From 00ec84d14ec963bb0500a1e9b4fecd1b11294a1e Mon Sep 17 00:00:00 2001 From: RuoJi6 Date: Sat, 24 Oct 2026 19:52:26 +0800 Subject: [PATCH] feat(findings): reduce duplicate-confirmation noise with grouped suggestions and ranked candidates - Prompt: evidenced endpoint/handler/param differences justify independent; suggest only with concrete shared points - Review acceptance: merged no longer blocked by unified report - Suggestions: 2..N member groups, one-click group confirm with per-member exclusion, related cards cleared - Candidates: score/tier/hit_reasons ranking, low-score kept paginated --- agent/finding_cases.go | 8 +- agent/finding_cases_test.go | 2 +- db/finding_cases.go | 321 +++++++++++++++++++---- db/finding_cases_test.go | 159 ++++++++++- db/schema.sql | 11 + server/finding_case_reviews.go | 15 +- server/finding_case_reviews_test.go | 22 +- server/finding_cases.go | 32 ++- web/src/components/finding-case-list.tsx | 59 +++-- web/src/lib/api.ts | 7 +- web/src/lib/types.ts | 1 + 11 files changed, 532 insertions(+), 105 deletions(-) diff --git a/agent/finding_cases.go b/agent/finding_cases.go index 7e69412..370e760 100644 --- a/agent/finding_cases.go +++ b/agent/finding_cases.go @@ -45,13 +45,13 @@ func findingCaseWorkflow(key string, tools []actool.CoreTool) ([]actool.CoreTool } return tools, ` -【同一漏洞的多次上报整理】保留本次独立报告，不覆盖其他原报告。默认报告完成即停止的旧指引在此扩展：先保存本次报告，然后 search_finding_duplicates(finding_id=独立记录ID) 分页读取同任务候选摘要，按需 get_finding_record 查看证据。仅凭漏洞类型/标题相似或高中低等级不能归并；须核对目标、接口、参数、权限边界、根因及实际验证证据。 -【判断准则】接口、路径、后端、读/写影响、评级不同，都不能单独作为不重复的理由；同一登录入口、同一令牌或同一漏洞类型也不能单独证明同一根因。比较实际缺失的鉴权检查、对象权限边界、会话门或共同组件，并引用每条记录中的具体证据。读和写可能是同一权限检查缺失的不同影响；也可能是独立 handler 的独立缺陷。只有证据证明共同实际根因才自动归并，证据证明不同根因才判独立；存在关联但无法确认时必须 suggest_finding_merge，说明已知共同点、缺失证据和待确认问题，不用“路径不同/影响不同”代替根因判断。明确同一缺陷时 merge_finding_records(task_id,finding_ids,title,reason)，不确定时 suggest_finding_merge 留待人工确认，不能强行归并。不重复上报、不删除记录、不重新探测目标。人工选择整理时只能判断所选记录，不能加入未选记录（既有组成员可作为报告上下文）。 -归并成功或已有case_id时，get_finding_case 记录实际version，分页读取全部成员，按需get_finding_record、get_finding_traffic读取原报告、PoC及已保存流量。update_finding_case_report保存完整统一报告（包含有内容的概述、影响、前置条件、复现步骤、证据、PoC、根因分析、修复建议章节，不得半篇截断）和基于已验证影响的severity及severity_reason；不能直接取成员最高等级。集中给出前置条件、完整执行顺序、输入输出如何衔接、可运行的完整脚本及命令，不把必要步骤分散到原报告；缺失条件、未验证步骤如实标明。版本冲突重新读取材料并生成，不能仅换version重试。各成员仍有独立报告、评级和证据。没有明确重复则保留独立记录。工具被关闭时不绕过配置。` +【同一漏洞的多次上报整理】保留本次独立报告，不覆盖其他原报告。默认报告完成即停止的旧指引在此扩展：先保存本次报告，然后 search_finding_duplicates(finding_id=独立记录ID) 分页读取同任务候选摘要（按 score 排序；strong=同漏洞类型且资产重叠，优先核对；weak=单一条件命中，仅作参考），按需 get_finding_record 查看证据。仅凭漏洞类型/标题相似或高中低等级不能归并；须核对目标、接口、参数、权限边界、根因及实际验证证据。 +【判断准则】有证据表明端点/路由、处理器或后端组件、易受影响参数、缺失检查属于不同代码路径时，可以判定 independent；但仅凭路径、接口、登录入口、令牌、标题或漏洞类型相同/不同都不能证明根因。比较实际缺失的鉴权检查、对象权限边界、会话门或共同组件，并引用每条记录中的具体证据。只有根因明确相同才用 merge_finding_records；只有存在具体共同点（同端点/参数/组件/缺失检查）但缺关键证据确认时才用 suggest_finding_merge，并一次提交覆盖2..N条记录的分组，禁止仅因类型或资产相同而建议。没有具体共同点就保持 independent，不要因不确定而建议。不重复上报、不删除记录、不重新探测目标。人工选择整理时只能判断所选记录，不能加入未选记录（既有组成员可作为报告上下文）。 +归并不以统一报告为前提：根因明确就先归并，统一报告可随后补齐（未完成时文件夹显示“统一报告待更新”）。归并成功或已有case_id时，get_finding_case 记录实际version，分页读取全部成员，按需get_finding_record、get_finding_traffic读取原报告、PoC及已保存流量。update_finding_case_report保存完整统一报告（包含有内容的概述、影响、前置条件、复现步骤、证据、PoC、根因分析、修复建议章节，不得半篇截断）和基于已验证影响的severity及severity_reason；不能直接取成员最高等级。集中给出前置条件、完整执行顺序、输入输出如何衔接、可运行的完整脚本及命令，不把必要步骤分散到原报告；缺失条件、未验证步骤如实标明。版本冲突重新读取材料并生成，不能仅换version重试。各成员仍有独立报告、评级和证据。没有明确重复则保留独立记录。工具被关闭时不绕过配置。` } const historicalFindingReviewGuidance = ` 【历史整理模式，优先于旧报告撰写及自动绑流量指引】本次只整理已经存在的上报。原始报告即使为空或待更新，也只记录缺失情况，绝不调用 update_finding_report、report_finding、bind_finding_traffic，不修改原始标题、评级、状态、证据或报告，不执行 Bash、写文件或重新探测目标。 先逐条 get_finding_record 读取所选记录，已有原报告直接作为判断材料复用，不重复撰写。只在所选范围内检查同一实际缺陷，可按需读取已存在的证据。按上述根因准则形成明确归并/独立/疑似建议；无法确认时保存建议供人工确认。 -已有或新建文件夹必须读取全部成员并写入完整、最新的统一报告；统一报告是单独的材料，不覆盖原报告。最后必须调用 complete_finding_case_review，为每条所选 finding_id 提交 verdict（independent/merged/suggested）、reason 和 root_cause_evidence，并提供总结。根因证据应说明具体相同/不同的检查及证据来源；疑似建议明确待补证据。服务端验收通过后才可宣布完成。工具不可用或验收失败时如实报告，不能只用自然语言声称完成。` +已有或新建文件夹应读取全部成员并尽力用 update_finding_case_report 写入完整、最新的统一报告；归并不因报告暂缺而阻塞，未完成时文件夹会显示“统一报告待更新”，可由后续运行或人工补齐。统一报告是单独的材料，不覆盖原报告。最后必须调用 complete_finding_case_review，为每条所选 finding_id 提交 verdict（independent/merged/suggested）、reason 和 root_cause_evidence，并提供总结。根因证据应说明具体相同/不同的检查及证据来源；疑似建议明确待补证据。服务端验收通过后才可宣布完成。工具不可用或验收失败时如实报告，不能只用自然语言声称完成。` diff --git a/agent/finding_cases_test.go b/agent/finding_cases_test.go index a8177ea..0fc0788 100644 --- a/agent/finding_cases_test.go +++ b/agent/finding_cases_test.go @@ -46,7 +46,7 @@ func TestHistoricalFindingReviewFiltersMutationAndProbeTools(t *testing.T) { t.Fatal("historical mutation capability", tool.Name()) } } - if len(out) != 6 || def.FindingGuidance != "" || !strings.Contains(def.CaseGuidance, "历史整理模式") || !strings.Contains(def.CaseGuidance, "接口、路径、后端、读/写影响、评级不同") { + if len(out) != 6 || def.FindingGuidance != "" || !strings.Contains(def.CaseGuidance, "历史整理模式") || !strings.Contains(def.CaseGuidance, "端点/路由") { t.Fatal("missing historical isolation", def) } normal, _, cl := AugmentTools(t.Context(), "reporter", tools) diff --git a/db/finding_cases.go b/db/finding_cases.go index d7ec976..5878535 100644 --- a/db/finding_cases.go +++ b/db/finding_cases.go @@ -41,13 +41,14 @@ type FindingCaseEvent struct { CreatedAt time.Time `json:"created_at"` } type FindingCaseSuggestion struct { - ID int64 `json:"id,string"` - TaskID int64 `json:"task_id,string"` - LeftID int64 `json:"left_id,string"` - RightID int64 `json:"right_id,string"` - Title string `json:"title"` - Reason string `json:"reason"` - State string `json:"state"` + ID int64 `json:"id,string"` + TaskID int64 `json:"task_id,string"` + LeftID int64 `json:"left_id,string"` + RightID int64 `json:"right_id,string"` + Title string `json:"title"` + Reason string `json:"reason"` + State string `json:"state"` + FindingIDs []string `json:"finding_ids"` } type FindingCaseRow struct { Case *FindingCase `json:"case,omitempty"` @@ -198,9 +199,13 @@ func lockCaseTask(tx *sql.Tx, taskID int64) error { return tx.QueryRow(`SELECT id FROM tasks WHERE id=$1 FOR UPDATE`, taskID).Scan(&id) } func (d *DB) MergeFindingCase(ctx context.Context, taskID int64, ids []int64, title, reason, actor string) (int64, error) { - return d.mergeFindingCase(ctx, taskID, ids, title, reason, actor, 0) + return d.mergeFindingCase(ctx, taskID, ids, title, reason, actor, 0, nil) } -func (d *DB) mergeFindingCase(ctx context.Context, taskID int64, ids []int64, title, reason, actor string, suggestionID int64) (int64, error) { + +// mergeFindingCase merges ids into one folder. With suggestionID it also resolves +// that suggestion; dropped are suggestion members a human excluded, which are vetoed +// against the merged records so the same group is not suggested again. +func (d *DB) mergeFindingCase(ctx context.Context, taskID int64, ids []int64, title, reason, actor string, suggestionID int64, dropped []int64) (int64, error) { ids, err := sortedFindingIDs(ids) if err != nil { return 0, err @@ -308,8 +313,23 @@ func (d *DB) mergeFindingCase(ctx context.Context, taskID int64, ids []int64, ti return err } } - _, err := tx.Exec(`UPDATE finding_case_suggestions SET state='accepted' WHERE state='pending' AND left_id IN(SELECT value::bigint FROM jsonb_array_elements_text($1::jsonb)) AND right_id IN(SELECT value::bigint FROM jsonb_array_elements_text($1::jsonb))`, string(raw)) - return err + for _, x := range dropped { + if all[x] { + // Never pull an excluded record in through an existing folder. + return fmt.Errorf("排除的记录 #%d 已与所选记录在同一文件夹，请先在文件夹中移出后再确认", x) + } + for _, y := range ids { + if _, err := tx.Exec(`INSERT INTO finding_case_blocks(left_id,right_id,reason) VALUES(LEAST($1::bigint,$2::bigint),GREATEST($1::bigint,$2::bigint),'人工确认分组时排除') ON CONFLICT DO NOTHING`, x, y); err != nil { + return err + } + } + } + if suggestionID > 0 { + if _, err := tx.Exec(`UPDATE finding_case_suggestions SET state='accepted' WHERE id=$1`, suggestionID); err != nil { + return err + } + } + return sweepFindingCaseSuggestions(tx, taskID) }) return cid, err } @@ -355,8 +375,10 @@ func (d *DB) RemoveFindingCaseMember(ctx context.Context, cid, fid int64, reason if _, err := tx.Exec(`INSERT INTO finding_case_events(case_id,action,actor,reason,finding_ids) VALUES($1,'detach','human',$2,jsonb_build_array($3::bigint))`, cid, reason, fid); err != nil { return err } - _, err := tx.Exec(`DELETE FROM finding_case_members WHERE case_id=$1 AND finding_id=$2`, cid, fid) - return err + if _, err := tx.Exec(`DELETE FROM finding_case_members WHERE case_id=$1 AND finding_id=$2`, cid, fid); err != nil { + return err + } + return sweepFindingCaseSuggestions(tx, origin) }) } func (d *DB) UpdateFindingCaseReport(ctx context.Context, cid, version int64, title, report, severity, reason string) error { @@ -382,41 +404,174 @@ func (d *DB) UpdateFindingCaseReport(ctx context.Context, cid, version int64, ti return err }) } -func (d *DB) SuggestFindingCase(ctx context.Context, taskID, left, right int64, title, reason string) (int64, error) { + +// suggestionMembers returns the surviving member IDs of a suggestion in ascending order. +func suggestionMembers(q interface { + Query(string, ...any) (*sql.Rows, error) +}, id int64) ([]int64, error) { + rows, err := q.Query(`SELECT finding_id FROM finding_case_suggestion_members WHERE suggestion_id=$1 ORDER BY finding_id`, id) + if err != nil { + return nil, err + } + defer rows.Close() + out := []int64{} + for rows.Next() { + var fid int64 + if err := rows.Scan(&fid); err != nil { + return nil, err + } + out = append(out, fid) + } + return out, rows.Err() +} + +// A suggestion with a human-vetoed pair can never be accepted as a whole. +const pendingSuggestionHasBlockedPair = `EXISTS(SELECT 1 FROM finding_case_suggestion_members a JOIN finding_case_suggestion_members b ON a.suggestion_id=b.suggestion_id AND a.finding_id 20 { + return 0, errors.New("疑似重复分组最多20条记录") + } var id int64 err = d.WithEvidenceTx(ctx, func(tx *sql.Tx) error { if err := lockCaseTask(tx, taskID); err != nil { return err } + cases := map[int64]bool{} for _, fid := range ids { var tid sql.NullInt64 - if err := tx.QueryRow(`SELECT task_id FROM findings WHERE id=$1`, fid).Scan(&tid); err != nil { + var cid int64 + if err := tx.QueryRow(`SELECT f.task_id,COALESCE(m.case_id,0) FROM findings f LEFT JOIN finding_case_members m ON m.finding_id=f.id WHERE f.id=$1`, fid).Scan(&tid, &cid); err != nil { return err } if !tid.Valid || tid.Int64 != taskID { return errors.New("只能建议同一任务的归并") } + cases[cid] = true } - var blocked bool - if err := tx.QueryRow(`SELECT EXISTS(SELECT 1 FROM finding_case_blocks WHERE left_id=$1 AND right_id=$2)`, ids[0], ids[1]).Scan(&blocked); err != nil { + if len(cases) == 1 && !cases[0] { + return errors.New("这些记录已在同一文件夹中，无需建议") + } + raw, _ := json.Marshal(ids) + blocked := []string{} + rows, err := tx.Query(`SELECT left_id,right_id FROM finding_case_blocks WHERE left_id IN(SELECT value::bigint FROM jsonb_array_elements_text($1::jsonb)) AND right_id IN(SELECT value::bigint FROM jsonb_array_elements_text($1::jsonb)) ORDER BY left_id,right_id`, string(raw)) + if err != nil { return err } - if blocked { - return errors.New("人工已否决该归并") + for rows.Next() { + var l, r int64 + if err := rows.Scan(&l, &r); err != nil { + rows.Close() + return err + } + blocked = append(blocked, fmt.Sprintf("#%d/#%d", l, r)) } - return tx.QueryRow(`INSERT INTO finding_case_suggestions(task_id,left_id,right_id,title,reason) VALUES($1,$2,$3,$4,$5) ON CONFLICT(left_id,right_id) DO UPDATE SET reason=EXCLUDED.reason WHERE finding_case_suggestions.state='pending' RETURNING