billion-context 0.1.104 → 0.1.105-pr.691.546

package/README.md

@@ -24,6 +24,12 @@ Any agent that can set a base URL — <em>zero per-agent adapter code</em>.
 
 `billion-context` sits between **any** agent and its model API, rewriting Anthropic/OpenAI streams with [acp-kernel](https://github.com/ranxianglei/acp-kernel) compression. The model decides **when** and **what** to compress into high-fidelity summaries — not a hard truncation limit.
 
+## Community
+
+Discussion, help, and updates on QQ — one group covers all three projects (`billion-context`, `billion-context-pi`, `opencode-acp`):
+
+**QQ Group: 1056132097**
+
 ## Why
 
 Long coding sessions blow up context. Each provider charges per token, and once you pass the context window the session degrades or dies. `billion-context` compresses consumed conversation into layered summaries so you can run a single session for days — billions of tokens through one context window.
@@ -466,6 +472,10 @@ Early. Protocol handling and compression work against mock tests (500+ passing).
 
 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.
 
+## Community
+
+QQ group — one shared group for all three projects ([`billion-context`](https://github.com/ranxianglei/billion-context), [`billion-context-pi`](https://github.com/ranxianglei/billion-context-pi), [`opencode-acp`](https://github.com/ranxianglei/opencode-acp)): **1056132097**
+
 ## License
 
 MIT

package/README.zh-CN.md

@@ -24,6 +24,12 @@ AI 编程助手的<strong>通用上下文压缩代理</strong>
 
 `billion-context` 架在**任意**编程助手与其模型 API 之间,用 [acp-kernel](https://github.com/ranxianglei/acp-kernel) 压缩重写 Anthropic/OpenAI 流。何时压缩、压缩什么 —— <strong>由模型决定</strong>,而非硬截断。
 
+## 社区
+
+交流、求助与更新都在 QQ——同一个群覆盖三个项目(`billion-context`、`billion-context-pi`、`opencode-acp`):
+
+**QQ 群:1056132097**
+
 ## 为什么
 
 长编程会话会把上下文撑爆。各家 provider 按 token 计费,一旦超过上下文窗口,会话质量下降甚至崩掉。`billion-context` 把已消耗的对话压缩成分层摘要,让你**一个会话连跑数天** —— 海量 token 穿过同一个上下文窗口。
@@ -287,6 +293,10 @@ Windows 下会自动发现常见 Clash/Mihomo 静态系统代理;Web UI 会显
 
 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。
 
+## 社区
+
+QQ 群 —— 三个项目共用一个群（[`billion-context`](https://github.com/ranxianglei/billion-context)、[`billion-context-pi`](https://github.com/ranxianglei/billion-context-pi)、[`opencode-acp`](https://github.com/ranxianglei/opencode-acp)）：**1056132097**
+
 ## 许可证
 
