opencode-acp 1.14.16 → 1.14.17-pr.304.12

package/README.md

@@ -492,6 +492,16 @@ For the complete list with root cause analysis, see the [bug tracker](https://gi
 
 ## Changelog
 
+### v1.14.17 — Eliminate compress retry dead-ends (#301)
+
+**Problem**: Two compress arguments hard-failed on first use and produced confusing error loops instead of actionable feedback. (1) The model learned "add `acknowledgeRisk: true` to retry" from quality-gate rejection templates, then carried the flag into *every* retry — including retries after non-quality errors (e.g. a missing topic), where no quality-gate rejection was pending, hitting `Parameter "acknowledgeRisk": true was provided, but no quality gate rejection is pending` (issue #301). (2) `content[0] needs a topic — provide content[0].topic or the top-level topic` hard-failed compress calls whose entries had no topic and no top-level fallback.
+
+**Fix**: Both parameters are now tolerant. `acknowledgeRisk` without a pending rejection is a no-op — quality checks still run (the gate's protection is unchanged: bypass is only armed by a real quality-gate rejection), and a successful result carries an `⚠️ acknowledgeRisk was ignored: ...` note teaching correct usage. Topics are fully optional: an entry without its own topic falls back to the top-level topic, then to one derived from the summary's first line (markdown headings stripped, capped at 80 chars), so `search_context` always has something searchable.
+
+Files: `lib/compress/range.ts`, `lib/compress/range-utils.ts` (new `deriveFallbackTopic`), `lib/compress/quality-gate/rejection.ts` (removed `buildPreemptiveAcknowledgeError`), `lib/compress/types.ts`, `lib/prompts/compress-range.ts`, `lib/prompts/extensions/tool.ts`, `lib/state/types.ts`. Tests: 976 pass.
+
+**Install**: `opencode plugin opencode-acp@latest --global`
+
 ### v1.14.16 — Raise context limits to 80%
 
 **Problem**: The compression thresholds were `maxContextLimit: 55%` / `minContextLimit: 45%`. The strong "over-limit" nudge fired as soon as context exceeded 55% of the model window. Combined with the fixed 50K growth threshold (v1.14.15), this still engaged compression relatively early, leaving usable context unused.

package/README.zh-CN.md

@@ -446,6 +446,16 @@ ACP 是 DCP 的直接替代品。迁移步骤：
 
 ## 更新日志
 
+### v1.14.17 — 消除 compress 重试死循环（#301）
+
+**问题**：两个 compress 参数在首次使用时硬报错，产生令人困惑的错误循环而非可执行的反馈。(1) 模型从质量门拒绝模板学到「重试要加 `acknowledgeRisk: true`」，于是把它带进了所有重试——包括非质量错误（如缺 topic）之后的重试，此时没有 pending 的质量门拒绝，撞上 `Parameter "acknowledgeRisk": true was provided, but no quality gate rejection is pending`（issue #301）。(2) `content[0] needs a topic — provide content[0].topic or the top-level topic` 在 entry 无 topic 且无顶层 fallback 时硬报错。
+
+**修复**：两个参数均改为宽容处理。无 pending 拒绝时的 `acknowledgeRisk` 为 no-op——质量检查照常运行（质量门保护不变：bypass 只由真实的质量门拒绝解锁），成功结果附带 `⚠️ acknowledgeRisk was ignored: ...` 提示教模型正确用法。topic 完全可选：entry 无自有 topic 时依次回退到顶层 topic、再到从 summary 首行派生的 topic（去 markdown 标题符、截断 80 字符），保证 `search_context` 始终有可检索内容。
+
+文件：`lib/compress/range.ts`、`lib/compress/range-utils.ts`（新增 `deriveFallbackTopic`）、`lib/compress/quality-gate/rejection.ts`（移除 `buildPreemptiveAcknowledgeError`）、`lib/compress/types.ts`、`lib/prompts/compress-range.ts`、`lib/prompts/extensions/tool.ts`、`lib/state/types.ts`。测试：976 通过。
+
+**安装**：`opencode plugin opencode-acp@latest --global`
+
  ### v1.14.16 — 将上下文 limit 提升至 80%
 
 **问题**：压缩阈值原为 `maxContextLimit: 55%` / `minContextLimit: 45%`。上下文一旦超过模型窗口的 55% 就会触发强「超限」nudge。配合 v1.14.15 固定的 50K 增长阈值，压缩仍较早介入，可用上下文未被充分利用。

package/dist/index.js

@@ -2075,7 +2075,7 @@ THE FORMAT OF COMPRESS
   ]
 }
 \`\`\`
-Each entry needs a topic \u2014 either its own or the top-level fallback.`;
+Each entry MAY have a topic. Missing topics fall back to the top-level topic, then to an automatic one derived from the summary.`;
 
 // lib/state/persistence.ts
 import * as fs from "fs/promises";
@@ -3756,8 +3756,11 @@ function filterProtectedRecentMessages(selection, searchContext, state, compress
 
