Merge branch 'claude/bold-hawking': 파이프라인 3대 버그 수정 + CLAUDE.md 업데이트

CLAUDE.md

@@ -4,7 +4,9 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project Overview
 
-INFINITH Demo — AI 마케팅 분석 플랫폼의 프론트엔드 데모. 성형외과(뷰성형외과)를 대상으로 YouTube/Instagram/Facebook/웹사이트를 분석하고, 변환 전략/로드맵/KPI를 제안하는 마케팅 리포트를 보여준다. 현재는 mock 데이터 기반이며 실제 API 연동은 미구현 상태.
+INFINITH — AI 마케팅 분석 플랫폼. 성형외과를 대상으로 YouTube/Instagram/Facebook/네이버블로그/네이버플레이스/강남언니를 분석하고, 변환 전략/로드맵/KPI를 제안하는 마케팅 리포트를 생성한다.
+
+**프론트엔드(React) + 백엔드(Supabase Edge Functions) 모두 실제 구현 완료.** `src/data/` mock 데이터는 개발 fallback 용도로만 남아 있으며, 실제 파이프라인은 Supabase Edge Functions에서 실행된다.
 
 ## Commands
 
@@ -27,12 +29,43 @@ INFINITH Demo — AI 마케팅 분석 플랫폼의 프론트엔드 데모. 성
 - `/distribute` — 콘텐츠 배포
 - `/performance` — 성과 분석
 
-**Data flow:**
-1. Pages → custom hooks (`useReport`, `useMarketingPlan`) → mock data (`src/data/`)
-2. Hooks return `{ data, isLoading, error }` 패턴 (100ms delay로 API 시뮬레이션)
+**Frontend Data flow:**
+1. Pages → custom hooks (`useReport`, `useMarketingPlan`) → Supabase DB (`marketing_reports` 테이블)
+2. Hooks return `{ data, isLoading, error }` 패턴
 3. `ScreenshotContext`로 리포트 스크린샷 데이터 공유
 4. `useExportPDF` 훅으로 리포트 → PDF 변환
 
+**Backend Pipeline (Supabase Edge Functions):**
+
+| Phase | Function | 역할 |
+|---|---|---|
+| 1 | `discover-channels` | 병원 URL → SNS 채널 발견 (registry fast-path → clinics 캐시 → Firecrawl 스크래핑) |
+| 2 | `collect-channel-data` | 채널별 실제 데이터 수집 (YouTube API, Apify, Naver API, Firecrawl) |
+| 3 | `generate-report` | Perplexity AI로 리포트 생성 + groundTruth 주입으로 수치 보정 |
+| 4 | `generate-content-plan` | 마케팅 플랜 생성 |
+
+**실제 연동된 API:**
+- YouTube Data API v3 — 구독자/조회수/영상 수
+- Apify — Instagram/Facebook 스크래퍼
+- Naver Search API — 네이버플레이스 검색
+- Naver Blog RSS — 공식 블로그 포스트
+- Firecrawl — 홈페이지 footer SNS 링크 추출, 강남언니 스크래핑
+- Perplexity AI — 리포트 JSON 생성
+- Gemini 2.5 Flash — 콘텐츠 이미지 생성 (`src/services/geminiImageGen.ts`)
+
+**주요 DB 테이블 (Supabase):**
+- `marketing_reports` — 리포트 결과 (`channel_data`, `report`, `analysis_data` JSONB)
+- `clinic_registry` — 병원 SNS 채널 등록 (fast-path 캐시)
+- `clinics` — 병원 기본 정보 + 14일 TTL 캐시
+- `channel_snapshots` — 채널별 수집 이력
+- `analysis_runs` — 파이프라인 실행 로그
+
+**Shared utilities (`supabase/functions/_shared/`):**
+- `extractSocialLinks.ts` — URL → SNS handle 결정론적 추출 (regex 기반)
+- `verifyHandles.ts` — 핸들 실제 존재 여부 검증
+- `dataQuality.ts` — `isMissingValue()`, `validateReportQuality()`
+- `retry.ts` — 자동 재시도/타임아웃/rate limit 내장 fetch wrapper
+
 **Type definitions** in `src/types/`:
 - `report.ts` — `MarketingReport` 및 하위 타입 (YouTubeAudit, InstagramAudit, etc.)
 - `plan.ts` — 마케팅 플랜 타입
@@ -40,6 +73,18 @@ INFINITH Demo — AI 마케팅 분석 플랫폼의 프론트엔드 데모. 성
 
