billion-context 0.1.82 → 0.1.83-pr.399.349

package/README.md

@@ -112,6 +112,17 @@ turn is a semantic mismatch the model tolerates (it is clearly marked
   summaries (their tool call isn't in the agent's history and the agent's view
   skips `acp_summary`) — but that needs the id match above, which doesn't occur.
 
+## Which do I need?
+
+Pick by your client:
+
+| Client | Use |
+|---|---|
+| **pi** | [`billion-context-pi`](https://github.com/ranxianglei/billion-context-pi) (in-process extension) |
+| **opencode** | [`opencode-acp`](https://github.com/ranxianglei/opencode-acp) (in-process extension) |
+| **omp** | [`billion-context`](https://github.com/ranxianglei/billion-context) via `bili omp` (built-in plugin) |
+| **everything else** (no context hook) | [`billion-context`](https://github.com/ranxianglei/billion-context) — `bili <client>` (launcher, preferred) or `/bili/` prefix |
+
 ## Install
 
 ```bash
@@ -382,7 +393,7 @@ pass an explicit `x-acp-session` header per conversation to avoid collisions.
 
 Early. Protocol handling and compression work against mock tests (500+ passing). Real-model integration testing is the next milestone. Expect rough edges.
 
-See [billion-context-pi](https://github.com/ranxianglei/billion-context-pi) for the pi-extension mode (in-process, tighter integration, the reference implementation).
+Client-side plugins for pi / omp / opencode ship inside `billion-context` (`dist/agent/*.js`) for the cooperative-proxy path. See the **"Which do I need?"** section above for how `billion-context`, the standalone `billion-context-pi`, and `opencode-acp` relate.
 
 ## License
 

package/README.zh-CN.md

@@ -50,6 +50,17 @@ AI 编程助手的<strong>通用上下文压缩代理</strong>
 
 代理向对话注入四个上下文管理工具(`compress`、`decompress`、`search_context`、`acp_status`)。模型在对话增长时调用 `compress`,代理在服务端执行 —— 压缩后的范围在下一轮之前折叠进对话历史。
 
+## 该选哪个?
+
+按客户端选:
+
+| 客户端 | 用这个 |
+|---|---|
+| **pi** | [`billion-context-pi`](https://github.com/ranxianglei/billion-context-pi)(进程内扩展) |
+| **opencode** | [`opencode-acp`](https://github.com/ranxianglei/opencode-acp)(进程内扩展) |
+| **omp** | [`billion-context`](https://github.com/ranxianglei/billion-context),`bili omp`(内置插件) |
+| **其余所有**(没有上下文 hook) | [`billion-context`](https://github.com/ranxianglei/billion-context) —— `bili <client>`(启动器,优先)或 `/bili/` 前缀 |
+
 ## 安装
 
 ```bash
@@ -258,7 +269,7 @@ Windows 下会自动发现常见 Clash/Mihomo 静态系统代理;Web UI 会显
 
 早期。协议处理和压缩已通过 mock 测试(500+ 项通过)。真实模型集成测试是下一里程碑。预期会有粗糙的地方。
 
-pi 扩展模式(进程内、更紧密集成、参考实现)见 [billion-context-pi](https://github.com/ranxianglei/billion-context-pi)。
+针对 pi / omp / opencode 的客户端插件随 `billion-context` 一起发布(`dist/agent/*.js`),用于协作代理路径。三者(`billion-context`、独立的 `billion-context-pi`、`opencode-acp`)如何取舍,见上文「该选哪个?」一节。
 
