🤖OpenCode AI Coding Agent Hits 161k GitHub Stars
OpenCode's caching issues slow down responses to a crawl
TL;DR
OpenCode, an AI coding assistant with 161k stars on GitHub, faces significant performance hits due to cache misses and poor compaction. Prompt generation can take up to 10 minutes in extreme cases.
OpenCode, the popular AI coding agent with 161k GitHub stars, is grappling with serious performance issues stemming from its caching mechanism. When OpenCode can't find a suitable cached prefix for prompt evaluation, it resorts to lengthy GPU computations, leading to response times of up to ten minutes. This bottleneck affects developers using OpenCode in their daily workflows, particularly those relying on real-time feedback and efficiency. The agent re-reads AGENTS.md on every server-sent event (SSE) turn and prunes context from tool calls during user transitions, further complicating the issue.
Key Points
OpenCode has 161k stars on GitHub as of today
Prompt generation can take up to 10 minutes in extreme cases due to cache misses
Pruning discards tool call results more than 40k tokens behind the write head
Compaction is poorly implemented and interacts badly with pruning
The best method for summarizing a session is by manually writing out notes
Why It Matters
If you're using OpenCode as an AI coding assistant, be prepared for long wait times when cache misses occur. The agent's reliance on real-time context evaluation means that large projects or complex prompts can grind to a halt. Developers should consider alternative tools if they need more consistent performance in their daily workflows.
Frequently Asked Questions
Why does this matter?
If you're using OpenCode as an AI coding assistant, be prepared for long wait times when cache misses occur. The agent's reliance on real-time context evaluation means that large projects or complex prompts can grind to a halt. Developers should consider alternative tools if they need more consistent performance in their daily workflows.
What happened?
OpenCode, an AI coding assistant with 161k stars on GitHub, faces significant performance hits due to cache misses and poor compaction. Prompt generation can take up to 10 minutes in extreme cases.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 2,154 builders reading daily.