From fbf83f1cbbf7864c56f6bfa3e6ca7f0120b3cec4 Mon Sep 17 00:00:00 2001 From: Jeffery Date: Tue, 23 Jun 2026 15:36:19 +0800 Subject: [PATCH 1/5] =?UTF-8?q?feat(ai-review=20=E8=A1=8C=E8=99=9F):=20?= =?UTF-8?q?=E8=A7=92=E8=89=B2=20prompt=20=E5=BC=B7=E5=88=B6=E8=A1=8C?= =?UTF-8?q?=E8=99=9F=EF=BC=8C=E7=BC=BA=E8=A1=8C=E8=99=9F=E6=99=82=E5=8F=8D?= =?UTF-8?q?=E5=95=8F=E5=8E=9F=E8=A7=92=E8=89=B2=E4=BE=9D=20diff=20?= =?UTF-8?q?=E5=AE=9A=E4=BD=8D=EF=BC=88=E9=87=8D=E8=A9=A6=E4=B8=8A=E9=99=90?= =?UTF-8?q?=EF=BC=89?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- app/findings.js | 61 ++++++++++++++++++++++++++++++++++++++++++++++++- app/main.js | 4 +++- app/roles.js | 21 ++++++++++++++++- 3 files changed, 83 insertions(+), 3 deletions(-) diff --git a/app/findings.js b/app/findings.js index 2d8687a..c055672 100644 --- a/app/findings.js +++ b/app/findings.js @@ -1,7 +1,7 @@ import fs from 'fs'; import path from 'path'; import { chatJSON } from './llm.js'; -import { buildAnalysisPrompt, loadRole, buildVerdictPrompt } from './roles.js'; +import { buildAnalysisPrompt, loadRole, buildVerdictPrompt, buildLocateLinePrompt } from './roles.js'; import { FINDINGS_PATH, EXCLUSIONS_PATH } from './config.js'; import { line, ok, warn } from './log.js'; @@ -244,6 +244,65 @@ function fallback(label, findings, e) { return findings; } +const MAX_LOCATE_ATTEMPTS = 3; + +/** 從 location 取出行號；無 `檔案:行號`（或多檔逗號）時回 null。 */ +function findingLine(location) { + const s = String(location || '').trim(); + if (!s || s.includes(',')) return null; + const m = /^(.+?):(\d+)(?:-\d+)?$/.exec(s); + return m ? Number(m[2]) : null; +} + +/** 從整份 unified diff 擷取指定檔案的區段，找不到時回退整份 diff。 */ +function extractFileDiff(diff, file) { + const lines = String(diff || '').split('\n'); + const out = []; + let capturing = false; + for (const l of lines) { + if (l.startsWith('diff --git ')) capturing = l.includes(`b/${file}`) || l.includes(`a/${file}`); + if (capturing) out.push(l); + } + return out.length ? out.join('\n') : String(diff || ''); +} + +/** + * 對「只有檔名、缺行號」的 findings，反問原角色依該檔 diff 找出行號， + * 重複嘗試直到取得有效行號（每條最多 maxAttempts 次，避免無限迴圈）； + * 成功則把 location 補成 `檔案:行號`，否則保留原檔名。 + */ +export async function resolveMissingLineNumbers(findings, diff, deps = {}) { + const { chatFn = chatJSON, getRole = loadRole, maxAttempts = MAX_LOCATE_ATTEMPTS } = deps; + let resolved = 0; + let pending = 0; + for (const f of findings) { + if (findingLine(f.location) != null) continue; // 已有行號 + const file = String(f.location || '').split(',')[0].split(':')[0].trim(); + if (!file) continue; + pending += 1; + const systemPrompt = buildLocateLinePrompt(getRole(f.role) || { name: f.role }); + const userContent = `${JSON.stringify({ file, problem: f.problem, suggestion: f.suggestion })}\n\n--- ${file} Git Diff ---\n${extractFileDiff(diff, file)}`; + let located = null; + for (let attempt = 1; attempt <= maxAttempts && located == null; attempt++) { + try { + const res = await chatFn(systemPrompt, userContent); + const ln = Number(res?.line); + if (Number.isInteger(ln) && ln > 0) located = ln; + } catch (e) { + warn(`[${f.role}] 行號定位失敗（第 ${attempt}/${maxAttempts} 次）: ${e.message}`); + } + } + if (located != null) { + f.location = `${file}:${located}`; + resolved += 1; + } else { + warn(`[${f.role}] ${maxAttempts} 次嘗試後仍無法定位行號，保留檔名: ${file}`); + } + } + if (pending > 0) ok(`補行號: ${resolved}/${pending} 筆成功定位`); + return findings; +} + /** 只保留 AI 需要的欄位，減少 token 用量 */ function toAIPayload(findings) { return findings.map(({ level, role, location, problem, suggestion }) => ({ level, role, location, problem, suggestion })); diff --git a/app/main.js b/app/main.js index 23a70f9..6339511 100644 --- a/app/main.js +++ b/app/main.js @@ -2,7 +2,7 @@ import path from 'path'; import { GITEA_REPOSITORY, PR_NUMBER, PR_HEAD_BRANCH, PR_BASE_BRANCH, getLLMConfig, FINDINGS_PATH, EXCLUSIONS_PATH } from './config.js'; import { loadRoles, getRoleIntro } from './roles.js'; import { getPRDiff, postComment, getCommitMessageBySha, getBotReviewOutcome, shouldSkipBotCommit } from './gitea.js'; -import { analyzeWithRole, loadOldFindings, mergeFindings, sortByLevel, deduplicateWithAI, loadExclusions, applyExclusions, filterFalsePositivesWithAI, appendExclusions } from './findings.js'; +import { analyzeWithRole, loadOldFindings, mergeFindings, sortByLevel, deduplicateWithAI, loadExclusions, applyExclusions, filterFalsePositivesWithAI, appendExclusions, resolveMissingLineNumbers } from './findings.js'; import { reconcileConversations, dropResolvedFindings, addCarriedFindings } from './resolve.js'; import { saveFindings, postFindingsReview, formatFindingsStatsLine } from './comments.js'; import { getRunUsage, getRateLimit, fetchAccountQuota, formatUsageStats, formatUsageStatsLine } from './usage.js'; @@ -88,6 +88,8 @@ async function main() { if (analyses[i].status === 'fulfilled') newFindings.push(...analyses[i].value); else warn(`[${roles[i].name}] 分析失敗（跳過）: ${analyses[i].reason?.message}`); } + // 對只有檔名、缺行號的問題，反問原角色補上行號（最多重試數次），確保後續能行內標註 + await resolveMissingLineNumbers(newFindings, diff); output(`新 findings ${newFindings.length} 筆（${formatFindingsStatsLine(newFindings)}）`); // Step6 合併與去重：舊問題 + 對話收斂結果 + 新問題 → 語意去重 diff --git a/app/roles.js b/app/roles.js index 675e86e..67fd066 100644 --- a/app/roles.js +++ b/app/roles.js @@ -70,7 +70,7 @@ export function buildAnalysisPrompt(role) { '{', ' "level": "critical|warning|info",', ` "role": "${role.name}",`, - ' "location": "檔案路徑:行號 或 檔案路徑",', + ' "location": "檔案路徑:行號（行號為必填，例如 