billion-context 0.1.95 → 0.1.96-pr.647.472

package/dist/index.js

@@ -49133,7 +49133,9 @@ function isCodexClient(headers) {
   const ua = headers["user-agent"];
   if (!ua) return false;
   const s3 = Array.isArray(ua) ? ua[0] : ua;
-  return typeof s3 === "string" && CODEX_UA_PREFIXES.some((p2) => s3.startsWith(p2));
+  if (typeof s3 !== "string") return false;
+  if (CODEX_UA_PREFIXES.some((p2) => s3.startsWith(p2))) return true;
+  return s3.includes("codex");
 }
 function hasCompactionTrigger(input) {
   if (!Array.isArray(input)) return false;
@@ -51390,9 +51392,9 @@ ${body}`;
 var PAIRED = /\x3cacp\s[^<>]{0,256}>([^<>]{0,64})<\/acp>/;
 var LONE_OPEN = /\x3cacp(?:\s[^<>]{0,256})?>/;
 var LONE_CLOSE = /<\/acp(?=[\s>])[^<>]{0,32}>/;
-var PARTIAL_TAIL = /(\x3cacp\s[^<>]*|\x3c\/acp(?:\s[^<>]*)?|<\/?a?c?p?)$/;
+var PARTIAL_TAIL = /(\x3cacp\s[^<>]*|\x3c\/acp(?:\s[^<>]{0,32})?|<\/?a?c?p?)$/;
 var TRUNC_OPEN = /\x3cacp\s[^<>]*$/;
-var TRUNC_CLOSE = /\x3c\/acp(?:\s[^<>]*)?$/;
+var TRUNC_CLOSE = /\x3c\/acp(?:\s[^<>]{0,32})?$/;
 var CLOSE_TAG = "</acp";
 var HOLD_LIMIT = 128;
 var TAG_OPEN_CAP = 4096;
@@ -59337,9 +59339,10 @@ function diagNudge(turn, sessionId, tokenCount, limit, model, willInject) {
   const modelTag = model ? ` model=${model}` : "";
   return `[${sessionId}] nudge ${inject}: usage=${pct2} (${tokenCount}/${limit}), growth=${growth}/${floor} (ref=${ref}, interval=${interval}), pendingT1=${pendingT1}/${interval}${modelTag}, reason="${n.reason.slice(0, 120)}"`;
 }
-function armHostUsageCredit(session, originalMessages, processedMessages, log2) {
+function armHostUsageCredit(session, originalMessages, processedMessages, headers, log2) {
   session.hostCreditTokens = 0;
   if (session.metadata.pluginAgent === "pi" || session.metadata.pluginAgent === "omp") return;
+  if (isCodexClient(headers)) return;
   session.hostCreditTokens = processedMessages.length > 0 ? Math.max(0, estimateCoreMessages(originalMessages) - estimateCoreMessages(processedMessages)) : 0;
   if (session.hostCreditTokens > 0) {
     log2("info", `[${session.id}] host usage backfill armed: +${session.hostCreditTokens} tok (forwarded view is folded); host usage will report the uncompressed baseline`);
@@ -59419,7 +59422,7 @@ function prepareAnthropic(parsed, req, opts, core, config, prompts, log2, sessio
   markDirty(session);
   const rebuilt = { ...parsed, messages: rebuiltMessages, system: systemOut, tools: toolsOut };
   delete rebuilt.prompt_cache_key;
-  armHostUsageCredit(session, originalMessages, processedMessages, log2);
+  armHostUsageCredit(session, originalMessages, processedMessages, req.headers, log2);
   return { body: JSON.stringify(rebuilt), session, processedMessages, originalMessages, anthropicSystem: parsed.system, protocol: "anthropic", stream: stream2, compressInjected: injectTools, pluginMode, nudge, prompts, renderTags: "text-only" };
 }
 var OUTPUT_CLAMP_MARGIN_PCT = 0.05;
@@ -59550,7 +59553,7 @@ function prepareOpenai(parsed, req, opts, core, config, prompts, log2, session,
   if (stream2 && rebuilt.stream_options === void 0) {
     rebuilt.stream_options = { include_usage: true };
   }
-  armHostUsageCredit(session, originalMessages, processedMessages, log2);
+  armHostUsageCredit(session, originalMessages, processedMessages, req.headers, log2);
   if (!isTitleGen && openaiOutboundSystem !== void 0) {
     session.metadata.systemPromptTokens = countSystemAndToolsTokens(openaiOutboundSystem, toolsOut);
   }
@@ -59701,7 +59704,7 @@ function prepareResponses(parsed, req, opts, core, config, prompts, log2, sessio
     });
     log2("info", `[${sessionId}] responses forward tools=[${fwdTools.join(",")}] injectTool=${injectTools}${pluginMode ? " (plugin mode: wire injection suppressed)" : ""} NO_INJECT_TOOL=${!!process.env.ACP_NO_INJECT_TOOL} NO_COMPRESS_PROMPT=${!!process.env.ACP_NO_COMPRESS_PROMPT}`);
   }
-  armHostUsageCredit(session, originalMessages, processedMessages, log2);
+  armHostUsageCredit(session, originalMessages, processedMessages, req.headers, log2);
   if (transformOk) {
     session.metadata.systemPromptTokens = countSystemAndToolsTokens(responsesDevContent ?? "", toolsOut);
   }