 MIT

package/dist/index.js

@@ -48626,7 +48626,6 @@ function loadOptions(env = process.env) {
     passthrough: passthrough.enabled,
     passthroughSource: passthrough.source,
     autoUpdate: (env.ACP_AUTO_UPDATE ?? (fileConfig.autoUpdate === false ? "0" : "1")) !== "0",
-    hostUsageCredit: parseHostUsageCredit(env.BILI_HOST_USAGE_CREDIT ?? fileConfig.hostUsageCredit),
     updateTag: (env.ACP_UPDATE_TAG ?? fileConfig.updateTag ?? "latest").trim() || "latest",
     logFile: env.ACP_LOG_FILE !== void 0 ? env.ACP_LOG_FILE || void 0 : fileConfig.logFile,
     mitm: {
@@ -48704,9 +48703,6 @@ function parsePromptCacheRouting(value) {
 function parseUpstreamProxyMode(value) {
   return value === "manual" || value === "auto" ? value : "direct";
 }
-function parseHostUsageCredit(value) {
-  return value === "off" ? "off" : "auto";
-}
 function parseCompressSettings(v2) {
   if (!v2 || typeof v2 !== "object" || Array.isArray(v2)) return void 0;
   const obj = v2;
@@ -51305,8 +51301,6 @@ function resetSessionCompression(session) {
   session.blockContents.clear();
   session.stats.lastInputTokens = 0;
   session.stats.contextTokens = 0;
-  session.hostCreditTokens = 0;
-  session.hostContextTokens = 0;
   session.metadata.nativeCompactionAt = Date.now();
   markDirty(session);
 }
@@ -53340,23 +53334,6 @@ function promptInputTotal(protocol, input, cached) {
   const splitSemantics = !includesCached || typeof cached === "number" && input < cached;
   return input + (splitSemantics && typeof cached === "number" ? cached : 0);
 }
-function backfillHostUsage(protocol, usage, credit) {
-  if (!Number.isFinite(credit) || credit <= 0) return false;
-  let patched = false;
-  const add = (key) => {
-    if (typeof usage[key] === "number" && Number.isFinite(usage[key])) {
-      usage[key] = usage[key] + credit;
-      patched = true;
-    }
-  };
-  if (protocol === "openai") {
-    add("prompt_tokens");
-    add("total_tokens");
-  } else {
-    add("input_tokens");
-  }
-  return patched;
-}
 var CONTEXT_OVERFLOW_PATTERNS = [
   /context_length_exceeded/i,
   /context_window_exceeded/i,
@@ -53486,7 +53463,6 @@ function recordUsage(ctx, usage, round) {
   const total = promptInputTotal(ctx.protocol, prompt, cached);
   if (total > 0) ctx.session.stats.inputTokens += total;
   ctx.session.stats.lastInputTokens = Math.max(0, total - (ctx.session.stats.compressCreditTokens ?? 0));
-  ctx.session.hostContextTokens = total + (ctx.session.hostCreditTokens ?? 0);
   if (typeof cached === "number") {
     ctx.session.stats.cachedTokens += cached;
     ctx.session.stats.cacheSamples += 1;
@@ -53627,10 +53603,6 @@ async function* runCompressLoop(upstream, ctx, requestBody, requestOptions, adap
       if (usage.inputTokens !== void 0 || usage.outputTokens !== void 0 || usage.cachedTokens !== void 0) {
         recordUsage(ctx, usage, round);
       }
-      const hostCredit = ctx.session.hostCreditTokens ?? 0;
-      if (hostCredit > 0 && typeof usage.inputTokens === "number") {
-        usage.inputTokens += hostCredit;
-      }
       let resolvedText = assistantText;
       let allCalls = calls;
       if (ctx.textProtocol && assistantText.length > 0 && adapter.extractTextTriggers) {
@@ -54527,24 +54499,7 @@ function stripFinishReasonChunk(buf) {
     return buf;
   }
 }
-function patchUsageChunk(eventStr, parsed, u2, hostCredit) {
-  const pu = typeof u2.prompt_tokens === "number" ? u2.prompt_tokens : void 0;
-  const tu = typeof u2.total_tokens === "number" ? u2.total_tokens : void 0;
-  if (hostCredit > 0 && (pu !== void 0 || tu !== void 0)) {
-    const patched = {
-      ...parsed,
-      usage: {
-        ...u2,
-        ...pu !== void 0 ? { prompt_tokens: pu + hostCredit } : {},
-        ...tu !== void 0 ? { total_tokens: tu + hostCredit } : {}
-      }
-    };
-    const out = eventStr.split("\n").map((l) => l.startsWith("data:") ? `data: ${JSON.stringify(patched)}` : l).join("\n");
-    return Buffer.from(out + "\n\n", "utf8");
-  }
-  return Buffer.from(eventStr + "\n\n", "utf8");
-}
-function createOpenaiAdapter(requestBody, clientSystem, hostCredit = 0, absorbName) {
+function createOpenaiAdapter(requestBody, clientSystem, absorbName) {
   const model = requestBody.model ?? "unknown";
   let responseId = `chatcmpl-proxy-${Date.now()}`;
   let toolIndex = 0;
@@ -54736,7 +54691,7 @@ function createOpenaiAdapter(requestBody, clientSystem, hostCredit = 0, absorbNa
               cachedTokens: typeof pd?.cached_tokens === "number" ? pd.cached_tokens : void 0
             };
             if (sawRealToolCall) {
-              yield { kind: "meta", chunk: patchUsageChunk(eventStr, parsed, u2, hostCredit) };
+              yield { kind: "meta", chunk: rawBuf };
             }
           }
           continue;
@@ -54756,7 +54711,7 @@ function createOpenaiAdapter(requestBody, clientSystem, hostCredit = 0, absorbNa
             cachedTokens: typeof pd?.cached_tokens === "number" ? pd.cached_tokens : void 0
           };
           if (sawRealToolCall) {
-            const chunk = patchUsageChunk(eventStr, parsed, u2 ?? {}, hostCredit);
+            const chunk = rawBuf;
             yield { kind: "meta", chunk };
             maybeWarnDegenerate(finishReason);
             yield { kind: "done", finishReason, suppressCompletion: true };
@@ -54947,7 +54902,7 @@ data: ${JSON.stringify({ type: "content_block_delta", index, delta: { type: "tex
     "utf8"
   );
 }
-function createAnthropicAdapter(requestBody, originalSystem, hostCredit = 0) {
+function createAnthropicAdapter(requestBody, originalSystem) {
   const model = requestBody.model ?? void 0;
   let messageId;
   let clientIndex = 0;
@@ -55072,19 +55027,8 @@ ${systemPrompt}` : systemPrompt;
           if (typeof u2.input_tokens === "number") roundInput = u2.input_tokens;
           if (typeof u2.cache_read_input_tokens === "number") roundCached = u2.cache_read_input_tokens;
           if (round === 1) {
-            let chunk = rawBuf;
-            if (hostCredit > 0 && typeof u2.input_tokens === "number") {
-              const patched = structuredClone(data);
-              const pmsg = patched["message"];
-              const pu = pmsg?.["usage"] ?? {};
-              if (typeof pu.input_tokens === "number") {
-                pu.input_tokens += hostCredit;
-                const out = eventStr.split("\n").map((l) => l.startsWith("data:") ? `data: ${JSON.stringify(patched)}` : l).join("\n");
-                chunk = Buffer.from(out + "\n\n", "utf8");
-              }
-            }
             messageStartForwarded = true;
-            yield { kind: "meta", chunk, firstRoundOnly: true };
+            yield { kind: "meta", chunk: rawBuf, firstRoundOnly: true };
           }
         } else if (type === "ping") {
           yield { kind: "meta", chunk: rawBuf };
@@ -55273,10 +55217,10 @@ data: ${JSON.stringify({ type: "content_block_stop", index })}
 }
 