app/foo.js:42）",', ' "problem": "繁體中文（台灣用語）說明審查員認為這裡有問題的原因，不要只填檔案路徑或行號",', ' "suggestion": "繁體中文（台灣用語）的具體修改建議"', '}', @@ -80,10 +80,29 @@ export function buildAnalysisPrompt(role) { '- warning：建議修正的問題', '- info：可選的改善建議', '', + 'location 規則（務必遵守）：', + '- **每一條問題都必須帶行號**，格式一律為 `檔案路徑:行號`（單一行號，例如 `app/foo.js:42`）。', + '- 嚴禁只給檔名而省略行號；行號請取該問題在 Git Diff 新增/修改處的實際行號。', + '- 一條問題只對應一個檔案與一個行號，不要用逗號列多個檔案。', + '', '只回傳 JSON 陣列，不要有其他文字。如果沒有問題，回傳空陣列 []。', ].filter(l => l !== '').join('\n'); } +/** + * 由角色定義組出「補行號」的 system prompt： + * 當該角色先前提出的問題只有檔名、缺行號時，請它對照 Git Diff 找出實際行號。 + */ +export function buildLocateLinePrompt(role) { + const name = role?.name || 'AI Review'; + const badge = role?.badge ? `${role.badge} ` : ''; + return [ + `你是 ${badge}${name}${role?.focus ? `（負責「${role.focus}」面向）` : ''}。`, + '你先前提出了一個問題，但 location 只給了檔名、沒有行號。請對照下方提供的該檔案 Git Diff，找出這個問題對應的**實際行號**（新增/修改處在該檔案中的行號）。', + '只回傳 JSON 物件：{"line": 數字}，不要有其他文字。若 diff 中確實找不到對應行，回傳 {"line": 0}。', + ].join('\n'); +} + /** * 由防守方角色定義組出「單條 finding 誤報裁決」的 system prompt： * 套用其個性與裁決準則本文，要求對一條 finding 判定成立或誤報，回固定 JSON 物件。 From d03f08e19d058e1d78df44b3b771b96ec9a29eef Mon Sep 17 00:00:00 2001 From: Jeffery Date: Tue, 23 Jun 2026 15:36:19 +0800 Subject: [PATCH 2/5] =?UTF-8?q?test(ai-review):=20=E8=A3=9C=E8=A1=8C?= =?UTF-8?q?=E8=99=9F=E5=AE=9A=E4=BD=8D=E3=80=81=E8=AA=A4=E5=A0=B1=E8=A3=81?= =?UTF-8?q?=E6=B1=BA=E9=82=8A=E7=95=8C=E3=80=81=E7=B5=B1=E8=A8=88=E8=88=87?= =?UTF-8?q?=20appendExclusions=20=E6=B8=AC=E8=A9=A6?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- app/comments.test.js | 16 ++++++++++++ app/findings.test.js | 58 +++++++++++++++++++++++++++++++++++++++++++- app/roles.test.js | 25 ++++++++++++++++++- 3 files changed, 97 insertions(+), 2 deletions(-) diff --git a/app/comments.test.js b/app/comments.test.js index 06b3fbc..e1c0792 100644 --- a/app/comments.test.js +++ b/app/comments.test.js @@ -307,6 +307,22 @@ describe('postFindingsReview', () => { assert.equal(reviewCalls[0].body, reviewCalls[0].body.trimEnd()); }); + it('counts both new and old findings in the summary but only inline-comments new ones', async () => { + const reviewCalls = []; + await postFindingsReview([ + { level: 'critical', role: 'Rex', location: 'app/a.js:10', suggestion: 'old crit', is_new: false }, + { level: 'warning', role: 'Leo', location: 'app/b.js:20', suggestion: 'new warn', is_new: true }, + { level: 'info', role: 'Maya', location: 'app/c.js:30', suggestion: 'new info', is_new: true }, + ], { postReview: async (a) => { reviewCalls.push(a); } }); + + const body = reviewCalls[0].body; + assert.match(body, /\| 新問題 \| 0 筆 \| 1 筆 \| 1 筆 \| 0 筆 \|/); + assert.match(body, /\| 舊問題 \| 1 筆 \| 0 筆 \| 0 筆 \| 0 筆 \|/); + // 舊問題 app/a.js 不產生行內 