 ## 许可证
 

package/dist/index.js

@@ -43476,10 +43476,6 @@ function defaultCountTokens(text) {
   const cjkCount = cjk?.length ?? 0;
   return cjkCount + Math.ceil((text.length - cjkCount) / 4);
 }
-function estimateTokensFast(text) {
-  if (!text) return 0;
-  return Math.ceil(text.length / 4);
-}
 var COMPRESS_PHILOSOPHY = `Compression Philosophy:
 - All compression serves the primary task, but be frugal.
 - Context capacity is precious. Save context by compressing consumed outputs, not by avoiding tools.
@@ -48360,33 +48356,64 @@ async function contextFromRegistry(model, host) {
 function peekRegistryContext(model, host) {
   return registryLookup(cache2, model, host);
 }
+var VARIANT_SUFFIXES = [
+  "-interleaved-thinking",
+  "-thinking",
+  "-reasoning",
+  "-extended",
+  "-fast",
+  "-high",
+  "-medium",
+  "-low"
+];
+function modelVariants(name) {
+  const variants = [name];
+  let current = name;
+  for (; ; ) {
+    let stripped = false;
+    for (const suffix of VARIANT_SUFFIXES) {
+      if (current.length > suffix.length && current.endsWith(suffix)) {
+        current = current.slice(0, -suffix.length);
+        variants.push(current);
+        stripped = true;
+        break;
+      }
+    }
+    if (!stripped) break;
+  }
+  return variants;
+}
 function registryLookup(reg, model, host) {
   if (!reg || !model) return void 0;
   const provider = host ? providerFromHost(host) : void 0;
-  const candidates = provider ? [`${provider}/${model}`, model] : [model];
-  for (const key of candidates) {
-    const entry = reg[key];
-    const ctx = entry?.limit?.context;
-    if (typeof ctx === "number" && ctx > 0) return ctx;
-  }
-  if (provider === void 0) {
-    const suffix = `/${model}`;
-    let max;
-    const distinct = /* @__PURE__ */ new Set();
-    const parts = [];
-    for (const key of Object.keys(reg)) {
-      if (!key.endsWith(suffix)) continue;
-      const ctx = reg[key].limit?.context;
-      if (typeof ctx !== "number" || ctx <= 0) continue;
-      if (max === void 0 || ctx > max) max = ctx;
-      distinct.add(ctx);
-      parts.push(`${key}=${ctx}`);
-    }
-    if (max !== void 0 && distinct.size > 1 && !warnedConflicts.has(model)) {
-      warnedConflicts.add(model);
-      log("warn", `[acp-registry] conflicting context windows for "${model}" (${parts.join(", ")}) \u2014 using max ${max}`);
+  for (const name of modelVariants(model)) {
+    const candidates = provider ? [`${provider}/${name}`, name] : [name];
+    for (const key of candidates) {
+      const entry = reg[key];
+      const ctx = entry?.limit?.context;
+      if (typeof ctx === "number" && ctx > 0) return ctx;
+    }
+    if (provider === void 0) {
+      const suffix = `/${name}`;
+      let max;
+      const distinct = /* @__PURE__ */ new Set();
+      const parts = [];
+      for (const key of Object.keys(reg)) {
+        if (!key.endsWith(suffix)) continue;
+        const ctx = reg[key].limit?.context;
+        if (typeof ctx !== "number" || ctx <= 0) continue;
+        if (max === void 0 || ctx > max) max = ctx;
+        distinct.add(ctx);
+        parts.push(`${key}=${ctx}`);
+      }
+      if (max !== void 0) {
+        if (distinct.size > 1 && !warnedConflicts.has(name)) {
+          warnedConflicts.add(name);
+          log("warn", `[acp-registry] conflicting context windows for "${name}" (${parts.join(", ")}) \u2014 using max ${max}`);
+        }
+        return max;
+      }
     }
-    return max;
   }
   return void 0;
 }
@@ -50581,7 +50608,7 @@ function executeAnthropicProxyTool(toolName, args, ctx) {
 ${lines.join("\n\n")}`;
   }
   if (toolName === "acp_status") {
-    return buildStatusReport(ctx.session.state, ctx.messages, estimateTokensFast);
+    return buildStatusReport(ctx.session.state, ctx.messages, defaultCountTokens);
   }
   return `[Unknown proxy tool: ${toolName}]`;
 }
@@ -51702,7 +51729,7 @@ function handleAcpStatus(args, ctx) {
   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 });
+  const base = buildStatusReport(ctx.session.state, ctx.messages, defaultCountTokens, { scope, view, tool, sort, limit });
   if (scope) return base;
   const nudge = ctx.nudge;
   const ranges = nudge?.compressibleRanges ?? [];
@@ -53547,7 +53574,7 @@ function executeProxyTool2(toolName, args, ctx) {
 ${lines.join("\n\n")}`;
   }
   if (toolName === "acp_status") {
-    return buildStatusReport(ctx.session.state, ctx.messages, estimateTokensFast);
+    return buildStatusReport(ctx.session.state, ctx.messages, defaultCountTokens);
   }
   return `[Unknown proxy tool: ${toolName}]`;
 }
@@ -54705,7 +54732,7 @@ function handlePluginStatus(conversationId2, res, fallbackLatest = false) {
   if (entry) entry.lastSeen = Date.now();
