billion-context 0.1.75-pr.490.310 → 0.1.75-pr.493.315

package/dist/index.js

@@ -48479,11 +48479,14 @@ ${summary}` }]
   }
   return { items, replaced, dropped };
 }
-function codexCompactGate(session, effectiveLimit, transformOk) {
-  if (!transformOk || effectiveLimit <= 0) return false;
+function codexCompactGatePre(session, effectiveLimit) {
+  if (effectiveLimit <= 0) return false;
   if (session.stats.lastInputTokens >= effectiveLimit * 0.9) return false;
   return session.state.blocks.some((b2) => b2.active);
 }
+function codexCompactGate(session, effectiveLimit, transformOk) {
+  return transformOk && codexCompactGatePre(session, effectiveLimit);
+}
 function buildTriggerForgeBody(summaryText, usage, stream2) {
   const compactionItem = {
     type: "compaction",
@@ -51914,6 +51917,7 @@ var CONTEXT_OVERFLOW_PATTERNS = [
   /maximum context length/i,
   /max context length/i,
   /maximum context size/i,
+  /longer than the model'?s context length/i,
   /exceeds the context window/i,
   /out of room in the model/i,
   /exceeded model token limit/i,
@@ -51930,7 +51934,7 @@ function toTokenNumber(s3) {
 function parseOverflowWindow(text) {
   let m2 = text.match(/>\s*(\d[\d,]*)\s*maximum/i);
   if (m2) return toTokenNumber(m2[1]);
-  m2 = text.match(/maximum context length is (\d[\d,]*)/i) ?? text.match(/maximum context length of (\d[\d,]*)/i) ?? text.match(/maximum context size (?:is|of) (\d[\d,]*)/i) ?? text.match(/(?:maximum|max)\s+(?:context\s+)?length\s+(?:is\s+)?(\d[\d,]*)/i) ?? text.match(/limit of (\d[\d,]*)\s*token/i) ?? text.match(/(\d[\d,]*)\s*maximum\b/i);
+  m2 = text.match(/maximum context length is (\d[\d,]*)/i) ?? text.match(/maximum context length of (\d[\d,]*)/i) ?? text.match(/maximum context size (?:is|of) (\d[\d,]*)/i) ?? text.match(/(?:maximum|max)\s+(?:context\s+)?length\s+(?:is\s+)?(\d[\d,]*)/i) ?? text.match(/context length\s*\((\d[\d,]*)\s*token/i) ?? text.match(/limit of (\d[\d,]*)\s*token/i) ?? text.match(/(\d[\d,]*)\s*maximum\b/i);
   if (m2) return toTokenNumber(m2[1]);
   return void 0;
 }
@@ -57074,6 +57078,21 @@ ${bodyBuffer.toString("utf8")}`);
     try {
       await withSessionLock(session, async () => {
         const runPrepare = () => countTokens ? prepareCountTokens(parsed, core, reqConfig, log2, session) : protocol === "anthropic" ? prepareAnthropic(parsed, req, opts, core, reqConfig, reqPrompts, log2, session, pluginMode) : protocol === "openai" ? prepareOpenai(parsed, req, opts, core, reqConfig, reqPrompts, log2, session, pluginMode, nativeWindow) : responsesCompact ? prepareResponsesCompact(bodyBuffer, parsed, session, req, core, reqConfig, log2) : prepareResponses(parsed, req, opts, core, reqConfig, reqPrompts, log2, session, responsesIdentity, pluginMode, upstreamOrigin, nativeWindow);
+        const isCodexCompactTrigger = protocol === "responses" && !responsesCompact && isCodexClient(req.headers) && hasCompactionTrigger(parsed.input);
+        if (isCodexCompactTrigger) {
+          const mode = codexCompactMode();
+          const gatePre = codexCompactGatePre(session, reqConfig.modelContextLimit);
+          if (mode === "intercept" && gatePre) prepared = runPrepare();
+          if (prepared?.codexForge) {
+            await forward(req, res, opts, prepared.body, prepared, core, reqConfig, log2, route, instanceId, affinity);
+            rememberPluginMessages(sessionId, prepared.processedMessages, prepared.originalMessages, prepared.nudge);
+            return;
+          }
+          const why = mode !== "intercept" ? "BILI_CODEX_COMPACT=pass" : !gatePre ? "gate preconditions not met" : "transform/forge failed";
+          log2("info", `[${session.id}] codex compaction_trigger request not intercepted (${why}) \u2014 forwarding verbatim (no preflight, no rebuild, no window clamp)`);
+          await forward(req, res, opts, bodyBuffer, null, core, reqConfig, log2, route, instanceId, affinity);
+          return;
+        }
         prepared = runPrepare();
         if (!countTokens && !responsesCompact) {
           const outcome = await preflightCompressIfNeeded(
@@ -57519,7 +57538,7 @@ function prepareResponsesCompact(body, parsed, session, req, core, config, log2)
     compressInjected: false,
     resetAfterSuccess: true
   };
-  if (codexCompactMode() !== "intercept" || !isCodexClient(req.headers) || !Array.isArray(parsed.input)) {
+  if (codexCompactMode() !== "intercept" || !isCodexClient(req.headers) || !Array.isArray(parsed.input) || !codexCompactGatePre(session, config.modelContextLimit)) {
     return base;
   }
   const prevState = session.state;