Add human review feature and enhance data collection

Enhanced the web interface with comprehensive human review capabilities
for validating AI-flagged toxicity classifications. Added date filtering
and improved data collection to include engagement metrics for mentions.

Features added:
- Human review system with ✓/✗/? status buttons and filtering
- Date range filtering (from/to) for flagged content
- Review status tracking with database migrations
- Engagement metrics collection for mentions (likes, replies, reposts, quotes)
- Interactive review buttons that allow changing classifications
- Review filter to show unreviewed, correct, incorrect, or unsure items

UI improvements:
- Fixed Chart.js CDN URLs (switched to jsdelivr)
- Smart axis scaling for toxicity category charts with dynamic decimal places
- Clickable max toxicity badges linking to filtered content
- Improved mention author display using raw_json fallback
- Sortable table columns with visual indicators
- Review status preserved across pagination and filtering

Bug fixes:
- Commented out problematic account (stephanvanbaarle.bsky.social)
- Fixed filter parameter names (content_type, account_did)
- Fixed threshold boundary issues with 0.001 offset
- Added extra_js block to base template for JavaScript functionality

Database changes:
- Migration 03: Added engagement columns to mentions table
- Migration 04: Added human_reviewed, review_status, reviewed_at columns

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

.claude/settings.local.json

@@ -0,0 +1,9 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(docker compose:*)"
+    ],
+    "deny": [],
+    "ask": []
+  }
+}

config/accounts.yml

@@ -99,7 +99,7 @@ accounts:
   - handle: marietbosman62.bsky.social  # Mariet Bosman (BBB)
   - handle: natalienauta.bsky.social  # Natalie Nauta (BBB)
   - handle: nielsoosterom.bsky.social  # Niels Oosterom (BBB)
-  - handle: stephanvanbaarle.bsky.social  # Stephan van Baarle (DENK)
+  # - handle: stephanvanbaarle.bsky.social  # Stephan van Baarle (DENK) - DISABLED: Account returns 400 error
   - handle: ananninga.bsky.social  # Annabel Nanninga (JA21)
   - handle: djhvandijk.bsky.social  # Diederik van Dijk (SGP)
   - handle: carladikfaber.bsky.social  # Carla Dik-Faber (ChristenUnie)

scripts/03-mentions-engagement.sql

@@ -0,0 +1,10 @@
+-- Add engagement metrics to mentions table
+-- Migration to add like_count, reply_count, repost_count, quote_count
+
+ALTER TABLE mentions
+ADD COLUMN IF NOT EXISTS like_count INTEGER DEFAULT 0,
+ADD COLUMN IF NOT EXISTS reply_count INTEGER DEFAULT 0,
+ADD COLUMN IF NOT EXISTS repost_count INTEGER DEFAULT 0,
+ADD COLUMN IF NOT EXISTS quote_count INTEGER DEFAULT 0;
+
+CREATE INDEX IF NOT EXISTS idx_mentions_like_count ON mentions (like_count DESC);

scripts/04-human-review.sql

@@ -0,0 +1,20 @@
+-- Add human review functionality to toxicity scores
+-- Migration to add human_reviewed, review_status, and reviewed_at columns
+
+-- Add columns to toxicity_scores (for posts)
+ALTER TABLE toxicity_scores
+ADD COLUMN IF NOT EXISTS human_reviewed BOOLEAN DEFAULT FALSE,
+ADD COLUMN IF NOT EXISTS review_status VARCHAR(20) DEFAULT NULL, -- 'correct', 'incorrect', 'unsure'
+ADD COLUMN IF NOT EXISTS reviewed_at TIMESTAMP DEFAULT NULL;
+
+-- Add columns to mention_toxicity_scores (for mentions)
+ALTER TABLE mention_toxicity_scores
+ADD COLUMN IF NOT EXISTS human_reviewed BOOLEAN DEFAULT FALSE,
+ADD COLUMN IF NOT EXISTS review_status VARCHAR(20) DEFAULT NULL, -- 'correct', 'incorrect', 'unsure'
+ADD COLUMN IF NOT EXISTS reviewed_at TIMESTAMP DEFAULT NULL;
+
+-- Create indexes for filtering
+CREATE INDEX IF NOT EXISTS idx_toxicity_scores_human_reviewed ON toxicity_scores (human_reviewed);
+CREATE INDEX IF NOT EXISTS idx_toxicity_scores_review_status ON toxicity_scores (review_status);
+CREATE INDEX IF NOT EXISTS idx_mention_toxicity_scores_human_reviewed ON mention_toxicity_scores (human_reviewed);
+CREATE INDEX IF NOT EXISTS idx_mention_toxicity_scores_review_status ON mention_toxicity_scores (review_status);

