Unverified90% confidenceFactTime unknown
启用CodeGraph后提示词缓存命中量高达1.3M token，几乎是未启用时的两倍
1
Sources
90%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
Valid until: 8/29/2026
Sources
CodeGraph实测：代码图索引工具真能省Token吗？结果出乎意料
bilibiliRepinSky
Related Entities
Related Claims
Unverified每次重试都会重新发送完整的 prompt 输入 token 并被全额计费，一个 4000 输入 token 的请求重试 3 次会消耗 16000 个 token67% similarUnverifiedCodeGraph claims to save 47% of tokens compared to ad-hoc code searching by coding agents66% similarUnverifiedPrompt Caching在提示词前缀完全相同且超过约1024 token阈值时命中，缓存token通常按原价的25%甚至更低计费65% similarUnverified在Pi里说一句你好只占用约1100的上传Token，仅消耗上下文窗口的千分之四；而在Codex里打招呼要消耗18000 Token，约占7%的上下文窗口63% similarUnverifiedCodeGraph reduced Token consumption by approximately 47% compared to traditional AI coding assistant approaches62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/6390API
curl https://kongchang.com/api/v1/knowledge/claims/6390MCP
get_claim(id=6390)