 // lib/compress/range-utils.ts
 var BLOCK_PLACEHOLDER_REGEX = /\(b(\d+)\)|\{block_(\d+)\}/gi;
+function deriveFallbackTopic(summary) {
+  const firstLine = summary.split("\n").map((line) => line.replace(/^#{1,6}\s*/, "").trim()).find((line) => line.length > 0) ?? "compressed range";
+  return firstLine.length > 80 ? `${firstLine.slice(0, 77)}...` : firstLine;
+}
 function validateArgs(args) {
-  const hasTopLevelTopic = typeof args.topic === "string" && args.topic.trim().length > 0;
   if (!Array.isArray(args.content) || args.content.length === 0) {
     throw new Error("content is required and must be a non-empty array");
   }
@@ -3773,18 +3776,12 @@ function validateArgs(args) {
     if (typeof entry?.summary !== "string" || entry.summary.trim().length === 0) {
       throw new Error(`${prefix}.summary is required and must be a non-empty string`);
     }
-    const hasEntryTopic = typeof entry?.topic === "string" && entry.topic.trim().length > 0;
-    if (!hasEntryTopic && !hasTopLevelTopic) {
-      throw new Error(
-        `${prefix} needs a topic \u2014 provide ${prefix}.topic or the top-level topic`
-      );
-    }
   }
 }
 function resolveRanges(args, searchContext, state, logger) {
   return args.content.map((entry, index) => {
     const normalizedEntry = {
-      topic: typeof entry.topic === "string" && entry.topic.trim().length > 0 ? entry.topic.trim() : void 0,
+      topic: typeof entry.topic === "string" && entry.topic.trim().length > 0 ? entry.topic.trim() : typeof args.topic === "string" && args.topic.trim().length > 0 ? void 0 : deriveFallbackTopic(entry.summary),
       startId: entry.startId.trim(),
       endId: entry.endId.trim(),
       summary: entry.summary
@@ -5672,11 +5669,6 @@ exact values, errors). Then add "acknowledgeRisk": true to the compress tool cal
 Without acknowledgeRisk: true, the compression will be rejected again.`;
   return new Error(message);
 }
-function buildPreemptiveAcknowledgeError() {
-  return new Error(
-    'Parameter "acknowledgeRisk": true was provided, but no quality gate rejection is pending. This parameter is only valid immediately after a compression was rejected by the quality gate. Remove it and try again.'
-  );
-}
 
 // lib/compress/pipeline.ts
 function snapshotCompressionState(state) {
@@ -6150,13 +6142,13 @@ function createCompressRangeTool(factoryCtx) {
       }
       const acknowledgeRisk = args.acknowledgeRisk === true;
       const qualityGateRetryPendingBefore = ctx.state.qualityGateRetryPending;
-      if (acknowledgeRisk && !ctx.state.qualityGateRetryPending) {
-        throw buildPreemptiveAcknowledgeError();
+      const bypassQuality = acknowledgeRisk && ctx.state.qualityGateRetryPending;
+      const ignoredAcknowledgeRisk = acknowledgeRisk && !ctx.state.qualityGateRetryPending;
+      if (ignoredAcknowledgeRisk) {
+        ctx.logger.warn("compress: acknowledgeRisk ignored \u2014 no quality gate rejection pending");
       }
-      if (acknowledgeRisk) {
-        ctx.state.qualityGateRetryPending = false;
-      } else {
-        ctx.state.qualityGateRetryPending = false;
+      ctx.state.qualityGateRetryPending = false;
+      if (!bypassQuality) {
         for (const plan of preparedPlans) {
           const result = evaluatePreCommitQuality(
             rawMessages,
@@ -6232,7 +6224,10 @@ function createCompressRangeTool(factoryCtx) {
       const skippedNote = phantomSkipNotice !== null ? `
 \u26A0\uFE0F ${phantomSkipNotice}
 ` : "";
-      return `Compressed ${totalCompressedMessages} messages into ${COMPRESSED_BLOCK_HEADER}.${skippedNote}
+      const ackNote = ignoredAcknowledgeRisk ? `
+\u26A0\uFE0F acknowledgeRisk was ignored: no quality gate rejection was pending, so quality checks ran normally. Only pass it when retrying immediately after a quality gate rejection.
+` : "";
+      return `Compressed ${totalCompressedMessages} messages into ${COMPRESSED_BLOCK_HEADER}.${skippedNote}${ackNote}
 IMPORTANT: This was an automatic context compression. You MUST continue your previous task exactly where you left off. Do NOT ask the user what to do next.
 \u{1F4A1} Tip: Use search_context('keyword') to find compressed content when you need it later.`;
     }
@@ -8949,7 +8944,7 @@ import { writeFile as writeFile2, mkdir as mkdir2 } from "fs/promises";
 import { join as join3 } from "path";
 import { existsSync as existsSync3 } from "fs";
 import { homedir as homedir3 } from "os";
-var LOG_VERSION = true ? "1.14.16" : "dev";
+var LOG_VERSION = true ? "1.14.17-pr.304.12" : "dev";
 var Logger = class {
   logDir;
   enabled;
@@ -9277,7 +9272,7 @@ Rules:
 BATCHING
 When multiple independent ranges are ready and their boundaries do not overlap, include all of them as separate entries in the \`content\` array of a single tool call. Each entry should have its own \`startId\`, \`endId\`, and \`summary\`.
 
-When the ranges cover unrelated topics, give each entry its own \`topic\` for better summary quality \u2014 do not force unrelated content under a single shared topic. Omit the top-level \`topic\` when every entry has its own. Use the top-level \`topic\` only as a fallback when entries don't specify one.
+When the ranges cover unrelated topics, give each entry its own \`topic\` for better summary quality \u2014 do not force unrelated content under a single shared topic. The top-level \`topic\` is an optional fallback; when no topic is provided at all, one is derived from the summary's first line automatically.
 
 \`\`\`
 compress({ content: [