billion-context-omp 0.2.2 → 0.2.4

package/README.md

@@ -77,7 +77,7 @@ omp's built-in `/compact` is intercepted and replaced by an ACP model-summarized
 |------|-------------|
 | `compress` | Replace a contiguous message range with a detailed summary |
 | `decompress` | Restore a previously compressed block's content (to file by default; `inline:true` for single messages) |
-| `search_context` | Search compressed block summaries (and visible messages) by keyword |
+| `search_context` | Search compressed block summaries and the original messages folded into them by keyword (visible messages are not indexed) |
 | `acp_status` | Show context usage, compressed blocks, compressible ranges |
 
 > The `acp_delegate` sub-agent subsystem from the Pi build is intentionally **not** registered — omp ships its own multi-agent orchestration, and duplicate delegation tools would conflict.

package/README.zh-CN.md

@@ -77,7 +77,7 @@ omp 内置的 `/compact` 被拦截，替换为 ACP 模型摘要式 compaction，
 |------|------|
 | `compress` | 把一段连续消息区间替换为详细摘要 |
 | `decompress` | 恢复之前压缩的块内容（默认写文件；单条消息可 `inline:true`） |
-| `search_context` | 按关键字搜索压缩块摘要（及可见消息） |
+| `search_context` | 按关键字搜索压缩块摘要及被折叠其中的原始消息（可见消息不建索引） |
 | `acp_status` | 显示上下文用量、压缩块、可压缩区间 |
 
 > Pi 版中的 `acp_delegate` 子代理系统在此**有意不注册**——omp 自带多代理编排，重复的委派工具会冲突。

package/dist/auto-compress.d.ts

@@ -43,8 +43,11 @@ export declare function buildSummaryPrompt(prompts: Prompts): string;
  *  deliberately NOT used here: fold blocks only replay from in-stream
  *  compress tool calls, which the truncation removes. `previousSummary` (an
  *  earlier compaction's summary) is folded in so iterative compactions never
- *  drop it. Returns null on any failure so the caller falls back to Pi's
- *  native compaction. */
+ *  drop it. On a failed attempt the LLM call is
+ *  retried ONCE (stochastic formatting failures recover on a fresh call);
+ *  returns null only when the retry also fails — the caller then CANCELS the
+ *  compaction (no native fallback: ACP owns compression, host-default
+ *  summaries are the failure mode this hook exists to prevent). */
 export declare function summarizeMessages(ctx: ExtensionContext, messages: AgentMessage[], prompts: Prompts, configuredModel?: string | null, opts?: {
     previousSummary?: string;
     customInstructions?: string;
@@ -58,11 +61,9 @@ export declare function summarizeMessages(ctx: ExtensionContext, messages: Agent
     summary: string;
     model: string;
 } | null>;
-/** Generate a summary for a message range using the compression model. Shared
- *  entry point for the `/compact` handler. Returns null when no model is
- *  usable, the slice is empty, the model is unauthenticated, or the response
- *  is unparseable — the caller then returns `undefined` so Pi falls back to
- *  its native compaction. */
+/** Generate a summary for a message range using the compression model.
+ *  Currently unused by the extension (kept as the shared range-summary
+ *  surface); null = hard failure, caller decides. */
 export declare function summarizeRange(ctx: ExtensionContext, messages: CoreMessage[], state: CompressionState, startRef: string, endRef: string, prompts: Prompts, configuredModel?: string | null): Promise<{
     summary: string;
     model: string;

package/dist/decompress-tool.d.ts

@@ -12,4 +12,5 @@ declare const DecompressParams: import("@oh-my-pi/omptype").FluentType<{
     toFile?: string | undefined;
 }>;
 export declare function makeDecompressTool(runtime: AcpRuntime): ToolDefinition<typeof DecompressParams>;
+export declare function pruneAutoFiles(): void;
 export {};

package/dist/index.js

@@ -2769,8 +2769,12 @@ function resolveConfig(adapter, liveContextLimit) {
   return config;
 }
 function parsePercent(v) {
-  const n = typeof v === "number" ? v : v.trim().endsWith("%") ? Number(v.trim().slice(0, -1)) / 100 : Number(v);
-  if (!Number.isFinite(n)) return 0;
+  const raw = typeof v === "number" ? v : v.trim().endsWith("%") ? Number(v.trim().slice(0, -1)) : Number(v);
+  if (!Number.isFinite(raw)) return 0;
+  const n = raw > 1 ? raw / 100 : raw;
+  if (n > 1 || n < 0) {
+    logWarn("config", { event: "percent-clamped", value: String(v), clamped: Math.min(1, Math.max(0, n)) });
+  }
   return Math.min(1, Math.max(0, n));
 }
 