comment；只有新問題被標註 + assert.ok(!reviewCalls[0].comments.some(c => c.path === 'app/a.js')); + assert.deepEqual(reviewCalls[0].comments.map(c => c.path), ['app/b.js', 'app/c.js']); + }); + it('separates old and new findings in default review statistics', async () => { const reviewCalls = []; await postFindingsReview([ diff --git a/app/findings.test.js b/app/findings.test.js index 28bff17..a15ae03 100644 --- a/app/findings.test.js +++ b/app/findings.test.js @@ -3,7 +3,7 @@ import assert from 'node:assert/strict'; import fs from 'node:fs'; import os from 'node:os'; import path from 'node:path'; -import { loadOldFindings, loadExclusions, applyExclusions, filterFalsePositivesWithAI, appendExclusions } from './findings.js'; +import { loadOldFindings, loadExclusions, applyExclusions, filterFalsePositivesWithAI, appendExclusions, resolveMissingLineNumbers } from './findings.js'; import { EXCLUSIONS_PATH, FINDINGS_PATH } from './config.js'; describe('findings exclusions', () => { @@ -59,6 +59,18 @@ describe('findings exclusions', () => { assert.equal(merged.length, 2); }); + it('appendExclusions keeps same-path entries that have different original text', () => { + const fullPath = path.join(workspace, EXCLUSIONS_PATH); + fs.mkdirSync(path.dirname(fullPath), { recursive: true }); + fs.writeFileSync(fullPath, JSON.stringify([{ location: 'app/a.js:1', original_finding: '問題甲' }], null, 2)); + + appendExclusions(workspace, [{ location: 'app/a.js:5', original_finding: '問題乙', reason: 'r' }]); + + const onDisk = JSON.parse(fs.readFileSync(fullPath, 'utf8')); + assert.equal(onDisk.length, 2); // 同檔但原文不同 → 視為不同排除條目，兩者皆保留 + assert.deepEqual(onDisk.map(e => e.original_finding), ['問題甲', '問題乙']); + }); + it('writes appended exclusions to both workspace and mirror dir', () => { const repoRoot = path.join(workspace, 'repo'); fs.mkdirSync(repoRoot, { recursive: true }); @@ -225,6 +237,50 @@ describe('findings exclusions', () => { assert.equal(result.length, 2); }); + it('keeps a finding when the defender returns an out-of-range verdict value', async () => { + const findings = [{ level: 'warning', role: 'Mage', location: 'a.js:1', problem: 'p', suggestion: 's' }]; + const chatFn = async () => ({ verdict: 'maybe', reason: 'x' }); // 非 confirmed/false_positive + const result = await filterFalsePositivesWithAI(findings, [], chatFn); + assert.equal(result.length, 1); // 只有明確 false_positive 才剔除，其餘保守保留 + }); + + it('resolveMissingLineNumbers fills missing line numbers by re-asking the role', async () => { + const findings = [ + { level: 'critical', role: 'Maya', location: 'app/a.js', problem: 'p', suggestion: 's' }, + { level: 'warning', role: 'Leo', location: 'app/b.js:20', problem: 'p', suggestion: 's' }, // 已有行號 → 不動 + ]; + let