claude-code - 💡(How to fix) Fix [Bug] Resumed conversations incorrectly recount context tokens as new usage [1 comments, 2 participants]

[{"error":"Error: spawn /data/data/com.termux/files/usr/lib/node_modules/@anthropic-ai/claude-code/vendor/ripgrep/arm64-android/rg ENOENT\n at ChildProcess._handle.onexit (node:internal/child_process:286:19)\n at onErrorNT (node:internal/child_process:507:16)\n at process.processTicksAndRejections (node:internal/process/task_queues:90:21)","timestamp":"2026-04-10T02:29:40.484Z"}]

TL;DR

The issue can be mitigated by implementing a mechanism to track and exclude previously counted token usage when resuming conversations from local storage.

Guidance

• Investigate the token counting mechanism to identify why it's not distinguishing between new and resumed conversations.
• Consider implementing a unique conversation ID that persists across sessions to accurately track token usage.
• Review the local storage data to determine how the conversation context is being reloaded and if there's a way to flag it as a resumed conversation.
• Analyze the error log to see if the ENOENT error is related to the token counting issue or if it's a separate problem.

Example

No code example is provided as the issue lacks specific details about the implementation.

Notes

The provided error log seems unrelated to the token counting issue, but it's essential to investigate and resolve it to ensure the overall stability of the system.

Recommendation

Apply a workaround by implementing a conversation ID tracking mechanism to exclude previously counted token usage when resuming conversations. This approach can help mitigate the issue until a more permanent fix is available.