billion-context 0.1.59 → 0.1.60-pr.312.163

package/dist/index.js

@@ -30021,9 +30021,9 @@ var require_sha256 = __commonJS({
     var forge2 = require_forge();
     require_md();
     require_util7();
-    var sha256 = module.exports = forge2.sha256 = forge2.sha256 || {};
-    forge2.md.sha256 = forge2.md.algorithms.sha256 = sha256;
-    sha256.create = function() {
+    var sha2562 = module.exports = forge2.sha256 = forge2.sha256 || {};
+    forge2.md.sha256 = forge2.md.algorithms.sha256 = sha2562;
+    sha2562.create = function() {
       if (!_initialized) {
         _init();
       }
@@ -49830,15 +49830,17 @@ function normalizeSseLineEndings(buf) {
 }
 
 // src/loop/tag-echo-filter.ts
-var PAIRED = /\x3cacp\s[^<>]*>([^<>]{0,64})<\/acp>/;
-var LONE = /<\/?acp(?=[\s>])[^<>]*>/;
+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[^<>]*|<\/?a?c?p?)$/;
-var TRUNCATED_TAG = /\x3cacp\s[^<>]*tokens\s*=\s*"[^<>"]*"[^<>]*$/;
+var TRUNC_OPEN = /\x3cacp\s[^<>]*$/;
 var CLOSE_TAG = "</acp";
 var HOLD_LIMIT = 128;
+var TAG_OPEN_CAP = 4096;
 var SWALLOW_CAP = 80;
 function stripAcpTags(text) {
-  return text.replace(new RegExp(PAIRED.source, "g"), "").replace(new RegExp(LONE.source, "g"), "").replace(new RegExp(TRUNCATED_TAG.source), "");
+  return text.replace(new RegExp(PAIRED.source, "g"), "").replace(new RegExp(LONE_OPEN.source, "g"), "").replace(new RegExp(LONE_CLOSE.source, "g"), "").replace(new RegExp(TRUNC_OPEN.source), "");
 }
 function containsRenderTagText(s3) {
   return /\x3c\/?acp(?=[\s>])/.test(s3) || /\\u003c\/?acp(?=[\s>\\])/.test(s3);
@@ -49881,15 +49883,26 @@ function createTagEchoFilter(onDrop) {
         return out;
       }
       const p2 = PAIRED.exec(buf);
-      const l = LONE.exec(buf);
-      let m2;
-      if (p2 && l) m2 = p2.index <= l.index ? p2 : l;
-      else m2 = p2 ?? l;
+      const o = LONE_OPEN.exec(buf);
+      const c = LONE_CLOSE.exec(buf);
+      let m2 = null;
+      for (const cand of [p2, o, c]) {
+        if (cand && (m2 === null || cand.index < m2.index)) m2 = cand;
+      }
       if (!m2) {
         const t = PARTIAL_TAIL.exec(buf);
-        if (t && t[0].length <= HOLD_LIMIT) {
-          held = t[0];
-          out += buf.slice(0, buf.length - t[0].length);
+        if (t) {
+          const definite = /^\x3cacp\s/.test(t[0]);
+          const cap = definite ? TAG_OPEN_CAP : HOLD_LIMIT;
+          if (t[0].length <= cap) {
+            held = t[0];
+            out += buf.slice(0, buf.length - t[0].length);
+          } else if (definite) {
+            drop(t[0]);
+            out += buf.slice(0, buf.length - t[0].length);
+          } else {
+            out += buf;
+          }
         } else {
           out += buf;
         }
@@ -49898,8 +49911,7 @@ function createTagEchoFilter(onDrop) {
       drop(m2[0]);
       out += buf.slice(0, m2.index);
       buf = buf.slice(m2.index + m2[0].length);
-      const wasPaired = m2 === p2;
-      if (!wasPaired && /^\x3cacp\s/.test(m2[0])) {
+      if (m2 === o && /^\x3cacp\s/.test(m2[0])) {
         swallowUntilClose = true;
         swallowed = "";
       }
@@ -49913,14 +49925,19 @@ function createTagEchoFilter(onDrop) {
       return process2(chunk);
     },
     flush() {
-      let rest = swallowed + held;
+      const rest = swallowed + held;
+      const wasSwallowing = swallowUntilClose;
       swallowed = "";
       held = "";
       swallowUntilClose = false;
-      const t = new RegExp(TRUNCATED_TAG.source).exec(rest);
+      if (wasSwallowing) {
+        if (rest.length > 0) drop(rest);
+        return "";
+      }
+      const t = new RegExp(TRUNC_OPEN.source).exec(rest);
       if (t) {
         drop(t[0]);
-        rest = rest.slice(0, t.index);
+        return rest.slice(0, t.index);
       }
       return rest;
     },
@@ -52566,6 +52583,102 @@ function preferPromptCacheKeyIdentity(identity, body) {
   return { ...identity, value: pck, source: "prompt-cache-key", clientProvided: true };
 }
 