 **AI Integration:** `src/services/geminiImageGen.ts` — Gemini 2.5 Flash로 이미지 생성. `GEMINI_API_KEY` 환경변수 필요.
 
+## Environment Variables
+
+**Supabase Edge Functions (Supabase Dashboard → Settings → Secrets):**
+- `YOUTUBE_API_KEY` — YouTube Data API v3
+- `APIFY_API_KEY` — Instagram/Facebook 스크래퍼
+- `NAVER_CLIENT_ID` / `NAVER_CLIENT_SECRET` — Naver Search API
+- `FIRECRAWL_API_KEY` — Firecrawl 스크래핑
+- `PERPLEXITY_API_KEY` — 리포트 AI 생성
+
+**Frontend (`.env.local`):**
+- `GEMINI_API_KEY` — Gemini 이미지 생성
+
 ## Styling
 
 - Tailwind CSS 4 (`@theme` directive in `src/index.css`)
@@ -55,4 +100,5 @@ INFINITH Demo — AI 마케팅 분석 플랫폼의 프론트엔드 데모. 성
 
 ## Deployment
 
-Vercel (SPA 모드 — 모든 경로 `/index.html`로 리라이트)
+- **Frontend:** Vercel (SPA 모드 — 모든 경로 `/index.html`로 리라이트). `git push` 후 `vercel --prod` 수동 실행 필요 (Gitea 자동 트리거 없음)
+- **Backend:** Supabase Edge Functions — `supabase functions deploy <function-name>` 으로 개별 배포

supabase/functions/collect-channel-data/index.ts

@@ -414,7 +414,14 @@ Deno.serve(async (req) => {
         }
 
         if (!gangnamUnniUrl) {
-          throw new Error("강남언니 URL을 찾을 수 없습니다 (검색 실패)");
+          console.log("[gangnamUnni] URL not found via any search — marking skipped");
+          channelData.gangnamUnni = {
+            status: "skipped",
+            reason: "URL_NOT_FOUND",
+            reasonDetail: "Firecrawl search across 3 query variants returned no gangnamunni.com/hospitals/ link",
+            attemptedAt: new Date().toISOString(),
+          };
+          return;
         }
 
         const scrapeRes = await fetchWithRetry("https://api.firecrawl.dev/v1/scrape", {
@@ -438,11 +445,21 @@ Deno.serve(async (req) => {
             waitFor: 5000,
           }),
         }, { label: "firecrawl-gangnamunni", timeoutMs: 60000 });
-        if (!scrapeRes.ok) throw new Error(`Firecrawl 강남언니 scrape failed: ${scrapeRes.status}`);
+        if (!scrapeRes.ok) {
+          channelData.gangnamUnni = {
+            status: "skipped",
+            reason: "SCRAPE_FAILED",
+            reasonDetail: `Firecrawl returned HTTP ${scrapeRes.status}`,
+            attemptedAt: new Date().toISOString(),
+            sourceUrl: gangnamUnniUrl,
+          };
+          return;
+        }
         const data = await scrapeRes.json();
         const hospital = data.data?.json;
         if (hospital?.hospitalName) {
           channelData.gangnamUnni = {
+            status: "ok",
             name: hospital.hospitalName,
             rawRating: hospital.rating,
             rating: typeof hospital.rating === 'number' && hospital.rating > 0 ? hospital.rating : null,
@@ -452,9 +469,21 @@ Deno.serve(async (req) => {
             badges: hospital.badges || [], sourceUrl: gangnamUnniUrl,
           };
         } else {
-          throw new Error("강남언니 scrape returned no hospital data");
+          channelData.gangnamUnni = {
+            status: "skipped",
+            reason: "EMPTY_SCRAPE_RESULT",
+            reasonDetail: "Firecrawl scraped the page but could not extract hospital data",
+            attemptedAt: new Date().toISOString(),
+            sourceUrl: gangnamUnniUrl,
+          };
         }
       }));
+    } else {
+      channelData.gangnamUnni = {
+        status: "skipped",
+        reason: !FIRECRAWL_API_KEY ? "FIRECRAWL_API_KEY_MISSING" : "CLINIC_NAME_MISSING",
+        attemptedAt: new Date().toISOString(),
+      };
     }
 
