opencode-acp 1.14.6 → 1.14.7

package/README.md

@@ -509,6 +509,14 @@ For the complete list with root cause analysis, see the [bug tracker](https://gi
 
 ## Changelog
 
+### v1.14.7 — Deduplicate HOW_TO_COMPRESS_RULES (PR #228)
+
+**Problem**: `HOW_TO_COMPRESS_RULES` (~1.2K tokens) was injected 3-4 times per nudge turn — once in the system prompt, 1-2 times in nudge templates (turn/iteration/context-limit), and again in the breakdown block. In non-maxLimit scenarios the suffix message alone contained the rules 2-3 times. This wasted 2.4-3.6K tokens per nudge turn. The duplication became visible in v1.14.6 which persists debug nudge text to chat UI.
+
+**Fix**: Removed `HOW_TO_COMPRESS_RULES` from the breakdown block (`inject.ts:535-537`) and all 3 nudge templates (`turn-nudge.ts`, `iteration-nudge.ts`, `context-limit-nudge.ts`). Kept in `system.ts` (single source of truth — injected every turn per v1.8.2 invariant) and `quality-gate/rejection.ts` (retry guidance). Oracle-verified: no scenario loses compression guidance.
+
+Files: `lib/messages/inject/inject.ts`, `lib/prompts/{turn,iteration,context-limit}-nudge.ts`. 917 tests pass.
+
 ### v1.14.6 — Debug Nudge Chat Visibility (PR #226)
 
 **Problem**: With `debug: true`, ACP nudge injections (compression suggestions, context breakdown, tier triggers) were only visible via a 5-second toast and log files. The nudge suffix message is ephemeral — injected by the message-transform hook but never persisted to the conversation DB — so users could not see what the model was actually seeing in the chat UI. This made debugging nudge behavior very difficult.

package/README.zh-CN.md

@@ -463,6 +463,14 @@ ACP 在首次启动时自动将配置从 `dcp.jsonc` 迁移到 `acp.jsonc`，将
 
 ## 更新日志
 
+### v1.14.7 — 去重 HOW_TO_COMPRESS_RULES（PR #228）
+
+**问题**：`HOW_TO_COMPRESS_RULES`（~1.2K tokens）每次 nudge 注入重复 3-4 次——系统提示词 1 次、nudge 模板 1-2 次、breakdown 块 1 次。非 maxLimit 场景下 suffix 消息单独就包含 2-3 份规则。每次 nudge 浪费 2.4-3.6K tokens。v1.14.6 让 debug 模式持久化 nudge 文本到聊天界面后，重复变得可见。
+
+**修复**：从 breakdown 块（`inject.ts:535-537`）和全部 3 个 nudge 模板（`turn-nudge.ts`、`iteration-nudge.ts`、`context-limit-nudge.ts`）移除 `HOW_TO_COMPRESS_RULES`。保留在 `system.ts`（单一可信源——每轮注入，v1.8.2 不变量）和 `quality-gate/rejection.ts`（重试指导）。Oracle 验证：无任何场景丢失压缩指导。
+
+文件：`lib/messages/inject/inject.ts`、`lib/prompts/{turn,iteration,context-limit}-nudge.ts`。917 项测试通过。
+
 ### v1.14.6 — Debug Nudge 聊天界面可见性（PR #226）
 
 **问题**：开启 `debug: true` 时，ACP nudge 注入（压缩建议、上下文分类、分层触发器）只能通过 5 秒 toast 和日志文件查看。nudge 后缀消息是临时的——由消息变换 hook 注入但从未持久化到会话数据库——因此用户无法在聊天界面看到模型实际看到的内容。这使得调试 nudge 行为非常困难。

package/dist/index.js

@@ -8331,11 +8331,6 @@ ${formatCompressibleRanges(recommendedRanges, contextRanges.protected)}`;
       }
       breakdown += `
 Use \`acp_status({scope:"uncompressed"})\` to re-fetch compressible ranges after compressing, or \`acp_status\` for compressed block details.`;
-      if (effectiveTipsVariant !== "maxLimit") {
-        breakdown += `
-
-${HOW_TO_COMPRESS_RULES}`;
-      }
       appendToLastTextPart(suffixMessage, breakdown);
     }
     if (effectiveTipsVariant === "maxLimit") {
@@ -9543,7 +9538,7 @@ 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 ? "1.14.6" : "dev";
+var LOG_VERSION = true ? "1.14.7" : "dev";
 var Logger = class {
   logDir;
   enabled;
@@ -9968,8 +9963,6 @@ HOW TO CALL COMPRESS:
 - Do NOT use IDs from compressed block summaries \u2014 they are stale.
 - startId must appear BEFORE endId in the conversation.
 
-${HOW_TO_COMPRESS_RULES}
-
 RANGE STRATEGY:
 - Prefer one large range over multiple small ones.
 - Compress OLDER resolved history first. Keep recent active work.
@@ -9987,8 +9980,6 @@ Context is getting full. If you've finished reading tool outputs or exploration
 }
 
 \u26A0\uFE0F ONLY use IDs from  tags visible above. Do NOT invent or copy example IDs.
-
-${HOW_TO_COMPRESS_RULES}
 </system-reminder>
 `;
 
@@ -10002,9 +9993,7 @@ You've been iterating for a while. If any earlier work is closed and unlikely to
   "content": [{ "startId": "<visible message ID>", "endId": "<visible message ID>", "summary": "..." }]
 }
 
-\u26A0\uFE0F ONLY use IDs from <dcp-message-id> tags visible above. Do NOT invent or copy example IDs.
-
-${HOW_TO_COMPRESS_RULES}
+\u26A0\uFE0F ONLY use IDs from  tags visible above. Do NOT invent or copy example IDs.
 </system-reminder>
 `;