calls = 0; + const chatFn = async () => { calls += 1; return { line: 42 }; }; + + await resolveMissingLineNumbers(findings, 'diff --git a/app/a.js b/app/a.js\n@@ -1 +1 @@', { chatFn, getRole: () => ({ name: 'Maya' }) }); + + assert.equal(findings[0].location, 'app/a.js:42'); // 補上行號 + assert.equal(findings[1].location, 'app/b.js:20'); // 不變 + assert.equal(calls, 1); // 只對缺行號者呼叫 + }); + + it('resolveMissingLineNumbers retries until a valid line appears', async () => { + const findings = [{ level: 'warning', role: 'Leo', location: 'app/x.js', problem: 'p', suggestion: 's' }]; + let n = 0; + const chatFn = async () => { n += 1; return n < 3 ? { line: 0 } : { line: 7 }; }; + + await resolveMissingLineNumbers(findings, 'd', { chatFn, getRole: () => null, maxAttempts: 5 }); + + assert.equal(findings[0].location, 'app/x.js:7'); + assert.equal(n, 3); // 第三次才給出有效行號 + }); + + it('resolveMissingLineNumbers keeps the filename after exhausting retries', async () => { + const findings = [{ level: 'warning', role: 'Leo', location: 'app/y.js', problem: 'p', suggestion: 's' }]; + let n = 0; + const chatFn = async () => { n += 1; return { line: 0 }; }; + + await resolveMissingLineNumbers(findings, 'd', { chatFn, getRole: () => null, maxAttempts: 3 }); + + assert.equal(findings[0].location, 'app/y.js'); // 仍保留檔名 + assert.equal(n, 3); // 嘗試 3 次後放棄 + }); + it('logs exclusions file metadata and repo state when loading exclusions', () => { const fullPath = path.join(workspace, EXCLUSIONS_PATH); fs.mkdirSync(path.dirname(fullPath), { recursive: true }); diff --git a/app/roles.test.js b/app/roles.test.js index d06a502..80d7335 100644 --- a/app/roles.test.js +++ b/app/roles.test.js @@ -1,6 +1,6 @@ import { describe, it } from 'node:test'; import assert from 'node:assert/strict'; -import { parseRoleFile, loadRoles, loadRole, buildAnalysisPrompt, getRoleIntro } from './roles.js'; +import { parseRoleFile, loadRoles, loadRole, buildAnalysisPrompt, buildLocateLinePrompt, getRoleIntro } from './roles.js'; const SAMPLE = `--- name: Tester @@ -81,6 +81,29 @@ describe('buildAnalysisPrompt', () => { }); }); +describe('buildAnalysisPrompt 行號要求', () => { + it('requires a line number in location', () => { + const prompt = buildAnalysisPrompt(parseRoleFile(SAMPLE)); + assert.match(prompt, /行號為必填/); + assert.match(prompt, /每一條問題都必須帶行號/); + }); +}); + +describe('buildLocateLinePrompt', () => { + it('asks the same role to return a JSON line number', () => { + const prompt = buildLocateLinePrompt({ name: 'Maya', badge: '🧪', focus: 'testing' }); + assert.match(prompt, /Maya/); + assert.match(prompt, /找出.