     // ─── 5. Naver Blog + Place ───
@@ -490,6 +519,7 @@ Deno.serve(async (req) => {
           const totalMatch = xml.match(/<totalCount>(\d+)<\/totalCount>/) || xml.match(/<managedCount>(\d+)<\/managedCount>/);
           const totalPosts = totalMatch ? Number(totalMatch[1]) : items.length;
           channelData.naverBlog = {
+            status: "ok",
             officialBlogUrl, officialBlogHandle,
             totalResults: totalPosts,
             posts: items.slice(0, 10).map(i => ({
@@ -500,13 +530,28 @@ Deno.serve(async (req) => {
           };
           console.log(`[naverBlog] RSS: ${items.length} posts from verified handle ${officialBlogHandle}`);
         } catch (e) {
-          console.warn(`[naverBlog] RSS fetch failed:`, e);
-          // Fallback: at minimum expose the official URL even without post data
-          channelData.naverBlog = { officialBlogUrl, officialBlogHandle, totalResults: 0, posts: [], officialContent: null };
+          const reason = e instanceof Error ? e.message : String(e);
+          console.warn(`[naverBlog] RSS fetch failed:`, reason);
+          // Expose the official URL even without post data so the frontend can still link out
+          channelData.naverBlog = {
+            status: "skipped",
+            reason: "RSS_FETCH_FAILED",
+            reasonDetail: reason,
+            attemptedAt: new Date().toISOString(),
+            officialBlogUrl, officialBlogHandle,
+            totalResults: 0, posts: [], officialContent: null,
+          };
         }
       }));
     } else {
-      console.log(`[naverBlog] No verified handle in DB — skipping`);
+      console.log(`[naverBlog] No verified handle in DB — marking skipped`);
+      channelData.naverBlog = {
+        status: "skipped",
+        reason: "NO_VERIFIED_HANDLE",
+        attemptedAt: new Date().toISOString(),
+        totalResults: 0,
+        posts: [],
+      };
     }
 
     // naverPlace: use stored verified data if available, otherwise search once and save
@@ -564,7 +609,7 @@ Deno.serve(async (req) => {
         }
 
         if (found) {
-          channelData.naverPlace = found;
+          channelData.naverPlace = { status: "ok", ...found };
           // Save to clinics.verified_channels so future runs skip the search
           if (inputClinicId) {
             const { data: clinicRow } = await supabase.from('clinics').select('verified_channels').eq('id', inputClinicId).single();
@@ -576,9 +621,22 @@ Deno.serve(async (req) => {
             }
           }
         } else {
-          console.log(`[naverPlace] No confident match found — skipping to avoid wrong data`);
+          console.log(`[naverPlace] No confident match found — marking skipped`);
+          channelData.naverPlace = {
+            status: "skipped",
+            reason: "NO_CONFIDENT_MATCH",
+            reasonDetail: `Tried ${queries.length} queries but none matched the domain or exact clinic name`,
+            attemptedAt: new Date().toISOString(),
+            attemptedQueries: queries,
+          };
         }
       }));
+    } else {
+      channelData.naverPlace = {
+        status: "skipped",
+        reason: "NAVER_API_CREDENTIALS_MISSING",
+        attemptedAt: new Date().toISOString(),
+      };
     }
 
