chore: rename auto-scraper to auto-fetcher

Improves the perception of ethics

frontend/src/App.tsx

@@ -5,7 +5,7 @@ import DatasetsPage from "./pages/Datasets";
 import DatasetStatusPage from "./pages/DatasetStatus";
 import LoginPage from "./pages/Login";
 import UploadPage from "./pages/Upload";
-import AutoScrapePage from "./pages/AutoScrape";
+import AutoFetchPage from "./pages/AutoFetch";
 import StatPage from "./pages/Stats";
 import { getDocumentTitle } from "./utils/documentTitle";
 import DatasetEditPage from "./pages/DatasetEdit";
@@ -23,7 +23,7 @@ function App() {
         <Route path="/" element={<Navigate to="/login" replace />} />
         <Route path="/login" element={<LoginPage />} />
         <Route path="/upload" element={<UploadPage />} />
-        <Route path="/auto-scrape" element={<AutoScrapePage />} />
+        <Route path="/auto-fetch" element={<AutoFetchPage />} />
         <Route path="/datasets" element={<DatasetsPage />} />
         <Route path="/dataset/:datasetId/status" element={<DatasetStatusPage />} />
         <Route path="/dataset/:datasetId/stats" element={<StatPage />} />

frontend/src/pages/AutoFetch.tsx

@@ -37,7 +37,7 @@ const supportsSearch = (source?: SourceOption): boolean =>
 const supportsCategories = (source?: SourceOption): boolean =>
   Boolean(source?.categories_enabled ?? source?.categoriesEnabled);
 
-const AutoScrapePage = () => {
+const AutoFetchPage = () => {
   const navigate = useNavigate();
   const [datasetName, setDatasetName] = useState("");
   const [sourceOptions, setSourceOptions] = useState<SourceOption[]>([]);
@@ -106,11 +106,11 @@ const AutoScrapePage = () => {
     );
   };
 
-  const autoScrape = async () => {
+  const autoFetch = async () => {
     const token = localStorage.getItem("access_token");
     if (!token) {
       setHasError(true);
-      setReturnMessage("You must be signed in to auto scrape a dataset.");
+      setReturnMessage("You must be signed in to auto fetch a dataset.");
       return;
     }
 
@@ -243,7 +243,7 @@ const AutoScrapePage = () => {
       setReturnMessage("");
 
       const response = await axios.post(
-        `${API_BASE_URL}/datasets/scrape`,
+        `${API_BASE_URL}/datasets/fetch`,
         requestBody,
         {
           headers: {
@@ -255,7 +255,7 @@ const AutoScrapePage = () => {
       const datasetId = Number(response.data.dataset_id);
 
       setReturnMessage(
-        `Auto scrape queued successfully (dataset #${datasetId}). Redirecting to processing status...`,
+        `Auto fetch queued successfully (dataset #${datasetId}). Redirecting to processing status...`,
       );
 
       setTimeout(() => {
@@ -267,11 +267,11 @@ const AutoScrapePage = () => {
         const message = String(
           requestError.response?.data?.error ||
             requestError.message ||
-            "Auto scrape failed.",
+            "Auto fetch failed.",
         );
-        setReturnMessage(`Auto scrape failed: ${message}`);
+        setReturnMessage(`Auto fetch failed: ${message}`);
       } else {
-        setReturnMessage("Auto scrape failed due to an unexpected error.");
+        setReturnMessage("Auto fetch failed due to an unexpected error.");
       }
     } finally {
       setIsSubmitting(false);
@@ -283,9 +283,9 @@ const AutoScrapePage = () => {
       <div style={styles.containerWide}>
         <div style={{ ...styles.card, ...styles.headerBar }}>
           <div>
-            <h1 style={styles.sectionHeaderTitle}>Auto Scrape Dataset</h1>
+            <h1 style={styles.sectionHeaderTitle}>Auto Fetch Dataset</h1>
             <p style={styles.sectionHeaderSubtitle}>
-              Select sources and scrape settings, then queue processing
+              Select sources and fetch settings, then queue processing
               automatically.
             </p>
             <p
@@ -295,7 +295,7 @@ const AutoScrapePage = () => {
                 color: "#9a6700",
               }}
             >
-              Warning: Scraping more than 250 posts from any single site can
+              Warning: Fetching more than 250 posts from any single site can
               take hours due to rate limits.
             </p>
           </div>
@@ -305,10 +305,10 @@ const AutoScrapePage = () => {
               ...styles.buttonPrimary,
               opacity: isSubmitting || isLoadingSources ? 0.75 : 1,
             }}
-            onClick={autoScrape}
+            onClick={autoFetch}
             disabled={isSubmitting || isLoadingSources}
           >
-            {isSubmitting ? "Queueing..." : "Auto Scrape and Analyze"}
+            {isSubmitting ? "Queueing..." : "Auto Fetch and Analyze"}
           </button>
         </div>
 