@@ -3159,7 +3163,7 @@ function rangeFingerprints(ranges, coreMessages, byRef, blocks) {
 
 // src/runtime.ts
 function freshSlot(preserveFrom) {
-  const slot = { identities: [], foldedLen: 0, preview: false, state: createInitialState(), coreMessages: [], appliedCallIds: /* @__PURE__ */ new Set(), rejectStreak: 0 };
+  const slot = { identities: [], foldedLen: 0, preview: false, state: createInitialState(), coreMessages: [], appliedCallIds: /* @__PURE__ */ new Set(), rejectStreak: 0, lastRebuiltOutput: null };
   if (preserveFrom) {
     slot.state = { ...slot.state, nudge: preserveFrom.state.nudge };
     slot.rejectStreak = preserveFrom.rejectStreak;
@@ -3223,6 +3227,17 @@ function createRuntime(adapter) {
       slots.set(sid, slot);
     }
     const ids = stream.map(messageIdentity);
+    const lastOut = slot.lastRebuiltOutput;
+    if (lastOut !== null && ids.length === lastOut.length && ids.every((id, i) => id === lastOut[i])) {
+      const coreMessages2 = streamToCoreMessages(stream);
+      const originalById2 = /* @__PURE__ */ new Map();
+      stream.forEach((message, i) => originalById2.set(`p${i + 1}`, message));
+      slot.identities = ids;
+      slot.foldedLen = ids.length;
+      slot.coreMessages = coreMessages2;
+      debug.event("feedback-reuse", { sid, msgs: ids.length, blocks: slot.state.blocks.length });
+      return { state: slot.state, coreMessages: coreMessages2, originalById: originalById2, streamLen: ids.length };
+    }
     let lcp = 0;
     while (lcp < Math.min(ids.length, slot.identities.length) && ids[lcp] === slot.identities[lcp]) lcp++;
     if (lcp < slot.foldedLen) {
@@ -3311,6 +3326,10 @@ function createRuntime(adapter) {
     slot.state = state;
     if (toolCallId) slot.appliedCallIds.add(toolCallId);
   }
+  function recordRebuiltOutput(ctx, rebuilt) {
+    const slot = slotFor(sidOf(ctx));
+    slot.lastRebuiltOutput = rebuilt.map(messageIdentity);
+  }
   function noteCompressOutcome(ctx, ok) {
     const slot = slotFor(sidOf(ctx));
     slot.rejectStreak = ok ? 0 : slot.rejectStreak + 1;
@@ -3335,6 +3354,7 @@ function createRuntime(adapter) {
     foldStream,
     stateFor,
     commitFoldState,
+    recordRebuiltOutput,
     noteCompressOutcome,
     forgetSession,
     primeFold,
@@ -3435,11 +3455,13 @@ function makeCompressTool(runtime) {
 }
 async function handleCompress(args, runtime, ctx, toolCallId) {
   const ranges = args.content ?? [];
-  if (ranges.length === 0) return "No ranges provided.";
+  if (ranges.length === 0) {
+    return rejectionMessage(ctx, runtime, "No ranges provided. No changes applied \u2014 send content: [{ startId, endId, summary }].");
+  }
   const invalid = ranges.filter((r) => !r || typeof r.summary !== "string" || !r.summary.trim() || !r.startId || !r.endId);
   if (invalid.length > 0) {
     logWarn("compress", { sid: ctx.sessionManager.getSessionId(), event: "invalid-ranges", count: invalid.length });
-    return `Every range needs startId, endId and a summary (min 50 chars) \u2014 got ${invalid.length} range(s) missing fields. Re-send the FULL ranges array with summaries included.`;
+    return rejectionMessage(ctx, runtime, `Every range needs startId, endId and a summary (min 50 chars) \u2014 got ${invalid.length} range(s) missing fields. Re-send the FULL ranges array with summaries included. No changes applied \u2014 run acp_status for current refs.`);
   }
   const releaseLock = await runtime.acquireLock(ctx.sessionManager.getSessionId());
   try {
@@ -3550,10 +3572,12 @@ STOP: ${streak} compress calls rejected in a row. Do NOT retry the same range. R
 // src/decompress-tool.ts
 import { type as type2 } from "@oh-my-pi/omptype";
 import { writeFile, mkdir } from "fs/promises";
-import { realpathSync } from "fs";
+import { realpathSync, readdirSync, statSync as statSync2, unlinkSync } from "fs";
 import { resolve, relative, isAbsolute, join as join2, dirname as dirname2, basename } from "path";
 import { tmpdir } from "os";
-var AUTO_DIR = join2(homeDir() || tmpdir(), ".cache", "omp", "acp-decompress");
+function autoDir() {
+  return join2(homeDir() || tmpdir(), ".cache", "omp", "acp-decompress");
+}
 var PREVIEW_CHARS = 600;
 var MESSAGE_INLINE_THRESHOLD = 2e3;
 var DecompressParams = type2({
@@ -3618,7 +3642,31 @@ function resolveToFilePath(targetPath) {
   return resolved;
 }
 function autoFilePath(blockId) {
-  return join2(AUTO_DIR, `${blockId}-${Date.now()}.txt`);