     // ─── 6. Google Maps (Google Places API New) ───
@@ -587,6 +645,7 @@ Deno.serve(async (req) => {
         const place = await searchGooglePlace(clinicName, address || undefined, GOOGLE_PLACES_API_KEY);
         if (place) {
           channelData.googleMaps = {
+            status: "ok",
             name: place.name, rating: place.rating, reviewCount: place.reviewCount,
             address: place.address, phone: place.phone,
             clinicWebsite: place.clinicWebsite,
@@ -596,9 +655,20 @@ Deno.serve(async (req) => {
             topReviews: place.topReviews,
           };
         } else {
-          throw new Error("Google Maps: no matching place found");
+          channelData.googleMaps = {
+            status: "skipped",
+            reason: "PLACE_NOT_FOUND",
+            attemptedAt: new Date().toISOString(),
+            searchQuery: clinicName,
+          };
         }
       }));
+    } else {
+      channelData.googleMaps = {
+        status: "skipped",
+        reason: !GOOGLE_PLACES_API_KEY ? "API_KEY_MISSING" : "CLINIC_NAME_MISSING",
+        attemptedAt: new Date().toISOString(),
+      };
     }
 
     // ─── 7. Market Analysis (Perplexity) ───
@@ -841,34 +911,41 @@ Deno.serve(async (req) => {
           });
         }
 
+        // Helper: only snapshot channels that actually collected data.
+        // Skipped channels (status === 'skipped') still live in channelData for the
+        // generate-report step to render their skip reason, but shouldn't pollute
+        // the time-series channel_snapshots table.
+        const isCollected = (d: Record<string, unknown> | undefined): boolean =>
+          !!d && d.status !== "skipped";
+
         const guData = channelData.gangnamUnni as Record<string, unknown> | undefined;
-        if (guData) {
+        if (isCollected(guData)) {
           snapshotInserts.push({
             clinic_id: clinicId, run_id: runId, channel: 'gangnamUnni',
-            handle: guData.name, rating: guData.rating, rating_scale: 10,
-            reviews: guData.totalReviews,
-            health_score: computeHealthScore('gangnamUnni', guData),
+            handle: guData!.name, rating: guData!.rating, rating_scale: 10,
+            reviews: guData!.totalReviews,
+            health_score: computeHealthScore('gangnamUnni', guData!),
             details: guData,
           });
         }
 
         const gmData = channelData.googleMaps as Record<string, unknown> | undefined;
-        if (gmData) {
+        if (isCollected(gmData)) {
           snapshotInserts.push({
             clinic_id: clinicId, run_id: runId, channel: 'googleMaps',
-            handle: gmData.name, rating: gmData.rating, rating_scale: 5,
-            reviews: gmData.reviewCount,
-            health_score: computeHealthScore('googleMaps', gmData),
+            handle: gmData!.name, rating: gmData!.rating, rating_scale: 5,
+            reviews: gmData!.reviewCount,
+            health_score: computeHealthScore('googleMaps', gmData!),
             details: gmData,
           });
         }
 
         const nbData = channelData.naverBlog as Record<string, unknown> | undefined;
-        if (nbData) {
+        if (isCollected(nbData)) {
           snapshotInserts.push({
             clinic_id: clinicId, run_id: runId, channel: 'naverBlog',
-            handle: nbData.officialBlogHandle,
-            health_score: computeHealthScore('naverBlog', nbData),
+            handle: nbData!.officialBlogHandle,
+            health_score: computeHealthScore('naverBlog', nbData!),
             details: nbData,
           });
         }

