billion-context 0.1.118 → 0.1.119-pr.922.757

package/dist/index.js

@@ -52703,6 +52703,7 @@ function resetSessionCompression(session) {
   session.blockContents.clear();
   session.stats.lastInputTokens = 0;
   delete session.stats.lastInputTokensSource;
+  session.stats.localInputEstimate = 0;
   session.stats.contextTokens = 0;
   session.metadata.nativeCompactionAt = Date.now();
   markDirty(session);
@@ -64270,6 +64271,7 @@ ${bodyBuffer.toString("utf8")}`);
             prepared,
             runPrepare,
             req,
+            bodyBuffer,
             res,
             opts,
             core,
@@ -64482,8 +64484,11 @@ function diagNudge(turn, sessionId, tokenCount, limit, model, willInject) {
 }
 function effectiveTokenCount(session, msgs) {
   if (session.stats.lastInputTokens > 0) return session.stats.lastInputTokens;
-  if (!session.metadata.anonymousPrefixAffinity) return 0;
-  return estimateCoreMessagesUpper(msgs);
+  if (session.metadata.anonymousPrefixAffinity) return estimateCoreMessagesUpper(msgs);
+  const est = session.stats.localInputEstimate ?? 0;
+  if (est <= 0) return 0;
+  const raw = estimateCoreMessagesUpper(msgs);
+  return Math.min(est, raw);
 }
 function prepareAnthropic(parsed, req, opts, core, config, prompts, surface, log2, session, pluginMode, upstreamOrigin, reasoning) {
   const sessionId = session.id;
@@ -64555,6 +64560,7 @@ function prepareAnthropic(parsed, req, opts, core, config, prompts, surface, log
   const rebuilt = { ...parsed, messages: rebuiltMessages, system: systemOut, tools: toolsOut };
   warnAnthropicThinkingPairs(rebuiltMessages, log2, sessionId);
   delete rebuilt.prompt_cache_key;
+  session.stats.localInputEstimate = estimateCoreMessagesUpper(processedMessages.length > 0 ? processedMessages : originalMessages) + countSystemAndToolsTokens(extractSystem(systemOut), toolsOut) + imageTokensInParsedBody("anthropic", rebuilt);
   return { body: JSON.stringify(rebuilt), session, processedMessages, originalMessages, anthropicSystem: parsed.system, protocol: "anthropic", stream: stream2, compressInjected: injectTools, pluginMode, nudge, prompts, surface, renderTags: "text-only" };
 }
 function prepareOpenai(parsed, req, opts, core, config, prompts, surface, log2, session, pluginMode, upstreamOrigin, nativeWindow, reasoning, billingUpstream) {
@@ -64637,6 +64643,9 @@ function prepareOpenai(parsed, req, opts, core, config, prompts, surface, log2,
   if (!isTitleGen && openaiOutboundSystem !== void 0) {
     session.metadata.systemPromptTokens = countSystemAndToolsTokens(openaiOutboundSystem, toolsOut);
   }
+  if (!isTitleGen) {
+    session.stats.localInputEstimate = estimateCoreMessagesUpper(processedMessages.length > 0 ? processedMessages : originalMessages) + countSystemAndToolsTokens(openaiOutboundSystem || openaiSystemText, toolsOut) + imageTokensInParsedBody("openai", rebuilt);
+  }
   snapshotMessages(session, originalMessages);
   markDirty(session);
   return { body: JSON.stringify(rebuilt), session, processedMessages, originalMessages, protocol: "openai", stream: stream2, compressInjected: injectTools, pluginMode, nudge, prompts, surface, openaiSystemText, renderTags: "text-only" };
@@ -64789,6 +64798,9 @@ function prepareResponses(parsed, req, opts, core, config, prompts, surface, log
   if (transformOk) {
     session.metadata.systemPromptTokens = countSystemAndToolsTokens(responsesDevContent ?? "", toolsOut);
   }
+  if (!isCompactionTrigger) {
+    session.stats.localInputEstimate = estimateCoreMessagesUpper(processedMessages.length > 0 ? processedMessages : originalMessages) + countSystemAndToolsTokens(responsesDevContent ?? "", toolsOut) + imageTokensInParsedBody("responses", rebuilt);
+  }
   snapshotMessages(session, originalMessages);
   markDirty(session);
   return {
@@ -65056,7 +65068,7 @@ function beginPreflightHold(res, prepared, log2) {
   }, PREFLIGHT_KEEPALIVE_MS);
   return () => clearInterval(iv);
 }
-async function preflightCompressIfNeeded(prepared, runPrepare, req, res, opts, core, config, configuredWindow, model, route, affinity, anonymous, log2, instanceId) {
+async function preflightCompressIfNeeded(prepared, runPrepare, req, inboundBody, res, opts, core, config, configuredWindow, model, route, affinity, anonymous, log2, instanceId) {
   const session = prepared.session;
   const limit = config.modelContextLimit;
   const compressionTarget = prepared.protocol === "responses" && isCodexClient(req.headers) && codexCompactMode() === "intercept" ? limit * CODEX_COMPACT_HEALTH_RATIO : limit;
@@ -65094,7 +65106,7 @@ async function preflightCompressIfNeeded(prepared, runPrepare, req, res, opts, c
       return failFast(502, "no part of the conversation is compressible (nothing left to fold)", false);
     }
   }
-  const deadEndKey = `${model}\0${limit}\0${createHash8("sha256").update(prepared.body).digest("hex")}`;
+  const deadEndKey = `${model}\0${limit}\0${createHash8("sha256").update(inboundBody).digest("hex")}`;
   const deadEnd = session.metadata.preflightDeadEnd;
   if (deadEnd && typeof deadEnd === "object") {
     const de2 = deadEnd;