pjandro/OmniRoute
1
0
Fork 0
mirror of https://github.com/diegosouzapw/OmniRoute.git synced 2026-07-31 04:12:10 +03:00
Code Issues Packages Projects Releases Wiki Activity

fix(claude): respect client thinking/effort params to prevent forced quota drain (#1761)

Browse Source 
Previously, OmniRoute unconditionally injected thinking: {type: 'adaptive'} and
output_config: {effort: 'high'} for Claude Opus 4.7 in Claude Code client requests.
This caused Claude Max 5h quota to drain in ~15 minutes.

Now checks the original client body: if thinking or output_config are explicitly set
(even to null or a different value), the injection is skipped. Users can opt-out by
sending thinking: null or output_config: {effort: 'low'}.
This commit is contained in:
diegosouzapw
2026-04-29 09:54:52 -03:00
parent d12406f32a
commit 5fc126427c
1 changed files with 11 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
open-sse/executors/base.ts
View File

@@ -527,17 +527,25 @@ export class BaseExecutor {
const supportsAdaptiveThinking = supportsXHighEffort("claude", model);
if (supportsAdaptiveThinking && !tb.thinking) {
// Fix #1761: Only inject adaptive thinking/high effort if the client didn't
// explicitly set these fields. This allows users to opt-out by sending
// `thinking: null` or `output_config: { effort: "low" }` to prevent forced
// quota drain on Claude Max accounts.
const originalBody = body as Record<string, unknown>;
const clientExplicitThinking = originalBody?.thinking !== undefined;
const clientExplicitEffort = originalBody?.output_config !== undefined;
if (supportsAdaptiveThinking && !tb.thinking && !clientExplicitThinking) {
tb.thinking = { type: "adaptive" };
}
if (supportsAdaptiveThinking && !tb.context_management) {
if (supportsAdaptiveThinking && !tb.context_management && !clientExplicitThinking) {
tb.context_management = {
edits: [{ type: "clear_thinking_20251015", keep: "all" }],
};
}
if (supportsAdaptiveThinking && !tb.output_config) {
if (supportsAdaptiveThinking && !tb.output_config && !clientExplicitEffort) {
tb.output_config = { effort: "high" };
}