supabase/functions/discover-channels/index.ts

@@ -284,23 +284,122 @@ Deno.serve(async (req) => {
     }
 
     // ═══════════════════════════════════════════
-    // NOT REGISTERED: Return error for unregistered domains
-    // (Registry-only mode — no API fallback)
+    // REGISTRY MISS — secondary fast-path via `clinics` cache
+    // If this domain was previously auto-discovered and cached in `clinics.verified_channels`,
+    // reuse it instead of re-running the full Firecrawl+Perplexity discovery.
+    // Fresh runs only re-discover after 14 days so stale handles eventually refresh.
     // ═══════════════════════════════════════════
-    console.log(`[registry] Miss: ${registryDomain} — returning CLINIC_NOT_REGISTERED`);
-    return new Response(
-      JSON.stringify({
-        success: false,
-        error: "CLINIC_NOT_REGISTERED",
-        message: "현재 지원하지 않는 병원입니다. 등록된 병원만 분석 가능합니다.",
-        domain: registryDomain,
-      }),
-      { status: 404, headers: { ...corsHeaders, "Content-Type": "application/json" } },
-    );
+    console.log(`[registry] Miss: ${registryDomain} — trying clinics cache + Firecrawl fallback`);
+
+    if (registryDomain) {
+      try {
+        const { data: cachedClinic } = await supabase
+          .from("clinics")
+          .select("id, name, name_en, address, phone, services, branding, verified_channels, last_analyzed_at")
+          .eq("domain", registryDomain)
+          .maybeSingle();
+
+        const cachedChannels = cachedClinic?.verified_channels as VerifiedChannels | null | undefined;
+        const hasAny = cachedChannels && (
+          ((cachedChannels as Record<string, unknown>).instagram as unknown[] | undefined)?.length
+          || (cachedChannels as Record<string, unknown>).youtube
+          || (cachedChannels as Record<string, unknown>).facebook
+          || (cachedChannels as Record<string, unknown>).naverBlog
+        );
+        const lastAnalyzed = cachedClinic?.last_analyzed_at ? new Date(cachedClinic.last_analyzed_at).getTime() : 0;
+        const CACHE_TTL_MS = 14 * 24 * 60 * 60 * 1000;
+        const cacheFresh = lastAnalyzed > 0 && (Date.now() - lastAnalyzed) < CACHE_TTL_MS;
+
+        if (cachedClinic && hasAny && cacheFresh) {
+          console.log(`[clinics-cache] Hit: ${cachedClinic.name} (${registryDomain}) — reusing cached channels`);
+
+          const scrapeDataFromCache = {
+            clinic: {
+              clinicName: cachedClinic.name,
+              clinicNameEn: cachedClinic.name_en,
+              address: cachedClinic.address,
+              phone: cachedClinic.phone,
+              services: cachedClinic.services || [],
+            },
+            branding: cachedClinic.branding || {},
+            siteLinks: [],
+            siteMap: [],
+            sourceUrl: url,
+            scrapedAt: new Date().toISOString(),
+            source: "clinics-cache",
+          };
+
+          const { data: saved, error: saveError } = await supabase
+            .from("marketing_reports")
+            .insert({
+              url,
+              clinic_name: cachedClinic.name,
+              status: "discovered",
+              verified_channels: cachedChannels,
+              scrape_data: scrapeDataFromCache,
+              report: {},
+              pipeline_started_at: new Date().toISOString(),
+            })
+            .select("id")
+            .single();
+
+          if (saveError) throw new Error(`DB save failed: ${saveError.message}`);
+
+          // Refresh last_analyzed_at so the cache stays warm
+          await supabase.from("clinics")
+            .update({ last_analyzed_at: new Date().toISOString() })
+            .eq("id", cachedClinic.id);
+
+          // V3 run record
+          let runId: string | null = null;
+          try {
+            const { data: runRow } = await supabase
+              .from("analysis_runs")
+              .insert({
+                clinic_id: cachedClinic.id,
+                status: "discovering",
+                scrape_data: scrapeDataFromCache,
+                discovered_channels: cachedChannels,
+                trigger: "manual",
+                pipeline_started_at: new Date().toISOString(),
+              })
+              .select("id")
+              .single();
+            runId = runRow?.id || null;
+          } catch (e) {
+            console.error("V3 dual-write error (clinics-cache):", e);
+          }
+
+          return new Response(
+            JSON.stringify({
+              success: true,
+              reportId: saved.id,
+              clinicId: cachedClinic.id,
+              runId,
+              clinicName: cachedClinic.name,
+              verifiedChannels: cachedChannels,
+              address: cachedClinic.address || "",
+              services: cachedClinic.services || [],
+              scrapeData: scrapeDataFromCache,
+              source: "clinics-cache",
+            }),
+            { headers: { ...corsHeaders, "Content-Type": "application/json" } },
+          );
+        } else if (cachedClinic && !cacheFresh) {
+          console.log(`[clinics-cache] Stale for ${registryDomain} — re-running discovery`);
+        }
+      } catch (e) {
+        console.warn("[clinics-cache] Lookup failed, falling through to full discovery:", e instanceof Error ? e.message : e);
+      }
+    }
 
