homunculus-code - npm Package Compare versions

Comparing version 0.10.0 to 0.10.1

core/evaluate-session.js

@@ -304,2 +304,12 @@ #!/usr/bin/env node
     log(`Extraction failed: ${err.message?.slice(0, 100)}`);
+    // When called from a CC hook subprocess, claude --print cannot access
+    // Keychain/OAuth tokens and will fail with auth errors. Queue the prompt
+    // so the next manual or scheduled run (outside hook context) can process it.
+    try {
+      ensureDir(REPORTS_DIR);
+      const queueFile = path.join(REPORTS_DIR, 'extraction-queue.jsonl');
+      const entry = JSON.stringify({ timestamp: new Date().toISOString(), prompt });
+      fs.appendFileSync(queueFile, entry + '\n');
+      log(`Queued for retry → ${queueFile} (next run outside hook context will process)`);
+    } catch {}
     return null;
@@ -309,2 +319,43 @@ }
 
+// Process any queued extraction prompts (written when called from hook context)
+function processQueue({ existing, todayCount, dynamicLimit }) {
+  const queueFile = path.join(REPORTS_DIR, 'extraction-queue.jsonl');
+  if (!fs.existsSync(queueFile)) return { instincts: 0, memory: 0, research: 0 };
+  // Still in hook context — skip, let the next run handle it
+  if (process.env.CLAUDECODE) return { instincts: 0, memory: 0, research: 0 };
+
+  const lines = fs.readFileSync(queueFile, 'utf8').trim().split('\n').filter(Boolean);
+  if (!lines.length) return { instincts: 0, memory: 0, research: 0 };
+
+  log(`Processing ${lines.length} queued extraction(s)...`);
+  const stats = { instincts: 0, memory: 0, research: 0 };
+
+  for (const line of lines) {
+    let entry;
+    try { entry = JSON.parse(line); } catch { continue; }
+    if (!entry.prompt) continue;
+
+    const env = { ...process.env };
+    delete env.CLAUDECODE;
+    const model = process.env.HOMUNCULUS_HARVEST_MODEL || 'claude-sonnet-4-6';
+    try {
+      const result = execSync(
+        `claude --print --model ${model} --max-turns 1 --no-session-persistence`,
+        { input: entry.prompt, encoding: 'utf8', timeout: 120000, env, stdio: ['pipe', 'pipe', 'pipe'] }
+      ).trim();
+      const s = processResult(result, { existing, todayCount: todayCount + stats.instincts, dynamicLimit });
+      stats.instincts += s.instincts;
+      stats.memory += s.memory;
+      stats.research += s.research;
+    } catch (err) {
+      log(`Queue entry failed: ${err.message?.slice(0, 80)}`);
+    }
+  }
+
+  // Clear processed queue
+  try { fs.unlinkSync(queueFile); } catch {}
+  log(`Queue done: ${stats.instincts} instincts, ${stats.memory} memory, ${stats.research} research`);
+  return stats;
+}
+
 // Process extraction result — parse JSON lines and route to appropriate handlers
@@ -427,9 +478,18 @@ function processResult(result, { existing, todayCount, dynamicLimit }) {
 
+  // Process any prompts queued from previous hook-context runs
+  const queueStats = processQueue({ existing, todayCount, dynamicLimit });
+  const afterQueue = todayCount + queueStats.instincts;
+  if (afterQueue >= dynamicLimit) {
+    log(`Limit reached after queue processing (${afterQueue}/${dynamicLimit}), skipping current session`);
+    if (queueStats.instincts > 0) updateCooldown(afterQueue);
+    return;
+  }
+
   log(`Analyzing: ${analysis.total_observations} observations, ${analysis.frequent_tools.length} patterns (limit: ${dynamicLimit})`);
 
   const result = extractFromSession(analysis);
-  const stats = processResult(result, { existing, todayCount, dynamicLimit });
+  const stats = processResult(result, { existing, todayCount: afterQueue, dynamicLimit });
 
-  if (stats.instincts > 0) {
-    updateCooldown(todayCount + stats.instincts);
+  if (stats.instincts > 0 || queueStats.instincts > 0) {
+    updateCooldown(afterQueue + stats.instincts);
   }
@@ -436,0 +496,0 @@

docs/zh-CN/README.md

@@ -169,2 +169,15 @@ **语言：** [English](../../README.md) | [繁體中文](../zh-TW/README.md) | [简体中文](README.md)
 