src/bluesky_client.py

@@ -297,5 +297,9 @@ def map_search_post_to_mention(post_data: dict, mentioned_did: str) -> Mention:
         mentioning_did=post_data.get("author", {}).get("did"),
         post_text=record.get("text"),
         post_created_at=_parse_dt(record.get("createdAt")),
+        like_count=post_data.get("likeCount", 0) or 0,
+        reply_count=post_data.get("replyCount", 0) or 0,
+        repost_count=post_data.get("repostCount", 0) or 0,
+        quote_count=post_data.get("quoteCount", 0) or 0,
         raw_json=post_data,
     )

src/db.py

@@ -136,15 +136,25 @@ class Database:
                         """
                         INSERT INTO mentions (
                             post_uri, mentioned_did, mentioning_did,
-                            post_text, post_created_at, raw_json
+                            post_text, post_created_at,
-                        ) VALUES ($1, $2, $3, $4, $5, $6)
+                            like_count, reply_count, repost_count, quote_count,
-                        ON CONFLICT (post_uri, mentioned_did) DO NOTHING
+                            raw_json
+                        ) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10)
+                        ON CONFLICT (post_uri, mentioned_did) DO UPDATE SET
+                            like_count = EXCLUDED.like_count,
+                            reply_count = EXCLUDED.reply_count,
+                            repost_count = EXCLUDED.repost_count,
+                            quote_count = EXCLUDED.quote_count
                         """,
                         m.post_uri,
                         m.mentioned_did,
                         m.mentioning_did,
                         m.post_text,
                         m.post_created_at,
+                        m.like_count,
+                        m.reply_count,
+                        m.repost_count,
+                        m.quote_count,
                         json.dumps(m.raw_json),
                     )
                     if "INSERT 0 1" in result:

src/models.py

@@ -46,6 +46,10 @@ class Mention:
     mentioning_did: str | None
     post_text: str | None
     post_created_at: datetime | None
+    like_count: int = 0
+    reply_count: int = 0
+    repost_count: int = 0
+    quote_count: int = 0
     raw_json: dict[str, Any] = field(default_factory=dict)
 
 

src/web/app.py

@@ -48,6 +48,7 @@ def create_app() -> Flask:
     from .routes.mentions import bp as mentions_bp
     from .routes.export import bp as export_bp
     from .routes.analysis import bp as analysis_bp
+    from .routes.review import bp as review_bp
 
     app.register_blueprint(dashboard_bp)
     app.register_blueprint(accounts_bp)
@@ -55,6 +56,7 @@ def create_app() -> Flask:
     app.register_blueprint(mentions_bp)
     app.register_blueprint(export_bp)
     app.register_blueprint(analysis_bp)
+    app.register_blueprint(review_bp)
 
     # Teardown
     @app.teardown_appcontext

src/web/db.py

@@ -495,6 +495,11 @@ def get_flagged_content(
     category: str | None = None,
     account_did: str | None = None,
     threshold: float = 0.5,
+    review_status: str | None = None,
+    date_from: str | None = None,
+    date_to: str | None = None,
+    sort: str = "overall",
+    direction: str = "desc",
     limit: int = 50,
     offset: int = 0,
 ) -> tuple[list[dict], int]:
@@ -522,6 +527,29 @@ def get_flagged_content(
         mention_conditions += " AND m.mentioned_did = %s"
         params_mentions.append(account_did)
 
+    if review_status:
+        if review_status == "unreviewed":
+            post_conditions += " AND ts.human_reviewed = false"
+            mention_conditions += " AND mts.human_reviewed = false"
+        elif review_status in ("correct", "incorrect", "unsure"):
+            post_conditions += " AND ts.review_status = %s"
+            params_posts.append(review_status)
+            mention_conditions += " AND mts.review_status = %s"
+            params_mentions.append(review_status)
+
+    if date_from:
+        post_conditions += " AND p.created_at >= %s"
+        params_posts.append(date_from)
+        mention_conditions += " AND m.post_created_at >= %s"
+        params_mentions.append(date_from)
+
+    if date_to:
+        # Add one day to include the entire "to" date
+        post_conditions += " AND p.created_at < (%s::date + interval '1 day')"
+        params_posts.append(date_to)
+        mention_conditions += " AND m.post_created_at < (%s::date + interval '1 day')"
+        params_mentions.append(date_to)
+
     type_filter_post = ""
     type_filter_mention = ""
     if content_type == "mention":
@@ -533,6 +561,14 @@ def get_flagged_content(
         elif content_type == "reply":
             post_conditions += " AND p.post_type = 'reply'"
 