+  return join2(autoDir(), `${blockId}-${Date.now()}.txt`);
+}
+var MAX_AUTO_FILES = 100;
+function pruneAutoFiles() {
+  const dir = autoDir();
+  try {
+    const files = readdirSync(dir).map((f) => {
+      try {
+        return { f, m: statSync2(join2(dir, f)).mtimeMs };
+      } catch {
+        return null;
+      }
+    }).filter((x) => x !== null);
+    if (files.length <= MAX_AUTO_FILES) return;
+    files.sort((a, b) => a.m - b.m);
+    const excess = files.slice(0, files.length - MAX_AUTO_FILES);
+    for (const { f } of excess) {
+      try {
+        unlinkSync(join2(dir, f));
+      } catch {
+      }
+    }
+    debug.event("decompress-pruned", { dir, removed: excess.length });
+  } catch {
+  }
 }
 function headPreview(text) {
   if (text.length <= PREVIEW_CHARS) return text;
@@ -3652,8 +3700,9 @@ ${text}`;
     logError("decompress", { sid: ctx.sessionManager.getSessionId(), event: "message-path-rejected", ref, toFile: args.toFile });
     return targetPath.error;
   }
-  await mkdir(AUTO_DIR, { recursive: true }).catch((e) => logError("decompress", { event: "mkdir-failed", dir: AUTO_DIR, error: e instanceof Error ? e.message : String(e) }));
+  await mkdir(autoDir(), { recursive: true }).catch((e) => logError("decompress", { event: "mkdir-failed", dir: autoDir(), error: e instanceof Error ? e.message : String(e) }));
   await writeFile(targetPath, text, "utf8");
+  if (!args.toFile) pruneAutoFiles();
   debug.event("decompress-message", { ref, ownerBlockId, mode: "file", path: targetPath, chars: text.length });
   logInfo("decompress", { sid: ctx.sessionManager.getSessionId(), event: "message", mode: "file", ref, ownerBlockId, path: targetPath, chars: text.length });
   return [
@@ -3695,8 +3744,9 @@ ${text}`;
     logError("decompress", { sid: ctx.sessionManager.getSessionId(), event: "block-path-rejected", blockId, toFile: args.toFile });
     return targetPath.error;
   }
-  await mkdir(AUTO_DIR, { recursive: true }).catch((e) => logError("decompress", { event: "mkdir-failed", dir: AUTO_DIR, error: e instanceof Error ? e.message : String(e) }));
+  await mkdir(autoDir(), { recursive: true }).catch((e) => logError("decompress", { event: "mkdir-failed", dir: autoDir(), error: e instanceof Error ? e.message : String(e) }));
   await writeFile(targetPath, text, "utf8");
+  if (!args.toFile) pruneAutoFiles();
   debug.event("decompress", { blockId, full, count, mode: "file", path: targetPath, chars: text.length });
   logInfo("decompress", { sid: ctx.sessionManager.getSessionId(), event: "block", mode: "file", blockId, full, count, path: targetPath, chars: text.length });
   const itemWord = count === 1 ? "item" : "items";
@@ -4508,7 +4558,7 @@ async function statusReport(runtime, ctx) {
   const coveredIds = collectCoveredMessageIds(state);
   const sentTokens = estimateTokens(coreMessages, coveredIds) + systemPromptTokens;
   const turn = runtime.core.processTurn({ messages: coreMessages, state, config, tokenCount: sentTokens });
-  const versionStr = "0.2.2" ? `billion-context-omp@${"0.2.2"}` : void 0;
+  const versionStr = "0.2.4" ? `billion-context-omp@${"0.2.4"}` : void 0;
   return buildStatusPanel({
     version: versionStr,
     tokenCount: sessionTokens,
@@ -4757,7 +4807,7 @@ import { join as join3 } from "path";
 import { complete } from "@oh-my-pi/pi-ai";
 import { CONFIG_DIR_NAME as CONFIG_DIR_NAME2 } from "@oh-my-pi/pi-utils";
 var TIMEOUT_MS = 6e4;
-var MAX_OUTPUT_TOKENS = 3e3;
+var MAX_OUTPUT_TOKENS = 8e3;
 var MAX_SLICE_CHARS = 15e4;
 var MAX_MSG_CHARS = 4e3;
 function readCompressModel() {
@@ -4807,9 +4857,10 @@ function parseSummary(text) {
   try {
     const obj = JSON.parse(cleaned);
     if (typeof obj.summary === "string" && obj.summary.length > 0) return obj.summary;
+    return null;
   } catch {
   }
-  return null;
+  return cleaned.length >= 50 ? cleaned : null;
 }
 function buildSummaryPrompt(prompts) {
   return prompts.compressPhilosophy.trim() + "\n\n" + prompts.howToCompressRules.trim() + '\n\nCompress the message range provided below into ONE dense, self-contained technical summary following the rules above. Output ONLY a JSON object: {"summary": "..."} where the value is the full summary as a single string.';
@@ -4846,14 +4897,32 @@ User instructions for this compaction: ${custom}`;
     const userText = `ENTIRE conversation to compress (${slice.length} messages, ~${tokens} tokens). Compress it:
 