   const limit = session.metadata.effectiveContextLimit;
   const mem = remembered.get(session.id);
-  const modelContextLimit = typeof limit === "number" && limit > 0 ? limit : 2e5;
+  const modelContextLimit = typeof limit === "number" && limit > 0 ? limit : 0;
   let panel;
   try {
     panel = buildStatusPanel({
@@ -56729,6 +56756,9 @@ function logUnrecognizedPath(log2, url) {
     log2("info", `unrecognized path ${key}: forwarding unchanged; further occurrences suppressed`);
   }
 }
+function isModelDiscoveryPath(path18) {
+  return path18.replace(/\/+$/, "").endsWith("/models");
+}
 var BILI_HOP_HEADER = "x-bili-hop";
 function parseLauncherModelWindows(raw) {
   if (!raw) return {};
@@ -56930,7 +56960,7 @@ Content-Length: ${Buffer.byteLength(body)}\r
       );
     }
     if (opts.debug) {
-      log2("info", `[debug] build features: raw-HTTP-capture(on) | remote_compaction_v2-strip(on) | cert-MITM-launcher(on) | strip-acp-summary(on) \u2014 seeing this line confirms the launcher build (not registry 0.1.34)`);
+      log2("info", `[debug] build features: raw-HTTP-capture(${bodyDumpEnabled() ? "on" : "off"}) | remote_compaction_v2-strip(on) | cert-MITM-launcher(on) | strip-acp-summary(on) \u2014 seeing this line confirms the launcher build (not registry 0.1.34)`);
     }
   };
   const attemptListen = (port) => {
@@ -56998,6 +57028,13 @@ 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;
@@ -57274,11 +57311,11 @@ ${bodyBuffer.toString("utf8")}`);
       const pluginWindow = pluginReportedContextWindow(req.headers);
       const launcherWindow = launcherContextWindow(model);
       const configuredWindow = resolveConfiguredContextLimit(opts.routes, embeddedUrl, model);
-      const peekWindow = host ? peekRegistryContext(model, host) : void 0;
+      const peekWindow = peekRegistryContext(model, host);
       let native = betaWindow ?? pluginWindow ?? launcherWindow ?? configuredWindow ?? peekWindow ?? lookupContextLimit(model);
       const operatorWindowTuned = resolveCompress(opts.routes, embeddedUrl, model, opts.compress).modelContextLimit !== void 0;
       nativeFromFallback = !betaWindow && !pluginWindow && !launcherWindow && !peekWindow && !configuredWindow && !operatorWindowTuned;
-      if (!native && host) {
+      if (!native) {
         native = await contextFromRegistry(model, host);
         if (native) nativeFromFallback = false;
       }
@@ -57379,7 +57416,6 @@ ${bodyBuffer.toString("utf8")}`);
     if (pluginAgent && !pluginConversation) pluginConversation = conversation;
     if (pluginAgent) {
       if (session.metadata.pluginAgent !== pluginAgent) session.metadata.pluginAgent = pluginAgent;
-      session.metadata.effectiveContextLimit = reqConfig.modelContextLimit;
       recordPluginSession(pluginConversation ?? conversation, session.id);
     }
     if (protocol === "responses" || protocol === "openai" || protocol === "anthropic") {
@@ -57389,6 +57425,21 @@ ${bodyBuffer.toString("utf8")}`);
       }
     }
     const pluginMode = pluginAgent !== void 0;
+    if (!countTokens && !responsesCompact && protocol !== null && isSideRequest(parsed)) {
+      log2("info", `[${session.id}] side request (max_tokens<=${SIDE_REQUEST_MAX_TOKENS}) \u2192 passthrough + tag strip only, kernel state untouched`);
+      const sidePrepared = {
+        body: bodyBuffer,
+        session,
+        processedMessages: [],
+        originalMessages: [],
+        protocol,
+        stream: parsed.stream === true,
+        compressInjected: false,
+        sidePassthrough: true
+      };
+      await forward(req, res, opts, bodyBuffer, sidePrepared, 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;
@@ -57397,6 +57448,22 @@ ${bodyBuffer.toString("utf8")}`);
       reqConfig = { ...reqConfig, modelContextLimit: learnedLimit };
       nativeFromFallback = false;
       log2("info", `[${session.id}] self-healed context window: ${resolved} \u2192 ${learnedLimit} (learned from an upstream overflow)`);
+    } else if (nativeFromFallback && reqModel) {
+      const prevInput = session.stats.lastInputTokens ?? 0;
+      const prevWindow = session.metadata.lastTurnWindow;
+      const resolved = reqConfig.modelContextLimit;
+      if (prevWindow !== void 0 && prevInput > prevWindow && prevInput > resolved && prevInput >= 1e3) {
+        const map = session.metadata.learnedContextLimits ?? {};
+        const prev = map[reqModel];
+        if (prev === void 0 || prevInput > prev) {
+          map[reqModel] = prevInput;
+          session.metadata.learnedContextLimits = map;
+          markDirty(session);
+        }
+        reqConfig = { ...reqConfig, modelContextLimit: prevInput };
+        nativeFromFallback = false;