+    # Build ORDER BY clause
+    valid_sorts = {"overall", "created_at", "author_handle"}
+    sort_col = sort if sort in valid_sorts else "overall"
+    dir_sql = "DESC" if direction == "desc" else "ASC"
+    order_by = f"ORDER BY {sort_col} {dir_sql}"
+    if sort_col != "created_at":
+        order_by += ", created_at DESC"
+
     with get_cursor() as cur:
         # Count
         cur.execute(f"""
@@ -564,7 +600,8 @@ def get_flagged_content(
                     ts.overall, ts.toxic, ts.threat, ts.hate_speech,
                     ts.racism, ts.antisemitism, ts.islamophobia,
                     ts.sexism, ts.homophobia, ts.insult, ts.dehumanization,
-                    ts.extremism, ts.ableism
+                    ts.extremism, ts.ableism,
+                    ts.human_reviewed, ts.review_status, ts.reviewed_at
                 FROM toxicity_scores ts
                 JOIN posts p ON p.uri = ts.uri
                 LEFT JOIN accounts a ON a.did = p.author_did
@@ -578,20 +615,22 @@ def get_flagged_content(
                     m.post_uri AS item_id,
                     m.post_text AS text,
                     m.mentioning_did AS author_did,
-                    NULL AS author_handle,
+                    COALESCE(aa.handle, m.raw_json->'author'->>'handle') AS author_handle,
                     m.mentioned_did,
                     ma.handle AS mentioned_handle,
                     m.post_created_at AS created_at,
                     mts.overall, mts.toxic, mts.threat, mts.hate_speech,
                     mts.racism, mts.antisemitism, mts.islamophobia,
                     mts.sexism, mts.homophobia, mts.insult, mts.dehumanization,
-                    mts.extremism, mts.ableism
+                    mts.extremism, mts.ableism,
+                    mts.human_reviewed, mts.review_status, mts.reviewed_at
                 FROM mention_toxicity_scores mts
                 JOIN mentions m ON m.id = mts.mention_id
                 LEFT JOIN accounts ma ON ma.did = m.mentioned_did
+                LEFT JOIN accounts aa ON aa.did = m.mentioning_did
                 {mention_conditions} {type_filter_mention}
             ) combined
-            ORDER BY overall DESC, created_at DESC
+            {order_by}
             LIMIT %s OFFSET %s
         """, params_posts + params_mentions + [limit, offset])
         rows = [dict(r) for r in cur.fetchall()]
@@ -636,15 +675,18 @@ def get_account_toxicity_summary(
             SELECT
                 a.did, a.handle, a.display_name,
                 coalesce(post_agg.avg_tox, 0) AS avg_post_tox,
+                coalesce(post_agg.max_tox, 0) AS max_post_tox,
                 coalesce(post_agg.flagged, 0) AS flagged_posts,
                 coalesce(post_agg.total, 0) AS scored_posts,
                 coalesce(mention_agg.avg_tox, 0) AS avg_mention_tox,
+                coalesce(mention_agg.max_tox, 0) AS max_mention_tox,
                 coalesce(mention_agg.flagged, 0) AS flagged_mentions,
                 coalesce(mention_agg.total, 0) AS scored_mentions
             FROM accounts a
             LEFT JOIN (
                 SELECT p.author_did,
                        avg(ts.overall) AS avg_tox,
+                       max(ts.overall) AS max_tox,
                        count(*) FILTER (WHERE ts.flagged) AS flagged,
                        count(*) AS total
                 FROM toxicity_scores ts
@@ -654,6 +696,7 @@ def get_account_toxicity_summary(
             LEFT JOIN (
                 SELECT m.mentioned_did,
                        avg(mts.overall) AS avg_tox,
+                       max(mts.overall) AS max_tox,
                        count(*) FILTER (WHERE mts.flagged) AS flagged,
                        count(*) AS total
                 FROM mention_toxicity_scores mts

src/web/routes/analysis.py

@@ -55,10 +55,15 @@ def index():
 
 @bp.route("/flagged")
 def flagged():
-    content_type = request.args.get("type") or None
+    content_type = request.args.get("content_type") or None
     category = request.args.get("category") or None
-    account_did = request.args.get("account") or None
+    account_did = request.args.get("account_did") or None
     threshold = request.args.get("threshold", 0.5, type=float)
+    review_status = request.args.get("review_status") or None
+    date_from = request.args.get("date_from") or None
+    date_to = request.args.get("date_to") or None
+    sort = request.args.get("sort", "overall")
+    direction = request.args.get("dir", "desc")
     page = max(1, request.args.get("page", 1, type=int))
     per_page = 50
 
@@ -67,6 +72,11 @@ def flagged():
         category=category,
         account_did=account_did,
         threshold=threshold,
+        review_status=review_status,
+        date_from=date_from,
+        date_to=date_to,
+        sort=sort,
+        direction=direction,
         limit=per_page,
         offset=(page - 1) * per_page,
     )
@@ -85,6 +95,11 @@ def flagged():
         category=category or "",
         account_did=account_did or "",
         threshold=threshold,
+        review_status=review_status or "",
+        date_from=date_from or "",
+        date_to=date_to or "",
+        sort=sort,
+        direction=direction,
     )
 