 ` + formatSlice(slice, opts?.messageRefs ? { ...createInitialState(), messageRefs: opts.messageRefs } : createInitialState());
-    const response = await run(
-      model,
-      { systemPrompt: [instructions], messages: [{ role: "user", content: [{ type: "text", text: userText }], timestamp: Date.now() }] },
-      { apiKey: auth.apiKey, headers: auth.headers, maxTokens: MAX_OUTPUT_TOKENS, signal: ac.signal }
-    );
-    const summary = parseSummary(
-      response.content.filter((c) => c.type === "text").map((c) => c.text).join("\n")
-    );
+    const attempt = async () => {
+      const response = await run(
+        model,
+        { systemPrompt: [instructions], messages: [{ role: "user", content: [{ type: "text", text: userText }], timestamp: Date.now() }] },
+        { apiKey: auth.apiKey, headers: auth.headers, maxTokens: MAX_OUTPUT_TOKENS, signal: ac.signal }
+      );
+      return parseSummary(
+        response.content.filter((c) => c.type === "text").map((c) => c.text).join("\n")
+      );
+    };
+    let summary = null;
+    try {
+      summary = await attempt();
+    } catch (e) {
+      if (opts?.signal?.aborted || ac.signal.aborted) throw e;
+      logWarn("summarize-messages", { event: "attempt-failed", model: label, error: String(e) });
+    }
+    if (!summary) {
+      try {
+        summary = await attempt();
+        if (summary) logInfo("summarize-messages", { event: "recovered-on-retry", model: label, messages: slice.length });
+      } catch (e) {
+        if (opts?.signal?.aborted || ac.signal.aborted) throw e;
+        logWarn("summarize-messages", { event: "retry-failed", model: label, error: String(e) });
+      }
+    }
     if (!summary) {
       logWarn("summarize-messages", { event: "unparseable-summary", model: label, messages: slice.length });
       return null;
@@ -4893,7 +4962,7 @@ You have four context-management tools:
 