+// src/prefix-affinity.ts
+import { createHash as createHash5 } from "crypto";
+var MIN_CANONICAL_BYTES = 24;
+var MAX_TRACKED_SESSIONS = 256;
+var TTL_MS2 = 7 * 24 * 60 * 60 * 1e3;
+function stableStringify(value) {
+  return JSON.stringify(sortKeys(value));
+}
+function sortKeys(value) {
+  if (Array.isArray(value)) return value.map(sortKeys);
+  if (value && typeof value === "object") {
+    const record = value;
+    const out = {};
+    for (const key of Object.keys(record).sort()) out[key] = sortKeys(record[key]);
+    return out;
+  }
+  return value;
+}
+function sha256(text) {
+  return createHash5("sha256").update(text, "utf8").digest("hex");
+}
+function hasUserMessage(messages) {
+  return messages.some((m2) => !!m2 && typeof m2 === "object" && m2.role === "user");
+}
+function chainHashes(messages) {
+  const hashes = [];
+  let prev = "";
+  let bytes = 0;
+  for (const message of messages) {
+    const canonical = stableStringify(message);
+    bytes += canonical.length;
+    prev = sha256(`${prev}\0${canonical}`);
+    hashes.push(prev);
+  }
+  return bytes >= MIN_CANONICAL_BYTES ? hashes : [];
+}
+var PrefixAffinityResolver = class {
+  trackedChains = /* @__PURE__ */ new Map();
+  /**
+   * Resolve an anonymous request to a session id.
+   * Returns null when the request carries no usable conversation signal
+   * (caller keeps the #286 explicit 400).
+   */
+  resolve(messages) {
+    const hashes = chainHashes(messages);
+    if (hashes.length === 0 || !hasUserMessage(messages)) return null;
+    const tailHash = hashes[hashes.length - 1];
+    const tracked = this.trackedChains;
+    this.expire(tracked);
+    let best;
+    for (const entry of tracked.values()) {
+      if (entry.depth > hashes.length) continue;
+      if (hashes[entry.depth - 1] !== entry.tailHash) continue;
+      if (!best || entry.depth > best.depth || entry.depth === best.depth && entry.lastSeen > best.lastSeen) best = entry;
+    }
+    if (best) {
+      return {
+        sessionId: best.sessionId,
+        matchedDepth: best.depth,
+        storedDepth: best.depth,
+        incomingDepth: hashes.length,
+        tailHash
+      };
+    }
+    const sessionId = `pfa-${sha256(`${tailHash}\0${hashes.length}`).slice(0, 16)}`;
+    return { sessionId, matchedDepth: 0, storedDepth: 0, incomingDepth: hashes.length, tailHash };
+  }
+  /** Record the chain of a session (on creation and after every anonymous
+   *  request — the incoming history is the truth, appends extend it). */
+  note(sessionId, depth, tailHash) {
+    const tracked = this.trackedChains;
+    tracked.delete(sessionId);
+    tracked.set(sessionId, { sessionId, depth, tailHash, lastSeen: Date.now() });
+    while (tracked.size > MAX_TRACKED_SESSIONS) {
+      const oldest = [...tracked.values()].sort((a, b2) => a.lastSeen - b2.lastSeen)[0];
+      if (!oldest) break;
+      tracked.delete(oldest.sessionId);
+    }
+  }
+  /** Drop tracking for a removed session. */
+  forget(sessionId) {
+    this.trackedChains.delete(sessionId);
+  }
+  trackedSessionIds() {
+    return [...this.trackedChains.keys()];
+  }
+  expire(tracked) {
+    if (tracked.size === 0) return;
+    const now = Date.now();
+    for (const [id, entry] of tracked) {
+      if (now - entry.lastSeen > TTL_MS2) tracked.delete(id);
+    }
+  }
+};
+var prefixAffinity = new PrefixAffinityResolver();
+
 // node_modules/acp-kernel/dist/panel/index.js
 function topicFallback(summary) {
   const first = summary.split(/[.\n]/)[0] ?? "";
@@ -53791,7 +53904,7 @@ function collectModelWindows(config) {
 }
 