id`, taskID, ids[0], ids[1], title, reason).Scan(&id) + err = rows.Err() + rows.Close() + if err != nil { + return err + } + if len(blocked) > 0 { + return fmt.Errorf("人工已否决这些记录的归并：%s；请去掉其中一方后重新提交", strings.Join(blocked, "、")) + } + // Compare against every pending group of the task; read all rows before issuing more statements. + type pending struct { + id int64 + members map[int64]bool + } + groups := []pending{} + rows, err = tx.Query(`SELECT s.id,sm.finding_id FROM finding_case_suggestions s JOIN finding_case_suggestion_members sm ON sm.suggestion_id=s.id WHERE s.task_id=$1 AND s.state='pending' ORDER BY s.id`, taskID) + if err != nil { + return err + } + for rows.Next() { + var sid, fid int64 + if err := rows.Scan(&sid, &fid); err != nil { + rows.Close() + return err + } + if len(groups) == 0 || groups[len(groups)-1].id != sid { + groups = append(groups, pending{id: sid, members: map[int64]bool{}}) + } + groups[len(groups)-1].members[fid] = true + } + err = rows.Err() + rows.Close() + if err != nil { + return err + } + want := map[int64]bool{} + for _, fid := range ids { + want[fid] = true + } + superseded := []int64{} + for _, g := range groups { + inNew := 0 + for fid := range g.members { + if want[fid] { + inNew++ + } + } + switch { + case inNew == len(ids) && len(g.members) == len(ids): + id = g.id + _, err := tx.Exec(`UPDATE finding_case_suggestions SET title=$2,reason=$3 WHERE id=$1`, g.id, title, reason) + return err + case inNew == len(ids): + // An existing pending group already covers every requested record. + id = g.id + return nil + case inNew == len(g.members): + superseded = append(superseded, g.id) + } + } + for _, sid := range superseded { + if _, err := tx.Exec(`UPDATE finding_case_suggestions SET state='superseded' WHERE id=$1 AND state='pending'`, sid); err != nil { + return err + } + } + if err := tx.QueryRow(`INSERT INTO finding_case_suggestions(task_id,left_id,right_id,title,reason) VALUES($1,$2,$3,$4,$5) RETURNING id`, taskID, ids[0], ids[1], title, reason).Scan(&id); err != nil { + return err + } + _, err = tx.Exec(`INSERT INTO finding_case_suggestion_members(suggestion_id,finding_id) SELECT $1,value::bigint FROM jsonb_array_elements_text($2::jsonb)`, id, string(raw)) + return err }) return id, err } + +// FindingCaseSuggestions lists pending groups with at least two surviving, +// non-vetoed members. func (d *DB) FindingCaseSuggestions(taskID int64) ([]FindingCaseSuggestion, error) { - rows, err := d.Query(`SELECT id,task_id,left_id,right_id,title,reason,state FROM finding_case_suggestions WHERE state='pending' AND ($1::bigint=0 OR task_id=$1) ORDER BY id DESC`, taskID) + rows, err := d.Query(`SELECT s.id,s.task_id,s.left_id,s.right_id,s.title,s.reason,s.state,COALESCE((SELECT jsonb_agg(sm.finding_id ORDER BY sm.finding_id) FROM finding_case_suggestion_members sm WHERE sm.suggestion_id=s.id),'[]'::jsonb) + FROM finding_case_suggestions s WHERE s.state='pending' AND ($1::bigint=0 OR s.task_id=$1) AND NOT `+pendingSuggestionHasBlockedPair+` + AND (SELECT count(*) FROM finding_case_suggestion_members sm WHERE sm.suggestion_id=s.id)>=2 + AND NOT EXISTS(SELECT 1 FROM finding_case_suggestion_members sm JOIN findings f ON f.id=sm.finding_id WHERE sm.suggestion_id=s.id AND f.task_id IS DISTINCT FROM s.task_id) + ORDER BY s.id DESC`, taskID) if err != nil { return nil, err } @@ -424,14 +579,27 @@ func (d *DB) FindingCaseSuggestions(taskID int64) ([]FindingCaseSuggestion, erro out := []FindingCaseSuggestion{} for rows.Next() { var s FindingCaseSuggestion - if err := rows.Scan(&s.ID, &s.TaskID, &s.LeftID, &s.RightID, &s.Title, &s.Reason, &s.State); err != nil { + var raw []byte + if err := rows.Scan(&s.ID, &s.TaskID, &s.LeftID, &s.RightID, &s.Title, &s.Reason, &s.State, &raw); err != nil { return nil, err } + var members []int64 + if err := json.Unmarshal(raw, &members); err != nil { + return nil, err + } + s.FindingIDs = make([]string, 0, len(members)) + for _, fid := range members { + s.FindingIDs = append(s.FindingIDs, strconv.FormatInt(fid, 10)) + } out = append(out, s) } return out, rows.Err() } -func (d *DB) ResolveFindingCaseSuggestion(ctx context.Context, id int64, accept bool) (int64, error) { + +// ResolveFindingCaseSuggestion confirms or rejects a whole group. On accept the +// remaining members (minus exclude) are merged once; excluded records are vetoed +// against the remaining members. Rejecting vetoes every pair inside the group. +func (d *DB) ResolveFindingCaseSuggestion(ctx context.Context, id int64, accept bool, exclude []int64) (int64, error) { var s FindingCaseSuggestion if err := d.QueryRow(`SELECT task_id,left_id,right_id,title,reason,state FROM finding_case_suggestions WHERE id=$1`, id).Scan(&s.TaskID, &s.LeftID, &s.RightID, &s.Title, &s.Reason, &s.State); err != nil { return 0, err @@ -439,8 +607,35 @@ func (d *DB) ResolveFindingCaseSuggestion(ctx context.Context, id int64, accept if s.State != "pending" { return 0, ErrFindingCaseConflict } + members, err := suggestionMembers(d, id) + if err != nil { + return 0, err + } + hasLeft, hasRight := false, false + for _, x := range members { + hasLeft = hasLeft || x == s.LeftID + hasRight = hasRight || x == s.RightID + } + if len(members) < 2 || !hasLeft || !hasRight { + return 0, ErrFindingCaseConflict + } if accept { - return d.mergeFindingCase(ctx, s.TaskID, []int64{s.LeftID, s.RightID}, s.Title, s.Reason, "human", id) + ex := map[int64]bool{} + for _, x := range exclude { + ex[x] = true + } + keep, dropped := []int64{}, []int64{} + for _, x := range members { + if ex[x] { + dropped = append(dropped, x) + } else { + keep = append(keep, x) + } + } + if len(keep) < 2 { + return 0, errors.New("至少保留两条记录才能确认归并") + } + return d.mergeFindingCase(ctx, s.TaskID, keep, s.Title, s.Reason, "human", id, dropped) } return 