 - compress \u2014 Replace a contiguous range of older conversation with a single detailed summary you write. Use when content is genuinely consumed (no longer needed for the current task step). Single range: compress({ content: [{ topic: "Session Opener", startId: "m00150", endId: "m00220", summary: "..." }] }) \u2014 topic is recommended but optional. Batch (multiple unrelated ranges, each with its own topic): compress({ content: [{ topic: "Auth", startId: "m00150", endId: "m00220", summary: "..." }, { topic: "Deploy", startId: "m00300", endId: "m00350", summary: "..." }] }). Call it as a normal tool \u2014 summaries are plain string arguments.
 - decompress \u2014 Restore a previously compressed block's content. The block stays compressed \u2014 context and cache prefix are not disrupted. By DEFAULT content is written to an auto-generated file (avoids context bloat); use the read tool to view it. Pass inline:true to return content in the tool result instead (appends to context). full:true recurses to original messages. Example: decompress({ blockId: "b5" }) or decompress({ blockId: "b5", full: true }) or decompress({ blockId: "b5", inline: true }).
-- search_context \u2014 Search compressed block summaries (and optionally visible messages) by keyword. Use BEFORE decompressing to find the right block. Example: search_context({ query: "auth token refresh" }).
+- search_context \u2014 Search compressed block summaries AND the original messages folded into them by keyword (messages still visible in context are not indexed \u2014 you can already see them). Use BEFORE decompressing to find the right block. Example: search_context({ query: "auth token refresh" }).
 - acp_status \u2014 Context status with compressible ranges. No args = overview + totals. scope:"uncompressed" for range view; add view:"messages" for per-message listing. scope:"compressed" for block details.
 
 ${prompts.compressPhilosophy}
@@ -5216,7 +5285,7 @@ async function checkForUpdate(autoUpdate, notify) {
     const data = await res.json();
     const latest = data.version;
     if (!latest) return;
-    const current = runtimeVersion ?? "0.2.2";
+    const current = runtimeVersion ?? "0.2.4";
     const hasUpdate = isNewer(latest, current);
     debug.event("update-check", {
       current,
@@ -5251,20 +5320,20 @@ async function getRuntimeVersion() {
 }
 
