[GH-ISSUE #63] BUG: use opencode #43

New issue

Closed

opened 2026-02-27 07:17:40 +03:00 by kerem · 3 comments

kerem commented

2026-02-27 07:17:40 +03:00

Owner

Originally created by @qinjw404 on GitHub (Feb 2, 2026).
Original GitHub issue: https://github.com/jwadow/kiro-gateway/issues/63

Kiro Gateway Version

latest

What happened?

Debug Logs

2026-02-02 13:51:38 | ERROR | kiro.routes_openai:chat_completions:292 - Error from Kiro API: 400 - {"message":"Input is too long.","reason":"CONTENT_LENGTH_EXCEEDS_THRESHOLD"}
2026-02-02 13:51:38 | WARNING | kiro.routes_openai:chat_completions:306 - HTTP 400 - POST /v1/chat/completions - Input is too long. (reason: CONTENT_LENGTH_EXCEEDS_THRESHOLD)

Originally created by @qinjw404 on GitHub (Feb 2, 2026). Original GitHub issue: https://github.com/jwadow/kiro-gateway/issues/63 ### Kiro Gateway Version latest ### What happened? <img width="987" height="243" alt="Image" src="https://github.com/user-attachments/assets/ca737d3b-a071-4c19-bdfd-c41560b94ac1" /> ### Debug Logs 2026-02-02 13:51:38 | ERROR | kiro.routes_openai:chat_completions:292 - Error from Kiro API: 400 - {"message":"Input is too long.","reason":"CONTENT_LENGTH_EXCEEDS_THRESHOLD"} 2026-02-02 13:51:38 | WARNING | kiro.routes_openai:chat_completions:306 - HTTP 400 - POST /v1/chat/completions - Input is too long. (reason: CONTENT_LENGTH_EXCEEDS_THRESHOLD)

kerem

2026-02-27 07:17:40 +03:00

closed this issue
added the
upstream

workaround
labels

kerem commented

2026-02-27 07:17:41 +03:00

Author

Owner

@jwadow commented on GitHub (Feb 2, 2026):

Hi, this isn't a Gateway error; it's because the model context is full of 200K tokens. The only thing I can do is display a more understandable error.

This is most likely a shortcoming of OpenCode, which handles context ineffectively.

I can, of course, trim unnecessary messages by modifying the context input into the model, but that's beyond the scope of Gateway, which should be as transparent as possible without explicitly modifying the content. I'll see what I can do about it.

@jwadow commented on GitHub (Feb 2, 2026): Hi, this isn't a Gateway error; it's because the model context is full of 200K tokens. The only thing I can do is display a more understandable error. This is most likely a shortcoming of OpenCode, which handles context ineffectively. I can, of course, trim unnecessary messages by modifying the context input into the model, but that's beyond the scope of Gateway, which should be as transparent as possible without explicitly modifying the content. I'll see what I can do about it.

kerem commented

2026-02-27 07:17:41 +03:00

Author

Owner

@qinjw404 commented on GitHub (Feb 2, 2026):

Perhaps this is a limitation of kiro? How can we better use it? Currently, my opencode is using a custom provider.

@qinjw404 commented on GitHub (Feb 2, 2026): Perhaps this is a limitation of kiro? How can we better use it? Currently, my opencode is using a custom provider.

kerem commented

2026-02-27 07:17:41 +03:00

Author

Owner

@jwadow commented on GitHub (Feb 2, 2026):

No, this is a limitation of the Claude model. Even if you used their official Anthropic API, you would get the same error.

Unfortunately, I won’t tell you how to use OpenCode, because I don’t use it. But properly working agents (of which OpenCode is one) should automatically compress the context when it comes to an end, for example when 180K is used. Why this was not done in your situation, I cannot say.

@jwadow commented on GitHub (Feb 2, 2026): No, this is a limitation of the Claude model. Even if you used their official Anthropic API, you would get the same error. Unfortunately, I won’t tell you how to use OpenCode, because I don’t use it. But properly working agents (of which OpenCode is one) should automatically compress the context when it comes to an end, for example when 180K is used. Why this was not done in your situation, I cannot say.