 // src/loop/index.ts
-function pickAdapter(protocol, requestBody, textProtocol, responsesProjection, anthropicSystem, openaiSystem, hostCredit = 0, absorbName) {
+function pickAdapter(protocol, requestBody, textProtocol, responsesProjection, anthropicSystem, openaiSystem, absorbName) {
   if (protocol === "responses") return createResponsesAdapter(textProtocol, responsesProjection, absorbName);
-  if (protocol === "openai") return createOpenaiAdapter(requestBody, openaiSystem, hostCredit, absorbName);
-  if (protocol === "anthropic") return createAnthropicAdapter(requestBody, anthropicSystem, hostCredit);
+  if (protocol === "openai") return createOpenaiAdapter(requestBody, openaiSystem, absorbName);
+  if (protocol === "anthropic") return createAnthropicAdapter(requestBody, anthropicSystem);
   throw new Error(`[acp-loop] unknown protocol: ${protocol}`);
 }
 
@@ -56634,7 +56578,7 @@ function handlePluginStatus(conversationId2, res, deps, fallbackLatest = false)
     const systemPromptTokens = typeof sysTokRaw === "number" && Number.isFinite(sysTokRaw) && sysTokRaw > 0 ? sysTokRaw : 0;
     panel = buildStatusPanel({
       version: `billion-context@${PROXY_VERSION}`,
-      tokenCount: session.hostContextTokens ?? session.stats.lastInputTokens,
+      tokenCount: session.stats.lastInputTokens,
       systemPromptTokens,
       state: session.state,
       nudge,
@@ -56652,8 +56596,7 @@ function handlePluginStatus(conversationId2, res, deps, fallbackLatest = false)
     label: session.meta.label ?? null,
     pluginAgent: session.metadata.pluginAgent ?? null,
     contextLimit: typeof limit === "number" ? limit : null,
-    contextTokens: session.hostContextTokens ?? session.stats.lastInputTokens,
-    hostCredit: session.hostCreditTokens ?? 0,
+    contextTokens: session.stats.lastInputTokens,
     inputTokens: session.stats.inputTokens,
     outputTokens: session.stats.outputTokens,
     cachedTokens: session.stats.cachedTokens,
@@ -56775,11 +56718,10 @@ function applyUsageSample(session, sample, protocol) {
     session.stats.inputTokens += total;
     session.stats.lastInputTokens = Math.max(0, total - (session.stats.compressCreditTokens ?? 0));
     warnCacheCollapse(session, total, sample.cachedTokens ?? 0);
-    session.hostContextTokens = total + (session.hostCreditTokens ?? 0);
     const hit = sample.cachedTokens === void 0 || total <= 0 ? void 0 : Math.round(100 * (sample.cachedTokens ?? 0) / total);
     const foldNew = session.stats.pendingFoldUsage === true;
     if (foldNew) session.stats.pendingFoldUsage = false;
-    log("info", `[${session.id}] [plugin] [acp-usage] input=${total} cached=${sample.cachedTokens ?? "n/a"}${hit === void 0 ? "" : ` (cache hit ${hit}%)`} ctx=${session.hostContextTokens}${foldNew ? " fold=new" : ""}`);
+    log("info", `[${session.id}] [plugin] [acp-usage] input=${total} cached=${sample.cachedTokens ?? "n/a"}${hit === void 0 ? "" : ` (cache hit ${hit}%)`}${foldNew ? " fold=new" : ""}`);
   }
   if (sample.outputTokens !== void 0) session.stats.outputTokens += sample.outputTokens;
 }
@@ -56793,7 +56735,6 @@ async function pipePluginChatWithStrip(stream2, res, protocol, session, log2) {
   const decoder = new TextDecoder("utf-8");
   let buf = "";
   const acc = {};
-  const credit = session?.hostCreditTokens ?? 0;
   const onDrop = (snippet) => {
     log("warn", `[tag-echo] stripped model-emitted render tag (plugin passthrough): ${snippet.slice(0, 80).replace(/\n/g, " ")}`);
     log2?.(`[tag-echo] stripped model-emitted render tag from plugin passthrough text`);
@@ -57004,18 +56945,8 @@ data: ${JSON.stringify({ type: "content_block_delta", index, delta: { type: delt
           }
           const sample = usageFromSseEvent(ev);
           if (sample) mergeUsageSample(acc, sample);
-          let backfilled = false;
-          if (credit > 0 && protocol) {
-            const usage = ev["type"] === "message_start" ? ev["message"]?.["usage"] : ev["usage"];
-            const deltaEcho = ev["type"] === "message_delta" && (num2(usage?.["input_tokens"]) ?? 0) <= 0;
-            if (usage && !deltaEcho && backfillHostUsage(protocol, usage, credit)) backfilled = true;
-          }
           const out = protocol === "anthropic" ? processAnthropic(ev, rawEvent) : processOpenai(ev, rawEvent);
-          if (backfilled && out === rawEvent + "\n\n") {
-            await write(rebuildEvent(rawEvent, ev));
-          } else if (out.length > 0) {