 

src/web/routes/review.py

@@ -0,0 +1,50 @@
+"""Review API routes for human validation of toxicity scores."""
+
+from flask import Blueprint, request, jsonify
+from ..db import get_cursor
+
+bp = Blueprint("review", __name__, url_prefix="/api/review")
+
+
+@bp.route("/submit", methods=["POST"])
+def submit_review():
+    """Submit a human review for a flagged item."""
+    data = request.get_json()
+
+    item_id = data.get("item_id")
+    source_type = data.get("source_type")  # 'post' or 'mention'
+    review_status = data.get("review_status")  # 'correct', 'incorrect', 'unsure'
+
+    if not all([item_id, source_type, review_status]):
+        return jsonify({"error": "Missing required fields"}), 400
+
+    if review_status not in ['correct', 'incorrect', 'unsure']:
+        return jsonify({"error": "Invalid review_status"}), 400
+
+    if source_type not in ['post', 'mention']:
+        return jsonify({"error": "Invalid source_type"}), 400
+
+    try:
+        with get_cursor() as cur:
+            if source_type == 'post':
+                cur.execute("""
+                    UPDATE toxicity_scores
+                    SET human_reviewed = true,
+                        review_status = %s,
+                        reviewed_at = NOW()
+                    WHERE uri = %s
+                """, (review_status, item_id))
+            else:  # mention
+                cur.execute("""
+                    UPDATE mention_toxicity_scores mts
+                    SET human_reviewed = true,
+                        review_status = %s,
+                        reviewed_at = NOW()
+                    FROM mentions m
+                    WHERE mts.mention_id = m.id AND m.post_uri = %s
+                """, (review_status, item_id))
+
+        return jsonify({"success": True, "message": "Review submitted"}), 200
+
+    except Exception as e:
+        return jsonify({"error": str(e)}), 500

src/web/templates/account_toxicity.html

@@ -44,8 +44,10 @@
           <tr>
             <th>{{ sort_header('handle', 'Account') }}</th>
             <th>{{ sort_header('post_tox', 'Avg Post Toxicity') }}</th>
+            <th>Max Post Toxicity</th>
             <th>{{ sort_header('flagged_posts', 'Flagged Posts') }}</th>
             <th>{{ sort_header('mention_tox', 'Avg Mention Toxicity') }}</th>
+            <th>Max Mention Toxicity</th>
             <th>{{ sort_header('flagged_mentions', 'Flagged Mentions') }}</th>
           </tr>
         </thead>
@@ -80,12 +82,32 @@
               </div>
             </td>
 
+            <!-- Max Post Toxicity -->
+            <td class="col-max-score">
+              {% if account.max_post_tox > 0 %}
+              <a href="{{ url_for('analysis.flagged', content_type='post', account_did=account.did, threshold=(account.max_post_tox - 0.001)|round(3)) }}" class="max-score-link">
+                <div class="max-score-badge {% if account.max_post_tox >= 0.5 %}high{% elif account.max_post_tox >= 0.3 %}medium{% else %}low{% endif %}">
+                  {{ "%.2f" | format(account.max_post_tox) }}
+                </div>
+              </a>
+              {% else %}
+              <span class="text-muted">—</span>
+              {% endif %}
+            </td>
+
             <!-- Flagged Posts Count -->
             <td class="col-count">
+              {% if account.flagged_posts > 0 %}
+              <a href="{{ url_for('analysis.flagged', content_type='post', account_did=account.did) }}" class="count-badge count-badge-link">
+                {{ account.flagged_posts | format_number }}
+                <span class="count-total">/ {{ account.scored_posts | format_number }}</span>
+              </a>
+              {% else %}
               <span class="count-badge">
                 {{ account.flagged_posts | format_number }}
                 <span class="count-total">/ {{ account.scored_posts | format_number }}</span>
               </span>
+              {% endif %}
             </td>
 
             <!-- Avg Mention Toxicity -->
@@ -104,12 +126,32 @@
               </div>
             </td>
 
+            <!-- Max Mention Toxicity -->
+            <td class="col-max-score">
+              {% if account.max_mention_tox > 0 %}
+              <a href="{{ url_for('analysis.flagged', content_type='mention', account_did=account.did, threshold=(account.max_mention_tox - 0.001)|round(3)) }}" class="max-score-link">
+                <div class="max-score-badge {% if account.max_mention_tox >= 0.5 %}high{% elif account.max_mention_tox >= 0.3 %}medium{% else %}low{% endif %}">
+                  {{ "%.2f" | format(account.max_mention_tox) }}
+                </div>
+              </a>
+              {% else %}
+              <span class="text-muted">—</span>
+              {% endif %}
+            </td>
+
             <!-- Flagged Mentions Count -->
             <td class="col-count">
+              {% if account.flagged_mentions > 0 %}
+              <a href="{{ url_for('analysis.flagged', content_type='mention', account_did=account.did) }}" class="count-badge count-badge-link">
+                {{ account.flagged_mentions | format_number }}
+                <span class="count-total">/ {{ account.scored_mentions | format_number }}</span>
+              </a>
+              {% else %}
               <span class="count-badge">
                 {{ account.flagged_mentions | format_number }}
                 <span class="count-total">/ {{ account.scored_mentions | format_number }}</span>
               </span>
+              {% endif %}
             </td>
           </tr>
           {% endfor %}
@@ -274,11 +316,78 @@
     width: 180px;
   }
 
