bianhaifeng789-hue/context-utils

Context Utils Skill - Context Utils

Context Utils 上下文窗口工具。

为什么需要这个？

场景：

• 1M context support
• Context window calculation
• Usage percentages
• Max output tokens
• [1m] suffix detection

Claude Code 方案：context.ts + 220+ lines OpenClaw 飞书适配：Context utils + Context management

---

Constants

// Model context window size (200k tokens default)
export const MODEL_CONTEXT_WINDOW_DEFAULT = 200_000

// Maximum output tokens for compact operations
export const COMPACT_MAX_OUTPUT_TOKENS = 20_000

// Default max output tokens
const MAX_OUTPUT_TOKENS_DEFAULT = 32_000
const MAX_OUTPUT_TOKENS_UPPER_LIMIT = 64_000

// Capped default for slot-reservation
export const CAPPED_DEFAULT_MAX_TOKENS = 8_000
export const ESCALATED_MAX_TOKENS = 64_000

---

Functions

1. 1M Context Check

export function is1mContextDisabled(): boolean {
  return isEnvTruthy(process.env.CLAUDE_CODE_DISABLE_1M_CONTEXT)
}

export function has1mContext(model: string): boolean {
  if (is1mContextDisabled()) return false
  return /\[1m\]/i.test(model)
}

export function modelSupports1M(model: string): boolean {
  if (is1mContextDisabled()) return false
  const canonical = getCanonicalName(model)
  return canonical.includes('claude-sonnet-4') || canonical.includes('opus-4-6')
}

2. Get Context Window

export function getContextWindowForModel(
  model: string,
  betas?: string[],
): number {
  // Env override (ant-only)
  if (
    process.env.USER_TYPE === 'ant' &&
    process.env.CLAUDE_CODE_MAX_CONTEXT_TOKENS
  ) {
    const override = parseInt(process.env.CLAUDE_CODE_MAX_CONTEXT_TOKENS, 10)
    if (!isNaN(override) && override > 0) return override
  }

  // [1m] suffix
  if (has1mContext(model)) return 1_000_000

  // Model capabilities
  const cap = getModelCapability(model)
  if (cap?.max_input_tokens && cap.max_input_tokens >= 100_000) {
    if (
      cap.max_input_tokens > MODEL_CONTEXT_WINDOW_DEFAULT &&
      is1mContextDisabled()
    ) {
      return MODEL_CONTEXT_WINDOW_DEFAULT
    }
    return cap.max_input_tokens
  }

  // Betas + Coral reef exp
  if (betas?.includes(CONTEXT_1M_BETA_HEADER) && modelSupports1M(model)) {
    return 1_000_000
  }
  if (getSonnet1mExpTreatmentEnabled(model)) return 1_000_000

  // Ant model
  if (process.env.USER_TYPE === 'ant') {
    const antModel = resolveAntModel(model)
    if (antModel?.contextWindow) return antModel.contextWindow
  }

  return MODEL_CONTEXT_WINDOW_DEFAULT
}

3. Calculate Percentages

export function calculateContextPercentages(
  currentUsage: {
    input_tokens: number
    cache_creation_input_tokens: number
    cache_read_input_tokens: number
  } | null,
  contextWindowSize: number,
): { used: number | null; remaining: number | null } {
  if (!currentUsage) return { used: null, remaining: null }

  const totalInputTokens =
    currentUsage.input_tokens +
    currentUsage.cache_creation_input_tokens +
    currentUsage.cache_read_input_tokens

  const usedPercentage = Math.round(
    (totalInputTokens / contextWindowSize) * 100,
  )
  const clampedUsed = Math.min(100, Math.max(0, usedPercentage))

  return {
    used: clampedUsed,
    remaining: 100 - clampedUsed,
  }
}

---

飞书卡片格式

Context Utils 卡片

{
  "config": {"wide_screen_mode": true},
  "elements": [
    {
      "tag": "div",
      "text": {
        "tag": "lark_md",
        "content": "**📐 Context Utils**\n\n---\n\n**Constants**：\n• MODEL_CONTEXT_WINDOW_DEFAULT = 200k\n• COMPACT_MAX_OUTPUT_TOKENS = 20k\n• CAPPED_DEFAULT_MAX_TOKENS = 8k\n• ESCALATED_MAX_TOKENS = 64k\n\n---\n\n**Functions**：\n• is1mContextDisabled()\n• has1mContext(model)\n• modelSupports1M(model)\n• getContextWindowForModel()\n• calculateContextPercentages()\n\n---\n\n**1M Context**：\n• [1m] suffix detection\n• Sonnet-4 + Opus-4-6\n• Coral reef experiment"
      }
    }
  ]
}

---

持久化存储

// memory/context-utils-state.json
{
  "stats": {
    "defaultContextWindow": 200000,
    "1mEnabled": false
  },
  "lastUpdate": "2026-04-12T11:26:00Z",
  "notes": "Context Utils Skill 创建完成。"
}

---

与 Claude Code 的差异

| Claude Code | OpenClaw 飞书场景 | |-------------|------------------| | context.ts (220+ lines) | Skill + Context | | getContextWindowForModel() | Get window | | 1M context | 1M support | | calculateContextPercentages() | Percentages |

---

注意事项

1. 200k default - MODEL_CONTEXT_WINDOW_DEFAULT
2. 1M detection - [1m] suffix + model check
3. Env override - CLAUDE_CODE_MAX_CONTEXT_TOKENS (ant-only)
4. Betas - CONTEXT_1M_BETA_HEADER
5. Coral reef - Sonnet 1m experiment

---

自动启用

此 Skill 在 context calculation 时自动运行。

---

下一步增强

• 飞书 context 集成
• Context analytics
• Context debugging