant 回复”的回归覆盖
  • 保留并对齐 reactive compact 的多次降级重试语义
• 同步更新文档：
  • 配置指南不再暴露 max_loops
  • compact / runtime 流程文档不再描述 loop_limit
  • 明确当前 reactive compact 最多 3 次降级尝试
  • 明确 runtime 不再有内部轮数上限，run 结束依赖最终回复、错误或外部取消

影响

• config.yaml 中如果仍包含 max_loops，现在会按未知字段报错
• runtime 不再提供内部轮数兜底；如果模型持续输出 tool call，session 会一直运行，直到模型收口、报错或被取消
• compact 现仅保留 manual、auto、reactive 三种 mode

验证

• go test ./internal/runtime ./internal/config ./internal/context/compact
• go test ./...

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fennoai]

Focused review completed. I found 2 noteworthy issues (runtime safety bound removal and config migration break).

[fennoai]

Removing the loop budget turns the run loop into an unbounded execution path. If the model keeps returning tool_calls, this can run indefinitely (unbounded tool invocations/token spend/session growth) until external cancellation. Please keep a hard safety guard (for example a per-run max-turn fail-safe, or a separate safety budget) even if max_loops is removed from user-facing config.

[fennoai]

max_loops was removed from persisted config while strict YAML known-fields decoding is still enforced, so existing user configs containing max_loops now hard-fail to load. If this breaking change is intentional, consider adding a temporary compatibility shim (accept+ignore legacy key) or an explicit migration path in docs/release notes to avoid upgrade-time boot failures.