+  .col-max-score {
+    width: 100px;
+    text-align: center;
+  }
+
   .col-count {
     width: 140px;
     text-align: center;
   }
 
+  /* Max Score Badge */
+  .max-score-link {
+    text-decoration: none;
+    display: inline-block;
+  }
+
+  .max-score-badge {
+    display: inline-block;
+    padding: 0.5rem 0.85rem;
+    border-radius: 0.375rem;
+    font-weight: 700;
+    font-size: 0.95rem;
+    transition: all 0.2s ease;
+    cursor: pointer;
+  }
+
+  .max-score-badge.low {
+    background: rgba(46, 204, 113, 0.2);
+    color: var(--tox-low);
+    border: 1px solid rgba(46, 204, 113, 0.4);
+  }
+
+  .max-score-badge.medium {
+    background: rgba(243, 156, 18, 0.2);
+    color: var(--tox-medium);
+    border: 1px solid rgba(243, 156, 18, 0.4);
+  }
+
+  .max-score-badge.high {
+    background: rgba(231, 76, 60, 0.2);
+    color: var(--tox-high);
+    border: 1px solid rgba(231, 76, 60, 0.4);
+  }
+
+  .max-score-link:hover .max-score-badge {
+    transform: translateY(-2px);
+    box-shadow: 0 4px 12px rgba(0, 0, 0, 0.3);
+  }
+
+  .max-score-link:hover .max-score-badge.low {
+    background: rgba(46, 204, 113, 0.3);
+    border-color: var(--tox-low);
+  }
+
+  .max-score-link:hover .max-score-badge.medium {
+    background: rgba(243, 156, 18, 0.3);
+    border-color: var(--tox-medium);
+  }
+
+  .max-score-link:hover .max-score-badge.high {
+    background: rgba(231, 76, 60, 0.3);
+    border-color: var(--tox-high);
+  }
+
+  .max-score-link:active .max-score-badge {
+    transform: translateY(0);
+  }
+
+  .text-muted {
+    color: rgba(255, 255, 255, 0.3);
+  }
+
   /* Account Info */
   .account-info {
     display: flex;
@@ -381,6 +490,22 @@
     font-size: 0.85rem;
   }
 
+  .count-badge-link {
+    text-decoration: none;
+    transition: all 0.2s ease;
+    cursor: pointer;
+  }
+
+  .count-badge-link:hover {
+    background: rgba(0, 180, 216, 0.2);
+    transform: translateY(-1px);
+    box-shadow: 0 2px 8px rgba(0, 180, 216, 0.3);
+  }
+
+  .count-badge-link:active {
+    transform: translateY(0);
+  }
+
   .count-total {
     color: rgba(255, 255, 255, 0.5);
     font-weight: 400;
@@ -538,7 +663,7 @@
   }
 </style>
 
-<script src="https://cdnjs.cloudflare.com/ajax/libs/Chart.js/4.4.7/chart.umd.min.js"></script>
+<script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.7/dist/chart.umd.min.js"></script>
 <script>
   document.addEventListener('DOMContentLoaded', function() {
     const chartData = {{ top_targeted_json | safe }};
@@ -548,6 +673,10 @@
       const scores = chartData.map(item => item.avg_mention_tox);
       const flagged = chartData.map(item => item.flagged_mentions);
 
+      // Calculate dynamic max for better readability
+      const maxScore = Math.max(...scores);
+      const dynamicMax = Math.min(1, Math.max(0.1, maxScore * 1.2)); // At least 10%, at most 100%, with 20% padding
+
       const ctx = document.getElementById('toxicity-chart');
       if (ctx) {
         new Chart(ctx, {
@@ -580,6 +709,12 @@
                 display: false
               },
               tooltip: {
+                backgroundColor: '#0f3460',
+                titleColor: 'rgba(224, 224, 224, 1)',
+                bodyColor: 'rgba(224, 224, 224, 0.9)',
+                borderColor: '#00b4d8',
+                borderWidth: 1,
+                padding: 10,
                 callbacks: {
                   afterLabel: function(context) {
                     return 'Flagged: ' + flagged[context.dataIndex];
@@ -590,7 +725,7 @@
             scales: {
               x: {
                 beginAtZero: true,
-                max: 1,
+                max: dynamicMax,
                 ticks: {
                   color: 'rgba(224, 224, 224, 0.7)',
                   callback: function(value) {
@@ -600,6 +735,15 @@
                 grid: {
                   color: 'rgba(255, 255, 255, 0.1)',
                   drawBorder: false
+                },
+                title: {
+                  display: true,
+                  text: 'Average Mention Toxicity',
+                  color: 'rgba(224, 224, 224, 0.9)',
+                  font: {
+                    size: 12,
+                    weight: 'bold'
+                  }
                 }
               },
               y: {