0, d.WithEvidenceTx(ctx, func(tx *sql.Tx) error { if err := lockCaseTask(tx, s.TaskID); err != nil { @@ -453,10 +648,19 @@ func (d *DB) ResolveFindingCaseSuggestion(ctx context.Context, id int64, accept if n, _ := res.RowsAffected(); n != 1 { return ErrFindingCaseConflict } - if _, err := tx.Exec(`INSERT INTO finding_case_blocks(left_id,right_id,reason) VALUES($1,$2,'人工否决') ON CONFLICT DO NOTHING`, s.LeftID, s.RightID); err != nil { + // Re-read members under the task lock so the veto matches what was rejected. + current, err := suggestionMembers(tx, id) + if err != nil { return err } - return nil + for i := range current { + for j := i + 1; j < len(current); j++ { + if _, err := tx.Exec(`INSERT INTO finding_case_blocks(left_id,right_id,reason) VALUES($1,$2,'人工否决') ON CONFLICT DO NOTHING`, current[i], current[j]); err != nil { + return err + } + } + } + return sweepFindingCaseSuggestions(tx, s.TaskID) }) } @@ -479,13 +683,16 @@ func (d *DB) DistinctFindingStats(taskID string) (*FindingDistinctStats, error) } // Compact candidate pages exclude PoC/report bodies and do not treat severity as identity. -func (d *DB) FindingCaseCandidates(fid int64, page, size int, scope ...[]int64) ([]map[string]any, int, error) { +// Recall stays broad (type OR name OR shared asset) so nothing is dropped, but rows are +// ranked: same vulnclass AND shared asset is "strong"; a single-condition hit is "weak". +// strong is the total number of strong candidates across all pages. +func (d *DB) FindingCaseCandidates(fid int64, page, size int, scope ...[]int64) (items []map[string]any, total int, strong int, err error) { f, err := d.FindingCaseSummary(fid) if err != nil { - return nil, 0, err + return nil, 0, 0, err } if f == nil || f.TaskID == nil { - return nil, 0, errors.New("漏洞任务不可用") + return nil, 0, 0, errors.New("漏洞任务不可用") } if page < 1 { page = 1 @@ -497,40 +704,62 @@ func (d *DB) FindingCaseCandidates(fid int64, page, size int, scope ...[]int64) if f.AssetIDs == nil { assets = []byte("[]") } - // Type/asset/name only retrieve candidates; the Reporter must still prove the root cause. - where := ` WHERE f.task_id=$1 AND f.id<>$2 AND NOT EXISTS(SELECT 1 FROM finding_case_blocks b WHERE b.left_id=LEAST(f.id,$2::bigint) AND b.right_id=GREATEST(f.id,$2::bigint)) - AND ((trim($3::text)<>'' AND lower(f.vulnclass)=lower($3)) OR (trim($4::text)<>'' AND lower(f.name)=lower($4)) - OR EXISTS(SELECT 1 FROM jsonb_array_elements(COALESCE(NULLIF(f.asset_ids,'null'::jsonb),'[]'::jsonb)) a WHERE a.value IN(SELECT value FROM jsonb_array_elements($5::jsonb))))` + // Type/asset/name only retrieve and rank candidates; the Reporter must still prove the root cause. + hits := `SELECT f.*, + (trim($3::text)<>'' AND lower(f.vulnclass)=lower($3)) AS same_class, + (trim($4::text)<>'' AND lower(f.name)=lower($4)) AS same_name, + EXISTS(SELECT 1 FROM jsonb_array_elements(CASE WHEN jsonb_typeof(f.asset_ids)='array' THEN f.asset_ids ELSE '[]'::jsonb END) a WHERE a.value IN(SELECT value FROM jsonb_array_elements($5::jsonb))) AS shared_asset + FROM findings f WHERE f.task_id=$1 AND f.id<>$2 AND NOT EXISTS(SELECT 1 FROM finding_case_blocks b WHERE b.left_id=LEAST(f.id,$2::bigint) AND b.right_id=GREATEST(f.id,$2::bigint))` args := []any{*f.TaskID, fid, f.VulnClass, f.Name, string(assets)} if len(scope) > 0 { raw, err := json.Marshal(scope[0]) if err != nil { - return nil, 0, err + return nil, 0, 0, err } - where += ` AND f.id IN (SELECT jsonb_array_elements_text($6::jsonb)::bigint)` + hits += ` AND f.id IN (SELECT jsonb_array_elements_text($6::jsonb)::bigint)` args = append(args, string(raw)) } - var total int - if err := d.QueryRow(`SELECT count(*) FROM findings f`+where, args...).Scan(&total); err != nil { - return nil, 0, err + base := `WITH hits AS (` + hits + `), cand AS (SELECT h.*, + (CASE WHEN same_class THEN 2 ELSE 0 END + CASE WHEN shared_asset THEN 2 ELSE 0 END + CASE WHEN same_name THEN 1 ELSE 0 END + + CASE WHEN same_class AND shared_asset THEN 3 ELSE 0 END) AS score + FROM hits h WHERE same_class OR same_name OR shared_asset)` + if err := d.QueryRow(base+` SELECT count(*),count(*) FILTER(WHERE same_class AND shared_asset) FROM cand`, args...).Scan(&total, &strong); err != nil { + return nil, 0, 0, err } args = append(args, size, (page-1)*size) - rows, err := d.Query(`SELECT f.id,f.name,f.vulnclass,f.severity,left(f.summary,400),f.asset_ids,COALESCE(m.case_id,0) FROM findings f LEFT JOIN finding_case_members m ON m.finding_id=f.id`+where+fmt.Sprintf(` ORDER BY f.id DESC LIMIT $%d OFFSET $%d`, len(args)-1, len(args)), args...) + rows, err := d.Query(base+fmt.Sprintf(` SELECT c.id,c.name,c.vulnclass,c.severity,left(c.summary,400),c.asset_ids,COALESCE(m.case_id,0),c.score,c.same_class,c.shared_asset,c.same_name + FROM cand c LEFT JOIN finding_case_members m ON m.finding_id=c.id ORDER BY c.score DESC,c.id DESC LIMIT $%d OFFSET $%d`, len(args)-1, len(args)), args...) if err != nil { - return nil, 0, err + return nil, 0, 0, err } defer rows.Close() - out := []map[string]any{} + items = []map[string]any{} for rows.Next() { var id, cid int64 + var score int + var sameClass, sharedAsset, sameName bool var name, class, severity, summary string var assetIDs json.RawMessage - if err := rows.Scan(&id, &name, &class, &severity, &summary, &assetIDs, &cid); err != nil { - return nil, 0, err + if err := rows.Scan(&id, &name, &class, &severity, &summary, &assetIDs, &cid, &score, &sameClass, &sharedAsset, &sameName); err != nil { + return nil, 0, 0, err } - out = append(out, map[string]any{"finding_id": strconv.FormatInt(id, 10), "name": name, "vulnclass": class, "severity": severity, "summary": summary, "asset_ids": assetIDs, "case_id": strconv.FormatInt(cid, 10)}) + reasons := []string{} + if sameClass { + reasons = append(reasons, "同漏洞类型") + } + if sharedAsset { + reasons = append(reasons, "资产重叠") + } + if sameName { + reasons = append(reasons, "同名称") + } + tier := "weak" + if sameClass && sharedAsset { + tier = "strong" + } + items = append(items, map[string]any{"finding_id": strconv.FormatInt(id, 10), "name": name, "vulnclass": class, "severity": severity, "summary": summary, "asset_ids": assetIDs, "case_id": strconv.FormatInt(cid, 10), "score": score, "tier": tier, "hit_reasons": reasons}) } - return out, total, rows.Err() + return items, total, strong, rows.Err() } // Batch task-list counters avoid an additional database round trip per task. diff --git a/db/finding_cases_test.go b/db/finding_cases_test.go index 48398a5..a50cab4 100644 --- a/db/finding_cases_test.go +++ b/db/finding_cases_test.go @@ -118,11 +118,11 @@ func TestFindingCaseScopeSuggestionsAndDeletion(t *testing.T) { if _, err := d.MergeFindingCase(ctx, task.ID, []int64{ids[0], fid}, "x", "x", "reporter"); err == nil { t.Fatal("cross task accepted") } - sid, err := d.SuggestFindingCase(ctx, task.ID, ids[0], ids[1], "order", "uncertain") + sid, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[0], ids[1]}, "order", "uncertain") if err != nil { t.Fatal(err) } - if _, err := d.ResolveFindingCaseSuggestion(ctx, sid, false); err != nil { + if _, err := d.ResolveFindingCaseSuggestion(ctx, sid, false, nil); err != nil { t.Fatal(err) } if _, err := d.MergeFindingCase(ctx, task.ID, ids[:2], "x", "x", "reporter"); err == nil { @@ -233,7 +233,7 @@ func TestFindingCaseTaskDeletionAndRepeatedDetach(t *testing.T) { func TestFindingCaseSuggestionResolutionIsAtomic(t *testing.T) { d, task, ids := caseTestDB(t) ctx := context.Background() - sid, err := d.SuggestFindingCase(ctx, task.ID, ids[0], ids[1], "orders", "uncertain") + sid, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[0], ids[1]}, "orders", "uncertain") if err != nil { t.Fatal(err) } @@ -241,7 +241,7 @@ func TestFindingCaseSuggestionResolutionIsAtomic(t *testing.T) { var wg sync.WaitGroup for _, accept := range []bool{true, false} { wg.Add(1) - go func(a bool) { defer wg.Done(); _, e := d.ResolveFindingCaseSuggestion(ctx, sid, a); results <- e }(accept) + go func(a bool) { defer wg.Done(); _, e := d.ResolveFindingCaseSuggestion(ctx, sid, a, nil); results <- e }(accept) } wg.Wait() close(results) @@ -269,7 +269,7 @@ func TestFindingCaseCandidatesAreCompactAndRelated(t *testing.T) { if _, err := d.Exec(`UPDATE findings SET vulnclass='SQLi',name='search injection' WHERE id=$1`, ids[3]); err != nil { t.Fatal(err) } - items, total, err := d.FindingCaseCandidates(ids[0], 1, 1) + items, total, _, err := d.FindingCaseCandidates(ids[0], 1, 1) if err != nil || total != 2 || len(items) != 1 { t.Fatalf("candidate pagination %d %+v %v", total, items, err) } @@ -286,7 +286,7 @@ func TestFindingCaseCandidatesAreCompactAndRelated(t *testing.T) { if err := d.RemoveFindingCaseMember(context.Background(), cid, ids[1], "different cause"); err != nil { t.Fatal(err) } - _, total, err = d.FindingCaseCandidates(ids[0], 1, 20) + _, total, _, err = d.FindingCaseCandidates(ids[0], 1, 20) if err != nil || total != 1 { t.Fatalf("veto excluded %d %v", total, err) } @@ -355,3 +355,150 @@ func TestFindingCaseViewContextAndAssetFiltering(t *testing.T) { t.Fatalf("original severity order %+v %v", rows, err) } } + +func suggestionState(t *testing.T, d *DB, id int64) string { + t.Helper() + var state string + if err := d.QueryRow(`SELECT state FROM finding_case_suggestions WHERE id=$1`, id).Scan(&state); err != nil { + t.Fatal(err) + } + return state +} + +func TestFindingCaseClusterSuggestionLifecycle(t *testing.T) { + d, task, ids := caseTestDB(t) + ctx := context.Background() + pair, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[1], ids[0]}, "orders", "same handler, owner check unverified") + if err != nil { + t.Fatal(err) + } + group, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[0], ids[1], ids[2], ids[3]}, "orders", "same handler for all four") + if err != nil { + t.Fatal(err) + } + if group == pair || suggestionState(t, d, pair) != "superseded" { + t.Fatalf("subset pair not superseded: %d %d %s", pair, group, suggestionState(t, d, pair)) + } + // Re-suggesting a subset of a pending group reuses the group instead of adding a card. + again, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[2], ids[3]}, "orders", "dup") + if err != nil || again != group { + t.Fatalf("subset reuse %d %v", again, err) + } + list, err := d.FindingCaseSuggestions(task.ID) + if err != nil || len(list) != 1 || !reflect.DeepEqual(list[0].FindingIDs, []string{fmt.Sprint(ids[0]), fmt.Sprint(ids[1]), fmt.Sprint(ids[2]), fmt.Sprint(ids[3])}) { + t.Fatalf("grouped list %+v %v", list, err) + } + if _, err := d.ResolveFindingCaseSuggestion(ctx, group, true, []int64{ids[1], ids[2], ids[3]}); err == nil { + t.Fatal("accepted with fewer than two records") + } + cid, err := d.ResolveFindingCaseSuggestion(ctx, group, true, []int64{ids[3]}) + if err != nil || cid == 0 { + t.Fatal(cid, err) + } + c, err := d.GetFindingCase(cid) + if err != nil || c.Count != 3 { + t.Fatalf("one merge of remaining members %+v %v", c, err) + } + if other, _ := d.FindingCaseID(ids[3]); other != 0 { + t.Fatal("excluded record merged") + } + if suggestionState(t, d, group) != "accepted" { + t.Fatal("group not accepted") + } + var events int + if err := d.QueryRow(`SELECT count(*) FROM finding_case_events WHERE case_id=$1 AND action='merge'`, cid).Scan(&events); err != nil || events != 1 { + t.Fatalf("merge events %d %v", events, err) + } + if _, err := d.MergeFindingCase(ctx, task.ID, []int64{ids[0], ids[3]}, "x", "x", "reporter"); err == nil { + t.Fatal("human exclusion not kept as veto") + } + if _, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[2], ids[3]}, "x", "x"); err == nil { + t.Fatal("vetoed pair suggested again") + } + if _, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[0], ids[1]}, "x", "x"); err == nil { + t.Fatal("records already in one folder suggested") + } + if list, _ := d.FindingCaseSuggestions(task.ID); len(list) != 0 { + t.Fatalf("stale cards remain %+v", list) + } +} + +func TestFindingCaseClusterRejectVetoesAllPairsAndMergeClearsCards(t *testing.T) { + d, task, ids := caseTestDB(t) + ctx := context.Background() + a, err := d.SuggestFindingCase(ctx, task.ID, ids[:3], "orders", "shared component") + if err != nil { + t.Fatal(err) + } + if _, err := d.ResolveFindingCaseSuggestion(ctx, a, false, nil); err != nil { + t.Fatal(err) + } + var blocks int + if err := d.QueryRow(`SELECT count(*) FROM finding_case_blocks WHERE left_id=ANY(ARRAY[$1,$2,$3]::bigint[]) AND right_id=ANY(ARRAY[$1,$2,$3]::bigint[])`, ids[0], ids[1], ids[2]).Scan(&blocks); err != nil || blocks != 3 { + t.Fatalf("blocks %d %v", blocks, err) + } + b, err := d.SuggestFindingCase(ctx, task.ID, []int64{ids[2], ids[3]}, "orders", "same param") + if err != nil { + t.Fatal(err) + } + // A direct reporter merge covering a pending group clears the card. + if _, err := d.MergeFindingCase(ctx, task.ID, []int64{ids[2], ids[3]}, "orders", "proof", "reporter"); err != nil { + t.Fatal(err) + } + if suggestionState(t, d, b) != "accepted" { + t.Fatal("covered suggestion still pending") + } +} + +func TestFindingCaseCandidatesRankStrongBeforeWeak(t *testing.T) { + d, task, _ := caseTestDB(t) + add := func(class, name string, assets []int64) int64 { + id, err := d.AddFinding(task.ID, 0, class, name, "high", "s", "e", "worker", assets) + if err != nil { + t.Fatal(err) + } + return id + } + base := add("SSRF", "fetch proxy", []int64{900001}) + strong := add("SSRF", "image import", []int64{900001}) + assetOnly := add("XSS", "comment", []int64{900001}) + nameOnly := add("RCE", "fetch proxy", nil) + items, total, strongTotal, err := d.FindingCaseCandidates(base, 1, 20) + if err != nil { + t.Fatal(err) + } + if total != 3 || strongTotal != 1 || len(items) != 3 { + t.Fatalf("recall kept %d strong %d %+v", total, strongTotal, items) + } + if items[0]["finding_id"] != fmt.Sprint(strong) || items[0]["tier"] != "strong" || !reflect.DeepEqual(items[0]["hit_reasons"], []string{"同漏洞类型", "资产重叠"}) { + t.Fatalf("strong first %+v", items[0]) + } + if items[1]["finding_id"] != fmt.Sprint(assetOnly) || items[2]["finding_id"] != fmt.Sprint(nameOnly) || items[2]["tier"] != "weak" { + t.Fatalf("weak ordering %+v", items) + } + if items[0]["score"].(int) <= items[1]["score"].(int) { + t.Fatalf("scores %+v", items) + } + page2, total2, _, err := d.FindingCaseCandidates(base, 2, 2) + if err != nil || total2 != 3 || len(page2) != 1 || page2[0]["finding_id"] != fmt.Sprint(nameOnly) { + t.Fatalf("low score candidates stay paginated %+v %v", page2, err) + } +} + +func TestFindingCaseClusterExclusionCannotBePulledInByFolder(t *testing.T) { + d, task, ids := caseTestDB(t) + ctx := context.Background() + if _, err := d.MergeFindingCase(ctx, task.ID, ids[:2], "orders", "proof", "reporter"); err != nil { + t.Fatal(err) + } + sid, err := d.SuggestFindingCase(ctx, task.ID, ids[:3], "orders", "shared handler") + if err != nil { + t.Fatal(err) + } + if _, err := d.ResolveFindingCaseSuggestion(ctx, sid, true, []int64{ids[1]}); err == nil { + t.Fatal("excluded member merged through its existing folder") + } + if cid, _ := d.FindingCaseID(ids[2]); cid != 0 || suggestionState(t, d, sid) != "pending" { + t.Fatal("failed exclusion was not rolled back") + } +} diff --git a/db/schema.sql b/db/schema.sql index 1fe2311..44d7c4f 100644 --- a/db/schema.sql +++ b/db/schema.sql @@ -1488,6 +1488,17 @@ CREATE TABLE IF NOT EXISTS finding_case_suggestions ( UNIQUE(left_id,right_id), CHECK(left_id < right_id) ); +ALTER TABLE finding_case_suggestions DROP CONSTRAINT IF EXISTS finding_case_suggestions_left_id_right_id_key; +CREATE TABLE IF NOT EXISTS finding_case_suggestion_members ( + suggestion_id BIGINT NOT NULL REFERENCES finding_case_suggestions(id) ON DELETE CASCADE, + finding_id BIGINT NOT NULL REFERENCES findings(id) ON DELETE CASCADE, + PRIMARY KEY(suggestion_id,finding_id) +); +CREATE INDEX IF NOT EXISTS finding_case_suggestion_members_finding_idx ON finding_case_suggestion_members(finding_id); +INSERT INTO finding_case_suggestion_members(suggestion_id,finding_id) + SELECT id,left_id FROM finding_case_suggestions ON CONFLICT DO NOTHING; +INSERT INTO finding_case_suggestion_members(suggestion_id,finding_id) + SELECT id,right_id FROM finding_case_suggestions ON CONFLICT DO NOTHING; -- Rejected/detached pairs are a durable human veto, also checked on group unions. CREATE TABLE IF NOT EXISTS finding_case_blocks ( left_id BIGINT NOT NULL REFERENCES findings(id) ON DELETE CASCADE, diff --git a/server/finding_case_reviews.go b/server/finding_case_reviews.go index 16a9936..8303a20 100644 --- a/server/finding_case_reviews.go +++ b/server/finding_case_reviews.go @@ -35,7 +35,7 @@ func (s *Server) toolCompleteFindingCaseReview() actool.CoreTool { "verdict": map[string]any{"type": "string", "enum": []string{"independent", "merged", "suggested"}}, "reason": strParam("判断理由，接口或读写影响不同不能单独证明不同根因"), "root_cause_evidence": strParam("原始报告/证据中支持实际共同或不同根因的具体材料；不确定则说明缺失证据及已保存建议")}, "finding_id", "verdict", "reason", "root_cause_evidence") - return wrTool("complete_finding_case_review", "历史整理结束前提交每条所选记录的根因判断并验收：必须已读取全部记录，疑似必须已有建议，归并必须已有完整最新统一报告。