     // ═══════════════════════════════════════════
-    // LEGACY FALLBACK: Full API discovery (disabled — registry-only mode)
-    // Kept for reference; unreachable in production
+    // FULL DISCOVERY: Firecrawl + extractSocialLinks + verifyHandles
+    // Runs for: unregistered domains, stale cache, or cache miss.
+    // Deterministic footer/link extraction (Source 1-3) is prioritized;
+    // Perplexity/Apify/Naver (Source 4-5) are AI fallbacks that only contribute
+    // when the regex path is empty. Everything that succeeds gets saved to
+    // `clinics.verified_channels` so the next run hits the cache fast-path above.
     // ═══════════════════════════════════════════
 
     const FIRECRAWL_API_KEY = Deno.env.get("FIRECRAWL_API_KEY") || "";
@@ -806,6 +905,19 @@ Deno.serve(async (req) => {
       } catch { /* ignore secondary failure */ }
     }
 
+    // Log deterministic vs AI contribution so we can see whether the footer
+    // scraping alone was sufficient for this clinic (the developer's hypothesis).
+    const deterministicCount =
+      (linkHandles.instagram?.length || 0) +
+      (linkHandles.youtube?.length || 0) +
+      (linkHandles.facebook?.length || 0) +
+      (linkHandles.naverBlog?.length || 0) +
+      (buttonHandles.instagram?.length || 0) +
+      (buttonHandles.youtube?.length || 0) +
+      (buttonHandles.facebook?.length || 0) +
+      (buttonHandles.naverBlog?.length || 0);
+    console.log(`[discover] ${registryDomain} — deterministic handles: ${deterministicCount}, final verified channels: ig=${(verified.instagram || []).length}, yt=${verified.youtube ? 1 : 0}, fb=${verified.facebook ? 1 : 0}, blog=${verified.naverBlog ? 1 : 0}`);
+
     return new Response(
       JSON.stringify({
         success: true, reportId: saved.id,
@@ -815,6 +927,7 @@ Deno.serve(async (req) => {
         address: clinic.address || "",
         services: clinic.services || [],
         scrapeData: scrapeDataFull,
+        source: "firecrawl-fallback",
       }),
       { headers: { ...corsHeaders, "Content-Type": "application/json" } },
     );

supabase/functions/generate-report/index.ts

@@ -195,6 +195,157 @@ ${JSON.stringify(scrapeData.branding || {}, null, 2).slice(0, 1000)}
         report.visionAnalysis = vision;
       }
 