src/web/templates/analysis.html

@@ -429,7 +429,7 @@
 </div>
 
 <!-- Chart.js Script -->
-<script src="https://cdnjs.cloudflare.com/ajax/libs/Chart.js/4.4.7/chart.umd.min.js"></script>
+<script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.7/dist/chart.umd.min.js"></script>
 <script>
   // Chart color scheme
   const chartColors = {
@@ -619,6 +619,29 @@
     const categoriesData = {{ categories_json | safe }};
     const categoryNames = {{ categories | tojson | safe }};
 
+    // Calculate dynamic max for better readability with smart scaling
+    const categoryValues = categoryNames.map(cat => categoriesData[cat] || 0);
+    const maxCategoryValue = Math.max(...categoryValues);
+
+    // Smart scaling: round up to next nice number with 20% padding
+    let dynamicMax;
+    const paddedMax = maxCategoryValue * 1.2;
+    if (paddedMax <= 0.01) {
+      dynamicMax = 0.01; // 1%
+    } else if (paddedMax <= 0.02) {
+      dynamicMax = 0.02; // 2%
+    } else if (paddedMax <= 0.05) {
+      dynamicMax = 0.05; // 5%
+    } else if (paddedMax <= 0.1) {
+      dynamicMax = 0.1; // 10%
+    } else if (paddedMax <= 0.2) {
+      dynamicMax = 0.2; // 20%
+    } else if (paddedMax <= 0.5) {
+      dynamicMax = 0.5; // 50%
+    } else {
+      dynamicMax = 1.0; // 100%
+    }
+
     const categoriesCtx = document.getElementById('categoriesChart').getContext('2d');
     const categoriesChart = new Chart(categoriesCtx, {
       type: 'bar',
@@ -627,7 +650,7 @@
         datasets: [
           {
             label: 'Average Toxicity Score',
-            data: categoryNames.map(cat => categoriesData[cat] || 0),
+            data: categoryValues,
             backgroundColor: chartColors.categories,
             borderColor: chartColors.categories.map(c => c.replace('0.', '1.')),
             borderWidth: 1.5,
@@ -660,7 +683,7 @@
         scales: {
           x: {
             min: 0,
-            max: 1,
+            max: dynamicMax,
             grid: {
               color: chartColors.gridLine,
               drawBorder: false
@@ -671,7 +694,14 @@
                 size: 11
               },
               callback: function(value) {
-                return (value * 100).toFixed(0) + '%';
+                // Use more decimal places for small values
+                if (dynamicMax <= 0.01) {
+                  return (value * 100).toFixed(2) + '%';
+                } else if (dynamicMax <= 0.1) {
+                  return (value * 100).toFixed(1) + '%';
+                } else {
+                  return (value * 100).toFixed(0) + '%';
+                }
               }
             },
             title: {

src/web/templates/base.html

@@ -684,5 +684,6 @@
   <footer class="footer">
     <span>Bluesky Collector</span>
   </footer>
+  {% block extra_js %}{% endblock %}
 </body>
 </html>

src/web/templates/flagged.html

@@ -48,22 +48,55 @@
         <input type="number" id="threshold" name="threshold" min="0.0" max="1.0" step="0.1" value="{{ threshold or 0.5 }}" class="filter-input" placeholder="0.5">
       </div>
 
+      <div class="filter-group">
+        <label for="review-status">Review Status:</label>
+        <select id="review-status" name="review_status" class="filter-select">
+          <option value="">All</option>
+          <option value="unreviewed" {% if review_status == 'unreviewed' %}selected{% endif %}>Unreviewed</option>
+          <option value="correct" {% if review_status == 'correct' %}selected{% endif %}>✓ Correct</option>
+          <option value="incorrect" {% if review_status == 'incorrect' %}selected{% endif %}>✗ Incorrect</option>
+          <option value="unsure" {% if review_status == 'unsure' %}selected{% endif %}>? Unsure</option>
+        </select>
+      </div>
+
+      <div class="filter-group">
+        <label for="date-from">From Date:</label>
+        <input type="date" id="date-from" name="date_from" value="{{ date_from or '' }}" class="filter-input">
+      </div>
+
+      <div class="filter-group">
+        <label for="date-to">To Date:</label>
+        <input type="date" id="date-to" name="date_to" value="{{ date_to or '' }}" class="filter-input">
+      </div>
+
       <button type="submit" class="btn-apply">Apply Filters</button>
     </form>
   </div>
 