 // src/dump.ts
-import { mkdirSync as mkdirSync2, writeFileSync as writeFileSync2, readdirSync, unlinkSync } from "fs";
+import { mkdirSync as mkdirSync2, writeFileSync as writeFileSync2, readdirSync as readdirSync2, unlinkSync as unlinkSync2 } from "fs";
 import * as path2 from "path";
 import { CONFIG_DIR_NAME as CONFIG_DIR_NAME4 } from "@oh-my-pi/pi-utils";
 var counters = {};
 var MAX_FILES_PER_PREFIX = 200;
 function pruneDumps(dir, prefixTest, seqOf) {
   try {
-    const files = readdirSync(dir).filter(prefixTest).map((f) => ({ f, n: seqOf(f) })).filter((x) => !Number.isNaN(x.n));
+    const files = readdirSync2(dir).filter(prefixTest).map((f) => ({ f, n: seqOf(f) })).filter((x) => !Number.isNaN(x.n));
     if (files.length <= MAX_FILES_PER_PREFIX) return;
     files.sort((a, b) => a.n - b.n);
     const excess = files.slice(0, files.length - MAX_FILES_PER_PREFIX);
     for (const { f } of excess) {
       try {
-        unlinkSync(path2.join(dir, f));
+        unlinkSync2(path2.join(dir, f));
       } catch {
       }
     }
@@ -5282,7 +5351,7 @@ function dumpContextMessages(messages, meta) {
     mkdirSync2(dir, { recursive: true });
     if (!(dir in counters)) {
       try {
-        const existing = readdirSync(dir).filter((f) => /^\d{4}\.json$/.test(f));
+        const existing = readdirSync2(dir).filter((f) => /^\d{4}\.json$/.test(f));
         const max = existing.reduce((mx, f) => {
           const n = parseInt(f, 10);
           return Number.isNaN(n) ? mx : Math.max(mx, n);
@@ -5365,7 +5434,7 @@ function dumpProviderRequest(payload, meta) {
     mkdirSync2(dir, { recursive: true });
     if (!("req" in counters)) {
       try {
-        const existing = readdirSync(dir).filter((f) => /^req_\d+\.json$/.test(f));
+        const existing = readdirSync2(dir).filter((f) => /^req_\d+\.json$/.test(f));
         const max = existing.reduce((mx, f) => {
           const n = parseInt(f.slice(4, -5), 10);
           return Number.isNaN(n) ? mx : Math.max(mx, n);
@@ -5404,10 +5473,11 @@ import * as path3 from "path";
 import { CONFIG_DIR_NAME as CONFIG_DIR_NAME5 } from "@oh-my-pi/pi-utils";
 async function loadUserConfig(cwd) {
   const home = homeDir();
+  const globalDir = join8(home, CONFIG_DIR_NAME5);
   const merged = {};
-  for (const base of [join8(home, CONFIG_DIR_NAME5), join8(cwd, CONFIG_DIR_NAME5)]) {
+  for (const base of [globalDir, join8(cwd, CONFIG_DIR_NAME5)]) {
     const file = join8(base, "acp-omp.json");
-    const allowPrompts = base.startsWith(home);
+    const allowPrompts = base === globalDir;
     try {
       const raw = await fs.readFile(file, "utf8");
       const parsed = JSON.parse(raw);
@@ -5504,8 +5574,8 @@ function wireCompactionDisable(pi, runtime) {
         messageRefs: slot.state.messageRefs
       });
       if (!result) {
-        ctx.ui?.notify?.("ACP: compression fell back to Pi native compaction", "warning");
-        return void 0;
+        ctx.ui?.notify?.("ACP: /compact aborted \u2014 summary generation failed after retry (details in ~/.omp/acp-omp.log)", "error");
+        return { cancel: true };
       }
       logInfo("compact", {
         sid,
@@ -5532,9 +5602,9 @@ function wireCompactionDisable(pi, runtime) {
 function wireSessionLifecycle(pi, runtime) {
   pi.on("session_start", async (_event, ctx) => {
     const sid = ctx.sessionManager.getSessionId();
-    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.2.2" : null });
+    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.2.4" : null });
     const selfPath = import.meta.url;
-    const conflict = stampAndDetect(selfPath, true ? "0.2.2" : null);
+    const conflict = stampAndDetect(selfPath, true ? "0.2.4" : null);
     if (conflict) {
       logWarn("instance", { event: "dual-instance", self: selfPath, other: conflict.path, otherPid: conflict.pid, otherVersion: conflict.version });
       try {
@@ -5573,6 +5643,7 @@ function wireSessionLifecycle(pi, runtime) {
 async function transformStream(ctx, runtime, input, mode) {
   const sid = ctx.sessionManager.getSessionId();
   const release = await runtime.acquireLock(sid);
+  let result;
   try {
     if (input.length === 0) {
       debug.event("empty-stream-bypass", { sid });
@@ -5678,21 +5749,25 @@ ${rendered.text}${example}`);
         }
       }
     }
+    if (mode === "context") {
+      runtime.recordRebuiltOutput(ctx, rebuilt);
+    }
     dumpContextMessages(rebuilt, {
       sid,
       injected: nudgeInjected,
       emergency: turn.nudge?.breakdown?.emergencyOverride === 1
     });
-    await checkForUpdate(runtime.adapter.autoUpdate ?? true, (msg) => {
-      if (ctx.hasUI) ctx.ui.notify(msg);
-    });
-    return { rebuilt, nudgeInjected };
+    result = { rebuilt, nudgeInjected };
   } catch (e) {
     logThrow("context", e, { sid, phase: "transform", mode });
     throw e;
   } finally {
     release();
   }
+  await checkForUpdate(runtime.adapter.autoUpdate ?? true, (msg) => {
+    if (ctx.hasUI) ctx.ui.notify(msg);
+  });
+  return result;
 }
 function wireContextTransform(pi, runtime) {
   pi.on("context", async (event, ctx) => {