-            await write(out);
-          }
+          if (out.length > 0) await write(out);
         }
       }
       if (res.destroyed || res.writableEnded) break;
@@ -57155,13 +57086,6 @@ async function pipePluginResponsesWithStrip(stream2, res, session, log2) {
             let evOut = ev;
             let rebuild = containsRenderTagText(jsonStr);
             if (rebuild) evOut = stripResponsesText(ev);
-            if (type === "response.completed") {
-              const credit = session?.hostCreditTokens ?? 0;
-              const usage = evOut["response"]?.["usage"];
-              if (credit > 0 && usage && backfillHostUsage("responses", usage, credit)) {
-                rebuild = true;
-              }
-            }
             const out = rebuild ? rebuildEvent(rawEvent, evOut) : rawEvent + "\n\n";
             await write(flushTail(out));
             continue;
@@ -57258,10 +57182,6 @@ async function pipePluginJson(stream2, res, session, protocol) {
           cachedTokens: num2(usage["prompt_tokens_details"]?.["cached_tokens"]) ?? num2(usage["input_tokens_details"]?.["cached_tokens"]) ?? num2(usage["cache_read_input_tokens"])
         }, protocol);
         markDirty(session);
-        const credit = session.hostCreditTokens ?? 0;
-        if (credit > 0 && protocol && backfillHostUsage(protocol, usage, credit)) {
-          mutated = true;
-        }
       }
     }
   } catch {
@@ -60068,16 +59988,6 @@ 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, headers, hostUsageCredit, log2) {
-  session.hostCreditTokens = 0;
-  if (hostUsageCredit === "off") return;
-  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`);
-  }
-}
 function effectiveTokenCount(session, msgs) {
   if (session.stats.lastInputTokens > 0) return session.stats.lastInputTokens;
   if (!session.metadata.anonymousPrefixAffinity) return 0;
@@ -60087,7 +59997,6 @@ function prepareAnthropic(parsed, req, opts, core, config, prompts, log2, sessio
   const sessionId = session.id;
   const stream2 = parsed.stream === true;
   ++session.stats.requests;
-  session.hostCreditTokens = 0;
   const injectTools = opts.compress.injectTool && !pluginMode;
   const stripReasoning = (msgs) => withReasoningDrop(msgs, reasoning, log2, sessionId, isStrictReasoningEcho(session, upstreamOrigin));
   if (isAutoModeClassifier(parsed)) {
@@ -60154,7 +60063,6 @@ function prepareAnthropic(parsed, req, opts, core, config, prompts, log2, sessio
   const rebuilt = { ...parsed, messages: rebuiltMessages, system: systemOut, tools: toolsOut };
   warnAnthropicThinkingPairs(rebuiltMessages, log2, sessionId);
   delete rebuilt.prompt_cache_key;
-  armHostUsageCredit(session, originalMessages, processedMessages, req.headers, opts.hostUsageCredit, 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;
@@ -60214,7 +60122,6 @@ function prepareOpenai(parsed, req, opts, core, config, prompts, log2, session,
   const sessionId = session.id;
   const stream2 = parsed.stream === true;
   ++session.stats.requests;
-  session.hostCreditTokens = 0;
   let openaiSystemText = "";
   const stripReasoning = (msgs) => withReasoningDrop(msgs, reasoning, log2, sessionId, isStrictReasoningEcho(session, upstreamOrigin));
   let openaiOutboundSystem;
@@ -60287,7 +60194,6 @@ 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, req.headers, opts.hostUsageCredit, log2);
   if (!isTitleGen && openaiOutboundSystem !== void 0) {
     session.metadata.systemPromptTokens = countSystemAndToolsTokens(openaiOutboundSystem, toolsOut);
   }
@@ -60299,7 +60205,6 @@ function prepareResponses(parsed, req, opts, core, config, prompts, log2, sessio
   const sessionId = session.id;
   const stream2 = parsed.stream === true;
   ++session.stats.requests;
-  session.hostCreditTokens = 0;
   const stripReasoning = (msgs) => withReasoningDrop(msgs, reasoning, log2, sessionId, isStrictReasoningEcho(session, upstreamOrigin));
   if (reconcileNativeCompactionBoundary(session)) {
     log2("info", `[${sessionId}] reconciled ACP state after native Responses compact boundary`);
@@ -60440,7 +60345,6 @@ function prepareResponses(parsed, req, opts, core, config, prompts, log2, sessio
     });
     log2("info", `[${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, req.headers, opts.hostUsageCredit, log2);
   if (transformOk) {
     session.metadata.systemPromptTokens = countSystemAndToolsTokens(responsesDevContent ?? "", toolsOut);
   }
@@ -61218,7 +61122,7 @@ ${hdrText}
 
