Skip to content

Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors

fix: also apply adjacency guard inside compressContext

oyi7762652b1fc159
diff --git a/open-sse/services/contextManager.ts b/open-sse/services/contextManager.ts
index d4d48292ea..2834eaa26a 100644
--- a/open-sse/services/contextManager.ts
+++ b/open-sse/services/contextManager.ts
@@ -266,6 +266,8 @@ function purifyHistory(messages: Record<string, unknown>[], targetTokens: number
while (keep > 2) {
let candidate = [...system, ...nonSystem.slice(-keep)];
candidate = fixToolPairs(candidate);
+ candidate = fixToolAdjacency(candidate);
+ candidate = stripTrailingAssistantOrphanToolUse(candidate);
const tokens = estimateTokens(JSON.stringify(candidate));
if (tokens <= targetTokens) break;
keep = Math.max(2, Math.floor(keep * 0.7)); // Drop 30% each iteration
@@ -273,6 +275,8 @@ function purifyHistory(messages: Record<string, unknown>[], targetTokens: number
let result = [...system, ...nonSystem.slice(-keep)];
result = fixToolPairs(result);
+ result = fixToolAdjacency(result);
+ result = stripTrailingAssistantOrphanToolUse(result);
// Add summary of dropped messages
if (keep < nonSystem.length) {