+        log2("info", `[${session.id}] self-healed context window upward: ${resolved} \u2192 ${prevInput} (a prior turn used ${prevInput} input tokens, exceeding the fallback window)`);
+      }
     }
     const nativeWindow = reqConfig.modelContextLimit;
     if (shouldReserveOutputHeadroom(protocol)) {
@@ -57410,6 +57477,8 @@ ${bodyBuffer.toString("utf8")}`);
       }
       if (reserved !== reqConfig.modelContextLimit) reqConfig = { ...reqConfig, modelContextLimit: reserved };
     }
+    session.metadata.effectiveContextLimit = reqConfig.modelContextLimit;
+    session.metadata.lastTurnWindow = reqConfig.modelContextLimit;
     acquireInFlight(session);
     try {
       await withSessionLock(session, async () => {
@@ -57474,7 +57543,7 @@ ${bodyBuffer.toString("utf8")}`);
     }
   }
   if (!prepared) {
-    if (protocol === null && !opts.passthrough) {
+    if (protocol === null && !opts.passthrough && !isModelDiscoveryPath(urlPath)) {
       logUnrecognizedPath(log2, req.url ?? "");
     }
     await forward(req, res, opts, bodyBuffer, null, core, reqConfig, log2, route, instanceId, void 0);
@@ -58311,7 +58380,7 @@ ${hdrText}
   }
   let responseBody = upstream.body;
   try {
-    if (prepared !== null && prepared.stream && maxFakeCompletionRetries() > 0) {
+    if (prepared !== null && prepared.stream && !prepared.sidePassthrough && maxFakeCompletionRetries() > 0) {
       const resolvedBuf = await resolveFakeCompletion(upstream.body, {
         protocol: prepared.protocol,
         body,
@@ -62631,7 +62700,6 @@ function selfDistFile(name) {
   return path15.join(selfPackageRoot(), "dist", name);
 }
 var LAUNCHER_DEFAULT_HOST = "127.0.0.1";
-var LAUNCHER_DEFAULT_PORT = 8787;
 var LAUNCH_CLIENTS = ["pi", "codex", "claude", "omp", "opencode", "hermes", "dsh", "pi-test"];
 var HEALTH_PATH = "/__bili/health";
 var HEALTH_POLL_INTERVAL_MS = 200;
@@ -63805,6 +63873,19 @@ async function probeExistingInstance(readInstance, fetchHealthInfo) {
   if (health.instanceId !== void 0 && health.instanceId !== inst.instanceId) return void 0;
   return inst;
 }
+function pickEphemeralPort(host = LAUNCHER_DEFAULT_HOST) {
+  return new Promise((resolve, reject) => {
+    const srv = net2.createServer();
+    srv.once("error", reject);
+    srv.listen(0, host, () => {
+      const addr = srv.address();
+      srv.close(() => {
+        if (addr && typeof addr === "object") resolve(addr.port);
+        else reject(new Error("could not allocate a free port"));
+      });
+    });
+  });
+}
 var INHERITED_PROXY_VARS = ["http_proxy", "https_proxy", "all_proxy", "HTTP_PROXY", "HTTPS_PROXY", "ALL_PROXY"];
 function stripInheritedProxy(env) {
   const cleaned = { ...env };
@@ -63830,7 +63911,7 @@ async function ensureProxyRunning(opts, deps = {}) {
     return { origin: existing.origin, port: existing.port, attached: true };
   }
   const launchToken = randomUUID5();
-  const port = opts.port;
+  const port = opts.port > 0 ? opts.port : await pickEphemeralPort(opts.host);
   const script = process.argv[1];
   if (!script) throw new Error("bili: cannot resolve launcher script path");
   const logPath2 = path15.join(os5.tmpdir(), `bili-proxy-${port}.log`);
@@ -63839,7 +63920,7 @@ async function ensureProxyRunning(opts, deps = {}) {
   try {
     child = spawnImpl(
       process.execPath,
-      [script, ...proxyStartArgs(opts)],
+      [script, ...proxyStartArgs({ ...opts, port })],
       {
         detached: true,
         stdio: ["ignore", logFd, logFd],
@@ -63946,8 +64027,9 @@ function resolveClientCommand(client, env) {
   return { command: resolved ?? client, prefixArgs: [] };
 }
 function parsePort(raw) {
-  const port = raw && raw.trim() ? parseInt(raw, 10) : LAUNCHER_DEFAULT_PORT;
-  if (!Number.isFinite(port) || port <= 0 || port > 65535) {
+  if (!raw || !raw.trim()) return 0;
+  const port = parseInt(raw, 10);
+  if (!Number.isFinite(port) || port < 1 || port > 65535) {
     console.error(`bili: invalid --port "${raw}"`);
     process.exit(2);
   }
@@ -64386,7 +64468,7 @@ Launcher (bili pi / bili codex / bili claude / bili omp / bili opencode / bili h
 Options (override config file / env):
   -F <url>                         upstream proxy to forward through (gost-style;
                                    http://host:port; must precede the client name)
-  --port <N>                       listen port (default 8787)
+  --port <N>                       listen port (start: 8787; launcher default: random free port)
   --host <ADDR>                    listen host (default 127.0.0.1)
   --mitm-domain <domain>           extra MITM domain (repeatable; launcher only)
   --config <FILE>                  path to config JSON (default: XDG location)