 // src/discover.ts
-var TTL_MS2 = 2e3;
+var TTL_MS3 = 2e3;
 function unwrapUpstream(url) {
   const idx = url.indexOf("/bili/");
   return idx >= 0 ? url.slice(idx + "/bili/".length) : url;
@@ -53859,7 +53972,7 @@ function mtimesEqual(a, b2) {
 var cache3 = null;
 function discoverMitmDomains(env = process.env) {
   const now = Date.now();
-  if (cache3 && now - cache3.checkedAt < TTL_MS2) {
+  if (cache3 && now - cache3.checkedAt < TTL_MS3) {
     return cache3.domains;
   }
   const paths = configFilePaths(env);
@@ -54742,7 +54855,7 @@ function launcherContextWindow(model) {
 function safeSessionId(id) {
   return (id ?? "unknown").replace(/[^a-zA-Z0-9._-]/g, "_");
 }
-var NO_IDENTITY_MESSAGE = "Missing stable conversation identity. Send one of the headers: x-session-id, x-session-affinity, x-acp-session, x-opencode-session, x-claude-code-session-id, session-id \u2014 or body session_id / prompt_cache_key (responses/openai). Content-fingerprint sessions are disabled (#286).";
+var NO_IDENTITY_MESSAGE = "Missing stable conversation identity. Send one of the headers: x-session-id, x-session-affinity, x-acp-session, x-opencode-session, x-claude-code-session-id, session-id \u2014 or body session_id / prompt_cache_key (responses/openai). Requests replaying a conversation history are matched by content prefix affinity (#309); this one carries no usable history signal.";
 var UPSTREAM_HOP_HEADERS = /* @__PURE__ */ new Set([
   "host",
   "content-length",
@@ -55194,13 +55307,25 @@ ${bodyBuffer.toString("utf8")}`);
       parsed.instructions
     );
     const clientProvided = protocol === "responses" ? responsesIdentity?.clientProvided ?? false : protocol === "openai" ? openaiIdentity?.clientProvided ?? false : !!convHeader;
+    let anonAffinity = null;
     if (!clientProvided) {
-      log2("warn", `400: no stable conversation identity on ${protocol} request \u2192 ${upstreamOrigin}; refusing to create a content-fingerprint session (#286)`);
-      res.writeHead(400, { "content-type": "application/json" });
-      res.end(JSON.stringify(protocol === "anthropic" ? { type: "error", error: { type: "invalid_request_error", message: NO_IDENTITY_MESSAGE } } : { error: { type: "invalid_request_error", message: NO_IDENTITY_MESSAGE } }));
-      return;
+      const anonMessages = protocol === "responses" ? parsed.input ?? [] : parsed.messages ?? [];
+      anonAffinity = prefixAffinity.resolve(Array.isArray(anonMessages) ? anonMessages : []);
+      if (!anonAffinity) {
+        log2("warn", `400: no stable conversation identity on ${protocol} request \u2192 ${upstreamOrigin}; refusing to create a content-fingerprint session (#286)`);
+        res.writeHead(400, { "content-type": "application/json" });
+        res.end(JSON.stringify(protocol === "anthropic" ? { type: "error", error: { type: "invalid_request_error", message: NO_IDENTITY_MESSAGE } } : { error: { type: "invalid_request_error", message: NO_IDENTITY_MESSAGE } }));
+        return;
+      }
+      if (anonAffinity.matchedDepth > 0) {
+        log2("info", `[prefix-affinity] anonymous ${protocol} request \u2192 session ${anonAffinity.sessionId} (prefix match depth=${anonAffinity.matchedDepth}/${anonAffinity.incomingDepth}, tail=${anonAffinity.tailHash.slice(0, 8)}; fork semantics: diverged histories split on their next request)`);
+        log("info", `[prefix-affinity] session ${anonAffinity.sessionId} matched at depth ${anonAffinity.matchedDepth}/${anonAffinity.incomingDepth} (tail=${anonAffinity.tailHash.slice(0, 8)})`);
+      } else {
+        log2("info", `[prefix-affinity] new anonymous session ${anonAffinity.sessionId} (depth=${anonAffinity.incomingDepth}, tail=${anonAffinity.tailHash.slice(0, 8)})`);
+        log("info", `[prefix-affinity] new session ${anonAffinity.sessionId} at depth ${anonAffinity.incomingDepth} (tail=${anonAffinity.tailHash.slice(0, 8)})`);
+      }
     }
-    const sessionId = conversation;
+    const sessionId = anonAffinity ? anonAffinity.sessionId : conversation;
     const bodyIdentity = responsesIdentity ?? openaiIdentity;
     const affinity = affinityToken(bodyIdentity ?? {
       value: clientConv ?? conversation,
@@ -55208,8 +55333,15 @@ ${bodyBuffer.toString("utf8")}`);
       clientProvided: !!clientConv
     });
     const clientLabel = bodyIdentity?.clientProvided ? bodyIdentity.value : clientConversationHeader(req.headers);
-    const session = getSession(sessionId, { protocol, upstreamOrigin, label: clientLabel ?? void 0 });
-    if (!pluginAgent) {
+    const session = getSession(sessionId, { protocol, upstreamOrigin, label: clientLabel ?? (anonAffinity ? "prefix-affinity" : void 0) });
+    if (anonAffinity) {
+      prefixAffinity.note(sessionId, anonAffinity.incomingDepth, anonAffinity.tailHash);
+      session.metadata.anonymousPrefixAffinity = {
+        depth: anonAffinity.incomingDepth,
+        tailHash: anonAffinity.tailHash
+      };
+    }
+    if (!pluginAgent && !anonAffinity) {
       const identityAgent = consumePluginRegisterFor(clientConv ?? conversation);
       if (identityAgent) {
         pluginAgent = identityAgent;