验收失败不能宣布完成。", objSchema(map[string]any{"summary": strParam("整理结果总结"), "decisions": map[string]any{"type": "array", "items": decision}}, "summary", "decisions"), func(ctx context.Context, raw json.RawMessage) (actool.Result, error) { + return wrTool("complete_finding_case_review", "历史整理结束前提交每条所选记录的根因判断并验收：必须已读取全部记录，疑似必须已有建议；归并只需已在活动文件夹中，统一报告可稍后补齐。", objSchema(map[string]any{"summary": strParam("整理结果总结"), "decisions": map[string]any{"type": "array", "items": decision}}, "summary", "decisions"), func(ctx context.Context, raw json.RawMessage) (actool.Result, error) { var a findingReviewConclusion if err := json.Unmarshal(raw, &a); err != nil { return actool.Errorf(err.Error()), nil @@ -101,7 +101,7 @@ func (s *Server) saveFindingReviewConclusion(ctx context.Context, conv int64, a } case "suggested": var pending bool - if err := tx.QueryRow(`SELECT EXISTS(SELECT 1 FROM finding_case_suggestions WHERE state='pending' AND (left_id=$1 OR right_id=$1))`, id).Scan(&pending); err != nil { + if err := tx.QueryRow(`SELECT EXISTS(SELECT 1 FROM finding_case_suggestions s JOIN finding_case_suggestion_members m ON m.suggestion_id=s.id WHERE s.state='pending' AND m.finding_id=$1)`, id).Scan(&pending); err != nil { return err } if !pending { @@ -115,17 +115,12 @@ func (s *Server) saveFindingReviewConclusion(ctx context.Context, conv int64, a return errors.New("无效判断类型") } if cid.Valid { - var report string - var version, rv int64 var active bool - if err := tx.QueryRow(`SELECT report,version,report_version,active FROM finding_cases WHERE id=$1 FOR SHARE`, cid.Int64).Scan(&report, &version, &rv, &active); err != nil { + if err := tx.QueryRow(`SELECT active FROM finding_cases WHERE id=$1 FOR SHARE`, cid.Int64).Scan(&active); err != nil { return err } - if !active || rv != version { - return fmt.Errorf("文件夹 #%d 的统一报告待更新，请重新读取后更新", cid.Int64) - } - if err := validateUnifiedFindingReport(report); err != nil { - return err + if !active { + return fmt.Errorf("文件夹 #%d 已关闭", cid.Int64) } } } diff --git a/server/finding_case_reviews_test.go b/server/finding_case_reviews_test.go index 06e1b58..589aee0 100644 --- a/server/finding_case_reviews_test.go +++ b/server/finding_case_reviews_test.go @@ -43,7 +43,7 @@ func TestUnifiedFindingReportRejectsIncompleteAndTruncated(t *testing.T) { } } -func TestFindingCaseReviewCompletionRequiresReadsAndFreshReport(t *testing.T) { +func TestFindingCaseReviewCompletionRequiresReadsNotUnifiedReport(t *testing.T) { s, task, ids := testCaseServer(t) runs, err := s.m.pg.CreateFindingCaseReviews(t.Context(), map[int64][]int64{task: ids}) if err != nil { @@ -71,15 +71,9 @@ func TestFindingCaseReviewCompletionRequiresReadsAndFreshReport(t *testing.T) { if err != nil { t.Fatal(err) } - if err := s.saveFindingReviewConclusion(ctx, conv, a); err == nil { - t.Fatal("missing unified report accepted") - } - c, _ := s.m.pg.GetFindingCase(cid) - if err := s.m.pg.UpdateFindingCaseReport(ctx, cid, c.Version, "orders", completeFindingReviewReport, "high", "verified impact"); err != nil { - t.Fatal(err) - } + // A merged verdict no longer waits for the unified report; the folder shows it as pending. if err := s.saveFindingReviewConclusion(ctx, conv, a); err != nil { - t.Fatal(err) + t.Fatalf("merge blocked by missing unified report: %v", err) } if err := s.validateFindingReviewCompletion(conv); err != nil { t.Fatal(err) @@ -90,8 +84,14 @@ func TestFindingCaseReviewCompletionRequiresReadsAndFreshReport(t *testing.T) { if _, err := s.m.pg.Exec(`UPDATE findings SET evidence='new proof' WHERE id=$1`, ids[0]); err != nil { t.Fatal(err) } + if err := s.validateFindingReviewCompletion(conv); err != nil { + t.Fatalf("stale unified report should not fail a merged review: %v", err) + } + if err := s.m.pg.RemoveFindingCaseMember(ctx, cid, ids[0], "different cause"); err != nil { + t.Fatal(err) + } if err := s.validateFindingReviewCompletion(conv); err == nil { - t.Fatal("stale group reported completed") + t.Fatal("merged verdict accepted after the record left the folder") } } @@ -138,7 +138,7 @@ func TestFindingCaseReviewSuggestionNeedsSavedSuggestion(t *testing.T) { if err := s.saveFindingReviewConclusion(ctx, conv, a); err == nil { t.Fatal("unpersisted suggestion accepted") } - if _, err := s.m.pg.SuggestFindingCase(ctx, task, ids[0], ids[1], "orders", "common component unknown"); err != nil { + if _, err := s.m.pg.SuggestFindingCase(ctx, task, []int64{ids[0], ids[1]}, "orders", "common component unknown"); err != nil { t.Fatal(err) } if err := s.saveFindingReviewConclusion(ctx, conv, a); err != nil { diff --git a/server/finding_cases.go b/server/finding_cases.go index d8afaf8..bc08b0f 100644 --- a/server/finding_cases.go +++ b/server/finding_cases.go @@ -7,6 +7,7 @@ import ( "errors" "fmt" "net/http" + "strconv" "strings" "github.com/Autumn-27/artex/db" @@ -203,12 +204,22 @@ func (s *Server) caseSuggestions(w http.ResponseWriter, r *http.Request) { } func (s *Server) resolveCaseSuggestion(w http.ResponseWriter, r *http.Request) { var a struct { - Accept bool `json:"accept"` + Accept bool `json:"accept"` + ExcludeIDs []string `json:"exclude_ids"` } if !decodeCaseRequest(w, r, &a) { return } - id, err := s.m.pg.ResolveFindingCaseSuggestion(r.Context(), int64(atoiDefault(r.PathValue("id"), 0)), a.Accept) + exclude := make([]int64, 0, len(a.ExcludeIDs)) + for _, v := range a.ExcludeIDs { + id, err := strconv.ParseInt(strings.TrimSpace(v), 10, 64) + if err != nil || id <= 0 { + caseHTTPError(w, errors.New("无效的排除记录编号")) + return + } + exclude = append(exclude, id) + } + id, err := s.m.pg.ResolveFindingCaseSuggestion(r.Context(), int64(atoiDefault(r.PathValue("id"), 0)), a.Accept, exclude) if err != nil { caseHTTPError(w, err) return @@ -253,10 +264,10 @@ func (s *Server) findingCaseTools() []actool.CoreTool { name, desc string required []string }{ - {"search_finding_duplicates", "只读检索同任务候选，返回摘要与资产；等级不同也可能重复。支持分页，按需读取详情。", []string{"finding_id"}}, + {"search_finding_duplicates", "只读检索同任务候选摘要与资产，按分数排序并返回 score、tier、hit_reasons：strong=同漏洞类型且资产重叠，weak=单一条件命中仅作参考。