 ${buildAbsorbSystemPrompt(absorbToolName(loopConfig))}` : "";
         const systemPrompt = (textProtocol ? buildCompressHybridSystemPrompt(prepared.prompts ?? defaultPrompts) : buildCompressSystemPrompt(prepared.prompts ?? defaultPrompts)) + absorbSection;
-        const adapter = pickAdapter(prepared.protocol, parsedReq, textProtocol, prepared.responsesProjection, prepared.anthropicSystem, prepared.openaiSystemText, prepared.session.hostCreditTokens ?? 0, absorbActive ? absorbToolName(loopConfig) : void 0);
+        const adapter = pickAdapter(prepared.protocol, parsedReq, textProtocol, prepared.responsesProjection, prepared.anthropicSystem, prepared.openaiSystemText, absorbActive ? absorbToolName(loopConfig) : void 0);
         const refreshFolded = (current) => {
           const turn = core.processTurn({
             messages: prepared.originalMessages,
@@ -61301,10 +61205,6 @@ ${buildAbsorbSystemPrompt(absorbToolName(loopConfig))}` : "";
             const out = u2.completion_tokens ?? u2.output_tokens;
             if (typeof out === "number") prepared.session.stats.outputTokens += out;
           }
-          const credit = prepared.session.hostCreditTokens ?? 0;
-          if (credit > 0 && backfillHostUsage(prepared.protocol, u2, credit)) {
-            prepared.session.hostContextTokens = (typeof total === "number" ? total : 0) + credit;
-          }
           if (prepared.protocol === "openai") {
             rewriteOpenaiJsonResponse(json, ctx);
           } else if (prepared.protocol === "responses") {