+      // ─── Harness 4: GroundTruth channel injection ───
+      // Perplexity occasionally drops channel metrics even though they exist in channelData.
+      // For each channel we force-inject the numeric fields from the actually-collected data.
+      // Same pattern as the Vision injection above (foundingYear / gangnamUnni), extended
+      // to instagram / youtube / facebook / naverBlog / naverPlace.
+      report.channelAnalysis = report.channelAnalysis || {};
+      const injectedChannels: string[] = [];
+
+      // Instagram — prefer instagramAccounts[0], fall back to channelData.instagram
+      const igAccountsList = (channelData.instagramAccounts as Record<string, unknown>[] | undefined) || [];
+      const igPrimary = (igAccountsList[0] || channelData.instagram) as Record<string, unknown> | undefined;
+      const igPostsSummary = channelData.instagramPosts as Record<string, unknown> | undefined;
+      if (igPrimary) {
+        const target = (report.channelAnalysis.instagram ||= {} as Record<string, unknown>);
+        if (isMissingValue(target.followers) && igPrimary.followers) {
+          target.followers = igPrimary.followers;
+        }
+        // Prefer profile-level post count; fall back to scraped posts summary
+        if (isMissingValue(target.posts)) {
+          if (igPrimary.posts) target.posts = igPrimary.posts;
+          else if (igPostsSummary?.totalPosts) target.posts = igPostsSummary.totalPosts;
+        }
+        if (isMissingValue(target.status)) target.status = "active";
+        injectedChannels.push(`instagram(f=${target.followers},p=${target.posts})`);
+      }
+
+      // YouTube
+      const ytGT = channelData.youtube as Record<string, unknown> | undefined;
+      if (ytGT) {
+        const target = (report.channelAnalysis.youtube ||= {} as Record<string, unknown>);
+        if (isMissingValue(target.subscribers) && ytGT.subscribers) {
+          target.subscribers = ytGT.subscribers;
+        }
+        if (isMissingValue((target as Record<string, unknown>).videos) && ytGT.totalVideos) {
+          (target as Record<string, unknown>).videos = ytGT.totalVideos;
+        }
+        if (isMissingValue(target.status)) target.status = "active";
+        injectedChannels.push(`youtube(s=${target.subscribers})`);
+      }
+
+      // Facebook
+      const fbGT = channelData.facebook as Record<string, unknown> | undefined;
+      if (fbGT) {
+        const target = (report.channelAnalysis.facebook ||= {} as Record<string, unknown>);
+        if (isMissingValue(target.followers) && fbGT.followers) {
+          target.followers = fbGT.followers;
+        }
+        if (isMissingValue((target as Record<string, unknown>).likes) && fbGT.likes) {
+          (target as Record<string, unknown>).likes = fbGT.likes;
+        }
+        if (isMissingValue(target.status)) target.status = "active";
+        injectedChannels.push(`facebook(f=${target.followers})`);
+      }
+
+      // Naver Blog — RSS provides totalResults + posts
+      const nbGT = channelData.naverBlog as Record<string, unknown> | undefined;
+      if (nbGT && !nbGT.skipped) {
+        const target = (report.channelAnalysis.naverBlog ||= {} as Record<string, unknown>);
+        if (isMissingValue(target.posts) && nbGT.totalResults) {
+          target.posts = nbGT.totalResults;
+        }
+        if (isMissingValue(target.status)) {
+          target.status = (nbGT.totalResults as number) > 0 ? "active" : "inactive";
+        }
+        injectedChannels.push(`naverBlog(p=${target.posts})`);
+      }
+
+      // Naver Place — Naver Local Search doesn't include rating/reviews, but stores the place metadata.
+      // If the place was found we at least set status=active; rating/reviews often come via Vision.
+      const npGT = channelData.naverPlace as Record<string, unknown> | undefined;
+      if (npGT && !npGT.skipped && npGT.name) {
+        const target = (report.channelAnalysis.naverPlace ||= {} as Record<string, unknown>);
+        if (isMissingValue(target.status)) target.status = "active";
+        if (isMissingValue((target as Record<string, unknown>).name) && npGT.name) {
+          (target as Record<string, unknown>).name = npGT.name;
+        }
+        // Rating/reviews may be injected by Vision block above; don't overwrite
+        injectedChannels.push(`naverPlace(${npGT.name})`);
+      }
+
+      // GangnamUnni — Vision block handled the rating/reviews path already.
+      // Here we only fill status/rating when they come directly from the scraper (D).
+      const guGT = channelData.gangnamUnni as Record<string, unknown> | undefined;
+      if (guGT && !guGT.skipped) {
+        const target = (report.channelAnalysis.gangnamUnni ||= {} as Record<string, unknown>);
