billion-context 0.1.67-pr.397.236 → 0.1.67-pr.398.237

package/dist/index.js

@@ -50110,41 +50110,6 @@ function withStagedCompressGuidance(text) {
   return text + STAGED_COMPRESS_GUIDANCE;
 }
 
-// src/acp-status.ts
-function handleAcpStatus(args, ctx) {
-  const scope = typeof args.scope === "string" ? args.scope : void 0;
-  const view = typeof args.view === "string" ? args.view : void 0;
-  const tool = typeof args.tool === "string" ? args.tool : void 0;
-  const sort = typeof args.sort === "string" ? args.sort : void 0;
-  const limit = typeof args.limit === "number" ? args.limit : void 0;
-  const base = buildStatusReport(ctx.session.state, ctx.messages, estimateTokensFast, { scope, view, tool, sort, limit });
-  if (scope) return base;
-  const extra = [];
-  try {
-    const turn = ctx.core.processTurn({
-      messages: ctx.messages,
-      state: ctx.session.state,
-      config: ctx.config,
-      tokenCount: ctx.session.stats.lastInputTokens,
-      renderTags: "none"
-    });
-    const nudge = turn.nudge;
-    if (nudge) {
-      extra.push("");
-      extra.push(nudge.shouldInject ? `Nudge: ACTIVE \u2014 ${nudge.reason}` : `Nudge: idle \u2014 ${nudge.reason}`);
-      const ranges = viableRanges(nudge.compressibleRanges);
-      const protectedRanges = nudge.protectedRanges ?? [];
-      if (ranges.length > 0 || protectedRanges.length > 0) {
-        extra.push("");
-        extra.push(formatRanges(ranges, protectedRanges));
-      }
-    }
-  } catch {
-  }
-  return extra.length > 0 ? `${base}
-${extra.join("\n")}` : base;
-}
-
 // src/decompress-shared.ts
 import { mkdirSync as mkdirSync3, unlinkSync, writeFileSync as writeFileSync2 } from "fs";
 import { dirname as dirname2, join as join2 } from "path";
@@ -50425,7 +50390,7 @@ function executeAnthropicProxyTool(toolName, args, ctx) {
 ${lines.join("\n\n")}`;
   }
   if (toolName === "acp_status") {
-    return handleAcpStatus(args, ctx);
+    return buildStatusReport(ctx.session.state, ctx.messages, estimateTokensFast);
   }
   return `[Unknown proxy tool: ${toolName}]`;
 }
@@ -51247,6 +51212,29 @@ ${lines.join("\n\n")}`;
   }
   return `[Unknown proxy tool: ${toolName}]`;
 }
+function handleAcpStatus(args, ctx) {
+  const scope = typeof args.scope === "string" ? args.scope : void 0;
+  const view = typeof args.view === "string" ? args.view : void 0;
+  const tool = typeof args.tool === "string" ? args.tool : void 0;
+  const sort = typeof args.sort === "string" ? args.sort : void 0;
+  const limit = typeof args.limit === "number" ? args.limit : void 0;
+  const base = buildStatusReport(ctx.session.state, ctx.messages, estimateTokensFast, { scope, view, tool, sort, limit });
+  if (scope) return base;
+  const nudge = ctx.nudge;
+  const ranges = nudge?.compressibleRanges ?? [];
+  const protectedRanges = nudge?.protectedRanges ?? [];
+  const extra = [];
+  if (nudge) {
+    extra.push("");
+    extra.push(nudge.shouldInject ? `Nudge: ACTIVE \u2014 ${nudge.reason}` : `Nudge: idle \u2014 ${nudge.reason}`);
+  }
+  if (ranges.length > 0 || protectedRanges.length > 0) {
+    extra.push("");
+    extra.push(formatRanges(ranges, protectedRanges));
+  }
+  return extra.length > 0 ? `${base}
+${extra.join("\n")}` : base;
+}
 function recordUsage(ctx, usage, round) {
   const prompt = usage.inputTokens;
   const cached = usage.cachedTokens;
@@ -52913,7 +52901,7 @@ function executeProxyTool2(toolName, args, ctx) {
 ${lines.join("\n\n")}`;
   }
   if (toolName === "acp_status") {
-    return handleAcpStatus(args, ctx);
+    return buildStatusReport(ctx.session.state, ctx.messages, estimateTokensFast);
   }
   return `[Unknown proxy tool: ${toolName}]`;
 }
@@ -53911,7 +53899,8 @@ async function handlePluginTool(payload, res, deps) {
         config: deps.config,
         messages,
         session,
-        log: (m2) => deps.log("info", `[${session.id}] [plugin] ${m2}`)
+        log: (m2) => deps.log("info", `[${session.id}] [plugin] ${m2}`),
+        nudge: mem?.nudge
       }, callId);
     });
   } catch (err2) {
@@ -55786,6 +55775,13 @@ Content-Length: ${Buffer.byteLength(body)}\r
   }
   return server;
 }
+var SIDE_REQUEST_MAX_TOKENS = 200;
+function isSideRequest(parsed) {
+  if (!parsed || typeof parsed !== "object") return false;
+  const p2 = parsed;
+  const raw = p2.max_tokens ?? p2.max_completion_tokens ?? p2.max_output_tokens;
+  return typeof raw === "number" && raw > 0 && raw <= SIDE_REQUEST_MAX_TOKENS;
+}
 function isTrustedAdminOrigin(origin, host, trustedHosts) {
   if (!host || !trustedHosts.has(host.toLowerCase())) return false;
   if (!origin) return true;
@@ -56147,6 +56143,11 @@ ${bodyBuffer.toString("utf8")}`);
       }
     }
     const pluginMode = pluginAgent !== void 0;
+    if (!countTokens && !responsesCompact && protocol !== null && isSideRequest(parsed)) {
+      log2("info", `[${session.id}] side request (max_tokens<=${SIDE_REQUEST_MAX_TOKENS}) \u2192 pure passthrough, kernel state untouched`);
+      await forward(req, res, opts, bodyBuffer, null, core, reqConfig, log2, route, instanceId, affinity);
+      return;
+    }
     const reqModel = parsed.model;
     const learnedMap = session.metadata.learnedContextLimits;
     const learnedLimit = (reqModel && learnedMap ? learnedMap[reqModel] : void 0) ?? session.metadata.learnedContextLimit;
@@ -57021,7 +57022,7 @@ ${hdrText}
       const adapter = pickAdapter(prepared.protocol, parsedReq, textProtocol, prepared.responsesProjection, prepared.anthropicSystem, prepared.openaiSystemText);
       const loop = runCompressLoop(
         streamToRead,
-        { core, config, messages: prepared.processedMessages.length > 0 ? prepared.processedMessages : prepared.originalMessages, compressMessages: prepared.originalMessages, session: prepared.session, log: ctx.log, proxyUrl, protocol: prepared.protocol, textProtocol, debug: opts.debug },
+        { core, config, messages: prepared.processedMessages.length > 0 ? prepared.processedMessages : prepared.originalMessages, compressMessages: prepared.originalMessages, session: prepared.session, log: ctx.log, proxyUrl, protocol: prepared.protocol, textProtocol, debug: opts.debug, nudge: prepared.nudge },
         parsedReq,
         { url: upstreamUrl, headers: reqHeaders },
         adapter,