   <!-- Content Table -->
   {% if items %}
+
+  {% macro sort_header(col, label) %}
+    {% set new_dir = 'asc' if (sort == col and direction == 'desc') else 'desc' %}
+    <a href="{{ url_for('analysis.flagged', content_type=content_type, category=category, account_did=account_did, threshold=threshold, review_status=review_status, date_from=date_from, date_to=date_to, sort=col, dir=new_dir) }}" class="sort-link">
+      {{ label }}
+      {% if sort == col %}
+        <span class="sort-icon">{{ '▼' if direction == 'desc' else '▲' }}</span>
+      {% endif %}
+    </a>
+  {% endmacro %}
+
   <div class="table-wrapper">
     <table class="flagged-table">
       <thead>
         <tr>
           <th>Type</th>
-          <th>Author</th>
+          <th>{{ sort_header('author_handle', 'Author') }}</th>
           <th>Content</th>
-          <th>Score</th>
+          <th>{{ sort_header('overall', 'Score') }}</th>
           <th>Category</th>
-          <th>Created</th>
+          <th>{{ sort_header('created_at', 'Created') }}</th>
+          <th>Review</th>
         </tr>
       </thead>
       <tbody>
@@ -106,7 +139,15 @@
               {{ item.text | truncate_text(200) }}
             </a>
             {% else %}
-            <span class="content-text">{{ item.text | truncate_text(200) }}</span>
+            {# Convert at://did:plc:xxx/app.bsky.feed.post/yyy to https://bsky.app/profile/handle/post/yyy #}
+            {% set uri_parts = item.item_id.replace('at://', '').split('/') %}
+            {% if uri_parts|length >= 3 and item.author_handle %}
+              <a href="https://bsky.app/profile/{{ item.author_handle }}/post/{{ uri_parts[2] }}" target="_blank" rel="noopener" class="content-link">
+                {{ item.text | truncate_text(200) }}
+              </a>
+            {% else %}
+              <span class="content-text">{{ item.text | truncate_text(200) }}</span>
+            {% endif %}
             {% endif %}
           </td>
 
@@ -141,6 +182,15 @@
               {{ item.created_at | time_ago }}
             </span>
           </td>
+
+          <!-- Review Buttons -->
+          <td class="col-review">
+            <div class="review-buttons" data-item-id="{{ item.item_id }}" data-source-type="{{ item.source_type }}" data-current-status="{{ item.review_status or '' }}">
+              <button class="btn-review btn-correct {% if item.review_status == 'correct' %}active{% endif %}" data-status="correct" title="Mark as correctly flagged">✓</button>
+              <button class="btn-review btn-incorrect {% if item.review_status == 'incorrect' %}active{% endif %}" data-status="incorrect" title="Mark as incorrectly flagged">✗</button>
+              <button class="btn-review btn-unsure {% if item.review_status == 'unsure' %}active{% endif %}" data-status="unsure" title="Mark as unsure">?</button>
+            </div>
+          </td>
         </tr>
         {% endfor %}
       </tbody>
@@ -151,15 +201,15 @@
   {% if total_pages > 1 %}
   <div class="pagination">
     {% if page > 1 %}
-    <a href="{{ url_for('analysis.flagged', page=1, content_type=content_type, category=category, account_did=account_did, threshold=threshold) }}" class="btn-pagination">First</a>
+    <a href="{{ url_for('analysis.flagged', page=1, content_type=content_type, category=category, account_did=account_did, threshold=threshold, review_status=review_status, date_from=date_from, date_to=date_to, sort=sort, dir=direction) }}" class="btn-pagination">First</a>
-    <a href="{{ url_for('analysis.flagged', page=page-1, content_type=content_type, category=category, account_did=account_did, threshold=threshold) }}" class="btn-pagination">Previous</a>
+    <a href="{{ url_for('analysis.flagged', page=page-1, content_type=content_type, category=category, account_did=account_did, threshold=threshold, review_status=review_status, date_from=date_from, date_to=date_to, sort=sort, dir=direction) }}" class="btn-pagination">Previous</a>
     {% endif %}
 
     <span class="pagination-info">Page {{ page }} of {{ total_pages }}</span>
 
     {% if page < total_pages %}
-    <a href="{{ url_for('analysis.flagged', page=page+1, content_type=content_type, category=category, account_did=account_did, threshold=threshold) }}" class="btn-pagination">Next</a>
+    <a href="{{ url_for('analysis.flagged', page=page+1, content_type=content_type, category=category, account_did=account_did, threshold=threshold, review_status=review_status, date_from=date_from, date_to=date_to, sort=sort, dir=direction) }}" class="btn-pagination">Next</a>
-    <a href="{{ url_for('analysis.flagged', page=total_pages, content_type=content_type, category=category, account_did=account_did, threshold=threshold) }}" class="btn-pagination">Last</a>
+    <a href="{{ url_for('analysis.flagged', page=total_pages, content_type=content_type, category=category, account_did=account_did, threshold=threshold, review_status=review_status, date_from=date_from, date_to=date_to, sort=sort, dir=direction) }}" class="btn-pagination">Last</a>
     {% endif %}
   </div>
   {% endif %}
@@ -325,6 +375,25 @@
     white-space: nowrap;
   }
 