*行號|實際行號/); + assert.match(prompt, /\{"line": 數字\}/); + }); + + it('tolerates a bare role object without badge/focus', () => { + const prompt = buildLocateLinePrompt({ name: 'Leo' }); + assert.match(prompt, /Leo/); + assert.doesNotMatch(prompt, /undefined/); + }); +}); + describe('getRoleIntro', () => { it('renders a table row per role with its badge', () => { const intro = getRoleIntro([parseRoleFile(SAMPLE)]); From 86b343bec9876885377a26adf68ed4c2c8bad014 Mon Sep 17 00:00:00 2001 From: Jeffery Date: Tue, 23 Jun 2026 15:36:19 +0800 Subject: [PATCH 3/5] =?UTF-8?q?docs(ai-review):=20=E8=A3=9C=20location=20?= =?UTF-8?q?=E8=A1=8C=E8=99=9F=E5=BC=B7=E5=88=B6=E8=88=87=E5=8F=8D=E5=95=8F?= =?UTF-8?q?=E5=AE=9A=E4=BD=8D=E6=B5=81=E7=A8=8B=E8=AA=AA=E6=98=8E?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/README.md b/README.md index 1064621..9aab1a4 100644 --- a/README.md +++ b/README.md @@ -44,6 +44,7 @@ - 帳號額度：`fetchAccountQuota` 依平台採不同策略——OpenRouter（`openai` slot 指向 openrouter.ai 時）以 `GET /auth/key` 取得 USD credits 已用／上限／剩餘；Ollama、OpenCode 為本地／自架服務回報「不適用」；OpenAI、Claude、Gemini、Amazon Q 的帳號額度需 org/admin 權限，API key 無法取得時誠實回報原因 - 兩種來源皆無法取得（例如帳號無上限且回應無速率 header）時，降級為「無法計算百分比」並附原因，不中斷流程；`formatUsageStats` 產生 Review 本文區塊，`formatUsageStatsLine` 產生單行 log 摘要 14. 誤報判斷套用「防守方」角色：`app/findings.js` 的 `filterFalsePositivesWithAI` 以 `app/roles.js` 的 `loadRole('Paladin')` 載入防守方角色，並用 `buildVerdictPrompt(role, exclusionHint)` 組出帶其個性與裁決準則的 system prompt；對每一條 finding 各派一個防守方 sub-agent（`judgeFindingIsFalsePositive`）裁決 `confirmed`／`false_positive`，多條問題時以 `Promise.all` 平行處理；判為誤報者剔除、成立者保留，任一 sub-agent 失敗（含解析失敗）保守視為成立保留，不中斷流程。角色檔遺失時 `buildVerdictPrompt(null)` 退回通用裁判 prompt。 +15. location 行號強制：`buildAnalysisPrompt` 明確要求每條問題的 `location` 必須是 `檔案路徑:行號`（單一行號、不可只給檔名），否則該問題無法在 Review 行內標註、只剩統計數字。Step5 角色分析後由 `resolveMissingLineNumbers` 把關：對「只有檔名、缺行號」的新問題，用 `buildLocateLinePrompt(role)` 反問**原角色**、附該檔 diff 區段（`extractFileDiff`）請它回 `{"line": 數字}`，重複嘗試到取得有效行號為止（每條最多 `MAX_LOCATE_ATTEMPTS=3` 次，避免無限迴圈）；成功補成 `檔案:行號`，連續失敗則記錄警告並保留檔名。 # 使用說明 From 892a79c9bc5c843cb4657cc2c4ab007d7d1cc303 Mon Sep 17 00:00:00 2001 From: Jeffery Date: Tue, 23 Jun 2026 15:36:19 +0800 Subject: [PATCH 4/5] =?UTF-8?q?chore(ai-review=20=E7=8B=80=E6=85=8B):=20?= =?UTF-8?q?=E8=A7=A3=E6=B1=BA=E8=A1=8C=E8=99=9F=E7=9B=B8=E9=97=9C=20findin?= =?UTF-8?q?gs=E3=80=81=E6=8E=92=E9=99=A4=20extractUsage=20=E8=AA=A4?= =?UTF-8?q?=E5=A0=B1=EF=BC=8C=E6=B8=85=E7=A9=BA=20findings.json?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .gitea/ai-review/exclusions.json | 6 ++++++ .gitea/ai-review/findings.json | 35 +------------------------------- 2 files changed, 7 insertions(+), 34 deletions(-) diff --git a/.gitea/ai-review/exclusions.json b/.gitea/ai-review/exclusions.json index 56f7c61..4ffa10c 100644 --- a/.gitea/ai-review/exclusions.json +++ b/.gitea/ai-review/exclusions.json @@ -482,5 +482,11 @@ "role": "Bard", "original_finding": "`fetchAccountQuota` 中的 `QUOTA_STRATEGIES` 物件定義龐大，將所有平台策略硬編碼於此，未來新增供應商難以維護；建議抽離至獨立檔案或策略模式。", "reason": "過早最佳化（與先前已排除的 usage.js SRP 拆檔建議等價）。