+        if (isMissingValue(target.rating) && guGT.rating) {
+          target.rating = guGT.rating;
+          (target as Record<string, unknown>).ratingScale = 10;
+        }
+        if (isMissingValue(target.reviews) && (guGT.totalReviews || guGT.reviews)) {
+          target.reviews = guGT.totalReviews || guGT.reviews;
+        }
+        if (isMissingValue(target.status)) target.status = "active";
+        injectedChannels.push(`gangnamUnni(r=${target.rating})`);
+      }
+
+      if (injectedChannels.length > 0) {
+        console.log(`[report] GroundTruth injected: ${injectedChannels.join(", ")}`);
+      }
+
+      // ─── Harness 4b: Diagnosis fallback ───
+      // ProblemDiagnosis.tsx returns null when `problemDiagnosis` is empty, hiding the whole
+      // section. Make sure each weak channel contributes at least one diagnosis item so the
+      // frontend renders. Only runs when AI didn't fill diagnosis itself.
+      const DEFAULT_DIAGNOSIS: Record<string, { issue: string; recommendation: string }> = {
+        instagram: {
+          issue: "Instagram 업로드 빈도가 부족하거나 게시물 참여율이 낮습니다",
+          recommendation: "주 3회 이상 주제별 콘텐츠 발행 + 릴스 비중 40% 이상 확보",
+        },
+        youtube: {
+          issue: "YouTube 채널 활동이 저조하여 브랜드 신뢰도 확보가 어렵습니다",
+          recommendation: "시술 Before/After Shorts 주 2회, 롱폼 월 2회 업로드 루틴 수립",
+        },
+        facebook: {
+          issue: "Facebook 페이지 활동이 낮아 유입 기여가 제한적입니다",
+          recommendation: "광고 랜딩 채널로 재정의하고 월 4회 프로모션 포스트 자동화",
+        },
+        naverBlog: {
+          issue: "네이버 블로그 최신 포스팅 양이 부족하여 로컬 SEO 기회를 놓치고 있습니다",
+          recommendation: "주요 시술 키워드별 주 2건, 월 8건 이상 SEO 최적화 포스팅",
+        },
+        naverPlace: {
+          issue: "네이버 플레이스 리뷰/콘텐츠 관리가 체계화되어 있지 않습니다",
+          recommendation: "방문 리뷰 유도 프로세스 구축 + 키워드·사진 리뷰 월 20건 확보",
+        },
+        gangnamUnni: {
+          issue: "강남언니 프로필 정보와 답변률이 미흡합니다",
+          recommendation: "의사별 프로필 완성도 100% + 문의 24시간 내 응답 체계",
+        },
+      };
+      const weakStatuses = new Set(["inactive", "weak", "not_found", undefined]);
+      for (const ch of Object.keys(DEFAULT_DIAGNOSIS)) {
+        const node = (report.channelAnalysis as Record<string, unknown>)[ch] as
+          | Record<string, unknown>
+          | undefined;
+        if (!node) continue;
+        const existing = node.diagnosis as unknown[] | undefined;
+        if (Array.isArray(existing) && existing.length > 0) continue;
+        const score = typeof node.score === "number" ? node.score : null;
+        const status = node.status as string | undefined;
+        const needsFallback = (score !== null && score < 60) || weakStatuses.has(status);
+        if (!needsFallback) continue;
+        const def = DEFAULT_DIAGNOSIS[ch];
+        node.diagnosis = [{
+          issue: def.issue,
+          severity: score !== null && score < 40 ? "critical" : "warning",
+          recommendation: def.recommendation,
+        }];
+        console.log(`[report] Diagnosis fallback injected for ${ch} (score=${score}, status=${status})`);
+      }
+
       // Embed channel enrichment data for frontend mergeEnrichment()
       report.channelEnrichment = channelData;
       report.enrichedAt = new Date().toISOString();
@@ -320,10 +471,31 @@ ${JSON.stringify(scrapeData.branding || {}, null, 2).slice(0, 1000)}
         console.warn(`[harness] Low report quality (${qualityReport.score}/100):`, qualityReport.warnings);
       }
 
+      // Persist quality report into analysis_data.qualityReport so the frontend (and
+      // future debugging queries) can inspect why specific sections rendered empty.
+      // Downgrade status to 'partial' when too many important/critical fields are missing.
+      const totalMissing = qualityReport.missingCritical.length + qualityReport.missingImportant.length;
+      const reportStatus = (qualityReport.missingCritical.length > 0 || totalMissing >= 5)
+        ? "partial"
+        : "complete";
+      const enrichedAnalysisData = {
+        ...analysisData,
+        qualityReport: {
+          score: qualityReport.score,
+          missingCritical: qualityReport.missingCritical,
+          missingImportant: qualityReport.missingImportant,
+          missingOptional: qualityReport.missingOptional,
+          warnings: qualityReport.warnings,
+          injectedChannels,
+          generatedAt: new Date().toISOString(),
+        },
+      };
+
       // Legacy: marketing_reports
       await supabase.from("marketing_reports").update({
         report,
-        status: "complete",
+        analysis_data: enrichedAnalysisData,
+        status: reportStatus,
         data_quality_score: qualityReport.score,
         pipeline_completed_at: new Date().toISOString(),
         updated_at: new Date().toISOString(),