+  /* Sort Links */
+  .sort-link {
+    color: var(--dark-text);
+    text-decoration: none;
+    display: inline-flex;
+    align-items: center;
+    gap: 0.25rem;
+    transition: color 0.2s ease;
+  }
+
+  .sort-link:hover {
+    color: var(--accent-primary);
+  }
+
+  .sort-icon {
+    font-size: 0.8rem;
+    opacity: 0.7;
+  }
+
   .flagged-table td {
     padding: 1rem;
     border-bottom: 1px solid rgba(255, 255, 255, 0.05);
@@ -360,6 +429,10 @@
     width: 120px;
   }
 
+  .col-review {
+    width: 130px;
+  }
+
   /* Badges */
   .badge {
     display: inline-block;
@@ -478,6 +551,60 @@
     color: rgba(255, 255, 255, 0.3);
   }
 
+  /* Review Buttons */
+  .review-buttons {
+    display: flex;
+    gap: 0.25rem;
+    justify-content: center;
+  }
+
+  .btn-review {
+    background: var(--dark-bg);
+    border: 1px solid rgba(255, 255, 255, 0.2);
+    color: var(--dark-text);
+    width: 32px;
+    height: 32px;
+    border-radius: 0.25rem;
+    font-size: 1rem;
+    cursor: pointer;
+    transition: all 0.2s ease;
+    display: flex;
+    align-items: center;
+    justify-content: center;
+  }
+
+  .btn-review:hover {
+    transform: translateY(-2px);
+    border-color: rgba(255, 255, 255, 0.4);
+  }
+
+  .btn-correct:hover, .btn-correct.active {
+    background: var(--tox-low);
+    border-color: var(--tox-low);
+    color: var(--dark-bg);
+  }
+
+  .btn-incorrect:hover, .btn-incorrect.active {
+    background: var(--tox-high);
+    border-color: var(--tox-high);
+    color: var(--dark-bg);
+  }
+
+  .btn-unsure:hover, .btn-unsure.active {
+    background: var(--tox-medium);
+    border-color: var(--tox-medium);
+    color: var(--dark-bg);
+  }
+
+  .btn-review.active {
+    font-weight: 700;
+    transform: none;
+  }
+
+  .btn-review.active:hover {
+    opacity: 0.8;
+  }
+
   /* Empty State */
   .empty-state {
     text-align: center;
@@ -592,3 +719,67 @@
   }
 </style>
 {% endblock %}
+
+{% block extra_js %}
+<script>
+document.addEventListener('DOMContentLoaded', function() {
+  // Handle review button clicks
+  document.querySelectorAll('.btn-review').forEach(button => {
+    button.addEventListener('click', async function() {
+      const buttonContainer = this.parentElement;
+      const itemId = buttonContainer.dataset.itemId;
+      const sourceType = buttonContainer.dataset.sourceType;
+      const currentStatus = buttonContainer.dataset.currentStatus;
+      const clickedStatus = this.dataset.status;
+
+      // If clicking the same status, do nothing (it's already set)
+      if (currentStatus === clickedStatus) {
+        return;
+      }
+
+      // Disable all buttons while processing
+      buttonContainer.querySelectorAll('.btn-review').forEach(btn => btn.disabled = true);
+
+      try {
+        const response = await fetch('/api/review/submit', {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+          },
+          body: JSON.stringify({
+            item_id: itemId,
+            source_type: sourceType,
+            review_status: clickedStatus,
+          }),
+        });
+
+        if (response.ok) {
+          // Update current status
+          buttonContainer.dataset.currentStatus = clickedStatus;
+
+          // Remove active class from all buttons
+          buttonContainer.querySelectorAll('.btn-review').forEach(btn => {
+            btn.classList.remove('active');
+          });
+
+          // Add active class to clicked button
+          this.classList.add('active');
+
+          // Re-enable all buttons
+          buttonContainer.querySelectorAll('.btn-review').forEach(btn => btn.disabled = false);
+        } else {
+          // Re-enable buttons on error
+          buttonContainer.querySelectorAll('.btn-review').forEach(btn => btn.disabled = false);
+          alert('Failed to submit review. Please try again.');
+        }
+      } catch (error) {
+        // Re-enable buttons on error
+        buttonContainer.querySelectorAll('.btn-review').forEach(btn => btn.disabled = false);
+        console.error('Error submitting review:', error);
+        alert('Failed to submit review. Please try again.');
+      }
+    });
+  });
+});
+</script>
+{% endblock %}