目前 QUOTA_STRATEGIES 為精簡的查表物件、各平台策略短小且集中易讀；在供應商數量出現實際膨脹痛點前抽檔，徒增檔案與匯入複雜度。" + }, + { + "location": "app/usage.js", + "role": "Assassin", + "original_finding": "extractUsage 對不預期 payload 僅返回 null，過度信任 API 回應結構，可能導致計費或配額相關監控被繞過；建議增加 Schema Validation、異常明確記錄。", + "reason": "誤判／過度設計。extractUsage 僅用於「使用量顯示統計」，非計費或配額強制；回傳 null 是「此回應無可辨識 usage 資訊」的正確訊號，呼叫端以 0 計入並降級顯示，不影響任何金流或門檻判斷。對 best-effort 顯示統計加 schema validation 與錯誤記錄屬過度設計。" } ] diff --git a/.gitea/ai-review/findings.json b/.gitea/ai-review/findings.json index 467c2be..fe51488 100644 --- a/.gitea/ai-review/findings.json +++ b/.gitea/ai-review/findings.json @@ -1,34 +1 @@ -[ - { - "level": "critical", - "role": "Maya", - "location": "app/comments.test.js", - "problem": "在 `postFindingsReview` 測試中，雖然增加了驗證 usageSection 的案例，但對於舊問題（is_new: false）不應被標註的邏輯，缺乏針對「舊問題數量是否正確統計進總計」的邊界測試。", - "suggestion": "補上一個測試案例：驗證當存在新問題與舊問題時，統計表格中「新問題」與「舊問題」的行數與數字皆正確，且舊問題確實沒有產生對應的行內評論。", - "is_new": true - }, - { - "level": "warning", - "role": "Assassin", - "location": "app/usage.test.js", - "problem": "`extractUsage` 對不預期 payload 僅返回 `null`，過度信任 API 回應結構，可能導致計費或配額相關監控被繞過。", - "suggestion": "增加嚴格結構驗證（Schema Validation），異常時應明確記錄並標示，而非默默忽略。", - "is_new": false - }, - { - "level": "warning", - "role": "Maya", - "location": "app/findings.test.js", - "problem": "在 `appendExclusions` 測試中，測試案例僅檢查了「檔案路徑 + 原文」的去重，未測試當只有檔案路徑相同、但原問題內容不同時的行為（理應視為不同排除條目）。", - "suggestion": "補上測試案例：輸入兩條路徑相同但原問題不同的 exclusion，確保兩者皆被成功寫入。", - "is_new": true - }, - { - "level": "warning", - "role": "Maya", - "location": "app/findings.test.js", - "problem": "在 `filterFalsePositivesWithAI` 的測試中，測試了 LLM 呼叫失敗時會保守保留，但未測試當 `chatFn` 回傳結構不完整（例如缺少 verdict 欄位）時，是否真的有正確過濾或保留。", - "suggestion": "增加一個測試案例，模擬 `chatFn` 回傳一個包含錯誤 verdict 格式的物件，驗證該問題是否如預期被保守保留。", - "is_new": true - } -] +[] From d3dcb36cbdda7d345716a1e7a8990685d3326292 Mon Sep 17 00:00:00 2001 From: AI Review Bot Date: Tue, 23 Jun 2026 07:37:49 +0000 Subject: [PATCH 5/5] chore: update ai-review findings [ai-review-bot][success] --- .gitea/ai-review/findings.json | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/.gitea/ai-review/findings.json b/.gitea/ai-review/findings.json index fe51488..6cebf4b 100644 --- a/.gitea/ai-review/findings.json +++ b/.gitea/ai-review/findings.json @@ -1 +1,10 @@ -[] +[ + { + "level": "warning", + "role": "Maya", + "location": "app/findings.test.js:283", + "problem": "在測試 `resolveMissingLineNumbers` 時，僅測試了 `chatFn` 成功回傳有效或無效行號的情況，但未測試 `chatFn` 拋出例外（Exception）的失敗情境。", + "suggestion": "補上 `chatFn` throw error 的測試案例，驗證該函數是否能妥善處理例外並正確記錄警告資訊，而非讓整個執行流程中斷。", + "is_new": true + } +]