+### v0.10.1 — Hook 认证 Fallback (Apr 2026)
+
+- **Hook context queue fallback** — `evaluate-session.js` 现在能优雅地处理从 CC hook 子进程调用时 `claude --print` 失败的情况（Keychain/OAuth 无法访问）。失败的提取会写入 `homunculus/reports/extraction-queue.jsonl`，下次在 hook context 外执行时（手动调用或 `/hm-night`）自动消费
+
+### v0.10.0 — Durability Gate & Multi-Run Eval (Apr 2026)
+
+- **Durability Gate** — `durability_score < 0.7` 的 instinct 会被过滤；事件特定模式不再污染系统
+- **Gaming Gate** — 分数跳升 >5pp 但只改了 ≤3 行的情况标记为 `gaming_suspected` 并丢弃
+- **Multi-run eval** — `--runs N` 和 `--passes N` 支持多次独立 eval + 多数决，降低 LLM 评判误差
+- **CLAUDE.md 覆盖率检查** — 已实现的规则注入提取 prompt，避免重复提取既有行为
+- **Bash 失败断路器** — `observe.sh` 追踪最近 10 次失败，用于进化分析
+- **文档简化** — 移除对外文档中的 P0/P1A/B/C 内部命名，更清晰的心智模型
+
 ### v0.9.0 — 进化引擎升级 (Mar 2026)
@@ -171,0 +184,0 @@

docs/zh-TW/README.md

@@ -169,2 +169,15 @@ **語言：** [English](../../README.md) | [繁體中文](README.md) | [简体中文](../zh-CN/README.md)
 
+### v0.10.1 — Hook 認證 Fallback (Apr 2026)
+
+- **Hook context queue fallback** — `evaluate-session.js` 現在能優雅地處理從 CC hook 子進程呼叫時 `claude --print` 失敗的情況（Keychain/OAuth 無法存取）。失敗的提取會寫入 `homunculus/reports/extraction-queue.jsonl`，下次在 hook context 外執行時（手動呼叫或 `/hm-night`）自動消費
+
+### v0.10.0 — Durability Gate & Multi-Run Eval (Apr 2026)
+
+- **Durability Gate** — `durability_score < 0.7` 的 instinct 會被過濾；事件特定模式不再污染系統
+- **Gaming Gate** — 分數跳升 >5pp 但只改了 ≤3 行的情況標記為 `gaming_suspected` 並丟棄
+- **Multi-run eval** — `--runs N` 和 `--passes N` 支援多次獨立 eval + 多數決，降低 LLM 評判誤差
+- **CLAUDE.md 覆蓋率檢查** — 已實作的規則注入提取 prompt，避免重複提取既有行為
+- **Bash 失敗斷路器** — `observe.sh` 追蹤最近 10 次失敗，用於演化分析
+- **文件簡化** — 移除對外文件中的 P0/P1A/B/C 內部命名，更清晰的心智模型
+
 ### v0.9.0 — Evolution Engine 升級 (Mar 2026)
@@ -171,0 +184,0 @@

package.json

 {
   "name": "homunculus-code",
-  "version": "0.10.0",
+  "version": "0.10.1",
   "description": "A self-evolving AI assistant that grows smarter every night",
@@ -5,0 +5,0 @@ "bin": {

README.md

@@ -169,2 +169,15 @@ **Language:** [English](README.md) | [繁體中文](docs/zh-TW/README.md) | [简体中文](docs/zh-CN/README.md)
 
+### v0.10.1 — Hook Auth Fallback (Apr 2026)
+
+- **Queue fallback for hook context** — `evaluate-session.js` now gracefully handles the case where `claude --print` fails from a CC hook subprocess (Keychain/OAuth not accessible). Failed extractions are queued to `homunculus/reports/extraction-queue.jsonl` and automatically processed on the next run outside hook context (e.g. manual call or `/hm-night`)
+
+### v0.10.0 — Durability Gate & Multi-Run Eval (Apr 2026)
+
+- **Durability Gate** — Instincts with `durability_score < 0.7` are filtered out; event-specific patterns no longer pollute the system
+- **Gaming Gate** — Score jumps >5pp with ≤3 net new lines are flagged as `gaming_suspected` and discarded
+- **Multi-run eval** — `--runs N` and `--passes N` flags for multiple independent eval runs and majority vote per scenario, reducing LLM-judge variance
+- **CLAUDE.md coverage check** — Already-implemented rules are injected into the extraction prompt to avoid re-extracting existing behaviors
+- **Bash failure circuit breaker** — `observe.sh` tracks last 10 failures for evolution analysis
+- **Simplified phase pipeline docs** — P0/P1A/B/C internal naming removed from public docs; cleaner mental model
+
 ### v0.9.0 — Evolution Engine Upgrade (Mar 2026)
@@ -171,0 +184,0 @@

New alerts

Environment variable access

Supply chain risk

Package accesses environment variables, which may be a sign of credential stuffing or data theft.

AI-detected potential code anomaly

Supply chain risk

AI has identified unusual behaviors that may pose a security risk.

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Fixed alerts

AI-detected potential code anomaly

Supply chain risk

AI has identified unusual behaviors that may pose a security risk.

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Improved metrics

Total package byte prevSize

537223

1.17%

Lines of code

1476

3.72%

Number of lines in readme file

665

1.99%

Worsened metrics

Number of low supply chain risk alerts

21

16.67%

No dependency changes