等级不同也可能重复。支持分页，按需读取详情。", []string{"finding_id"}}, {"get_finding_record", "读取独立漏洞完整原始证据和报告，返回探索节点编号以便追溯执行记录。", []string{"finding_id"}}, {"merge_finding_records", "证据明确证明同一缺陷时归并；保留所有原始记录。不跨任务、不改变原评级，不可绕过人工否决。", []string{"task_id", "finding_ids", "title", "reason"}}, - {"suggest_finding_merge", "证据不足时提交疑似重复建议，待人工确认；不是自动合并。只允许两条同任务记录。", []string{"task_id", "finding_ids", "title", "reason"}}, + {"suggest_finding_merge", "证据存在具体共同点但缺少关键证据时，提交一个疑似重复分组（2..N 条同任务记录，一次覆盖全部疑似成员，不要拆成两两建议），待人工确认；不是自动合并。仅类型/资产相同不应提交。", []string{"task_id", "finding_ids", "title", "reason"}}, {"get_finding_case", "读取归并组和成员分页。先记录version，再读取成员的原报告、证据及流量；版本变化必须重新读取。", []string{"case_id"}}, {"update_finding_case_report", "保存统一报告及评级理由，version必须来自实际读取。保留各条原报告，不新建漏洞。", []string{"case_id", "version", "title", "report", "severity", "severity_reason"}}, } @@ -312,8 +323,9 @@ func (s *Server) performFindingCaseTool(ctx context.Context, name string, a find if scope != nil { scopes = append(scopes, scope) } - items, total, err := s.m.pg.FindingCaseCandidates(fid, page, 20, scopes...) - return map[string]any{"items": items, "total": total, "page": page, "page_size": 20}, err + items, total, strong, err := s.m.pg.FindingCaseCandidates(fid, page, 20, scopes...) + return map[string]any{"items": items, "total": total, "strong_total": strong, "page": page, "page_size": 20, + "note": "按分数排序：strong=同漏洞类型且资产重叠，优先核对；weak=单一条件命中，仅作参考，没有具体共同点时不应据此建议归并"}, err case "get_finding_record": f, err := s.m.pg.GetFinding(fid) if err != nil { @@ -341,11 +353,11 @@ func (s *Server) performFindingCaseTool(ctx context.Context, name string, a find return map[string]string{"case_id": i64s(id)}, err case "suggest_finding_merge": ids := a.ids() - if len(ids) != 2 { - return nil, errors.New("疑似重复建议需要两条记录") + if len(ids) < 2 || len(ids) > 20 { + return nil, errors.New("疑似重复分组需要2至20条记录") } - id, err := s.m.pg.SuggestFindingCase(ctx, parseProfileID(a.TaskID), ids[0], ids[1], a.Title, a.Reason) - return map[string]string{"suggestion_id": i64s(id)}, err + id, err := s.m.pg.SuggestFindingCase(ctx, parseProfileID(a.TaskID), ids, a.Title, a.Reason) + return map[string]any{"suggestion_id": i64s(id), "finding_ids": ids}, err case "get_finding_case": c, err := s.m.pg.GetFindingCase(cid) if err != nil { diff --git a/web/src/components/finding-case-list.tsx b/web/src/components/finding-case-list.tsx index 105ce1d..ff5508c 100644 --- a/web/src/components/finding-case-list.tsx +++ b/web/src/components/finding-case-list.tsx @@ -192,15 +192,18 @@ function FindingMergeSuggestionCard({ onResolve, }: { suggestion: FindingCaseSuggestion; - onResolve: (s: FindingCaseSuggestion, accept: boolean) => Promise; + onResolve: (s: FindingCaseSuggestion, accept: boolean, excludeIds?: string[]) => Promise; }) { const [expanded, setExpanded] = React.useState(false); const [busy, setBusy] = React.useState(false); + const ids = s.finding_ids?.length ? s.finding_ids : [s.left_id, s.right_id]; + const [excluded, setExcluded] = React.useState([]); + const included = ids.filter((id) => !excluded.includes(id)); async function resolve(accept: boolean) { if (busy) return; setBusy(true); try { - await onResolve(s, accept); + await onResolve(s, accept, accept ? excluded : []); } finally { setBusy(false); } @@ -209,17 +212,37 @@ function FindingMergeSuggestionCard({
- 疑似重复 · 待确认 + 疑似重复分组 · {ids.length} 条 · 待确认 任务 #{s.task_id}
{s.title} - - - 查看原报告 #{s.left_id} - - - 查看原报告 #{s.right_id} - + + {ids.map((id) => ( + + + setExcluded((old) => + v === true ? old.filter((x) => x !== id) : old.includes(id) ? old : [...old, id], + ) + } + aria-label={`归并时包含上报 #${id}`} + /> + + 查看原报告 #{id} + + + ))} + {excluded.length > 0 ? ( + 已排除 {excluded.length} 条：确认后这些记录不会再与本组建议归并 + ) : null} @@ -238,11 +261,12 @@ function FindingMergeSuggestionCard({ - @@ -294,10 +318,15 @@ export function FindingCaseReviewPanel({ taskId, onChange }: { taskId?: string; clearTimeout(timer); }; }, [taskId]); - async function resolve(s: FindingCaseSuggestion, accept: boolean) { + async function resolve(s: FindingCaseSuggestion, accept: boolean, excludeIds: string[] = []) { try { - const result = await api.resolveFindingCaseSuggestion(s.id, accept); + const result = await api.resolveFindingCaseSuggestion(s.id, accept, excludeIds); setSuggestions((v) => v.filter((x) => x.id !== s.id)); + // Server-side resolution may also clear related groups; refresh instead of waiting for the poll. + api + .findingCaseSuggestions(taskId) + .then(setSuggestions) + .catch(() => undefined); if (accept && result.case_id !== "0") { toast.success("已归并，请进入文件夹生成统一报告"); } else { diff --git a/web/src/lib/api.ts b/web/src/lib/api.ts index fedc913..821da2f 100644 --- a/web/src/lib/api.ts +++ b/web/src/lib/api.ts @@ -645,8 +645,11 @@ export const api = { get( `/exploration/finding-case-suggestions${task ? `?task_id=${encodeURIComponent(task)}` : ""}`, ), - resolveFindingCaseSuggestion: (id: string, accept: boolean) => - post<{ case_id: string }>(`/exploration/finding-case-suggestions/${id}`, { accept }), + resolveFindingCaseSuggestion: (id: string, accept: boolean, excludeIds?: string[]) => + post<{ case_id: string }>(`/exploration/finding-case-suggestions/${id}`, { + accept, + ...(excludeIds?.length ? { exclude_ids: excludeIds } : {}), + }), reviewFindingCases: (ids: string[]) => post<{ runs: { conversation_id: number; task_id: string }[] }>("/exploration/finding-case-review", { finding_ids: ids, diff --git a/web/src/lib/types.ts b/web/src/lib/types.ts index 54e26a6..d730691 100644 --- a/web/src/lib/types.ts +++ b/web/src/lib/types.ts @@ -1737,6 +1737,7 @@ export interface FindingCaseSuggestion { title: string; reason: string; state: string; + finding_ids: string[]; } export interface FindingCaseReviewRun { conversation_id: number;