@@ -527,4 +527,4 @@ const AutoScrapePage = () => {
   );
 };
 
-export default AutoScrapePage;
+export default AutoFetchPage;

frontend/src/pages/Datasets.tsx

@@ -108,9 +108,9 @@ const DatasetsPage = () => {
             <button
               type="button"
               style={styles.buttonSecondary}
-              onClick={() => navigate("/auto-scrape")}
+              onClick={() => navigate("/auto-fetch")}
             >
-              Auto Scrape Dataset
+              Auto Fetch Dataset
             </button>
           </div>
         </div>

frontend/src/pages/Stats.tsx

@@ -66,45 +66,88 @@ const EMPTY_EXPLORER_STATE: ExplorerState = {
   error: "",
 };
 
-const getExplorerRecordIdentity = (record: DatasetRecord) =>
+const parseJsonLikePayload = (value: string): unknown => {
-  JSON.stringify({
+  const normalized = value
-    post_id: record.post_id ?? null,
+    .replace(/\uFEFF/g, "")
-    parent_id: record.parent_id ?? null,
+    .replace(/,\s*([}\]])/g, "$1")
-    reply_to: record.reply_to ?? null,
+    .replace(/(:\s*)(NaN|Infinity|-Infinity)\b/g, "$1null")
-    author: record.author ?? null,
+    .replace(/(\[\s*)(NaN|Infinity|-Infinity)\b/g, "$1null")
-    type: record.type ?? null,
+    .replace(/(,\s*)(NaN|Infinity|-Infinity)\b/g, "$1null")
-    timestamp: record.timestamp ?? null,
+    .replace(/(:\s*)None\b/g, "$1null")
-    dt: record.dt ?? null,
+    .replace(/(:\s*)True\b/g, "$1true")
-    title: record.title ?? null,
+    .replace(/(:\s*)False\b/g, "$1false")
-    content: record.content ?? null,
+    .replace(/(\[\s*)None\b/g, "$1null")
-    source: record.source ?? null,
+    .replace(/(\[\s*)True\b/g, "$1true")
-    topic: record.topic ?? null,
+    .replace(/(\[\s*)False\b/g, "$1false")
-  });
+    .replace(/(,\s*)None\b/g, "$1null")
+    .replace(/(,\s*)True\b/g, "$1true")
+    .replace(/(,\s*)False\b/g, "$1false");
 
-const dedupeExplorerRecords = (records: DatasetRecord[]) => {
+  return JSON.parse(normalized);
-  const uniqueRecords: DatasetRecord[] = [];
+};
-  const seen = new Set<string>();
 
-  for (const record of records) {
+const parseRecordStringPayload = (payload: string): DatasetRecord[] | null => {
-    const identity = getExplorerRecordIdentity(record);
+  const trimmed = payload.trim();
-    if (seen.has(identity)) {
+  if (!trimmed) {
-      continue;
+    return [];
   }
 
-    seen.add(identity);
+  try {
-    uniqueRecords.push(record);
+    return normalizeRecordPayload(parseJsonLikePayload(trimmed));
+  } catch {
+    // Continue with additional fallback formats below.
   }
 
-  return uniqueRecords;
+  const ndjsonLines = trimmed
+    .split(/\r?\n/)
+    .map((line) => line.trim())
+    .filter(Boolean);
+  if (ndjsonLines.length > 0) {
+    try {
+      return ndjsonLines.map((line) => parseJsonLikePayload(line)) as DatasetRecord[];
+    } catch {
+      // Continue with wrapped JSON extraction.
+    }
+  }
+
+  const bracketStart = trimmed.indexOf("[");
+  const bracketEnd = trimmed.lastIndexOf("]");
+  if (bracketStart !== -1 && bracketEnd > bracketStart) {
+    const candidate = trimmed.slice(bracketStart, bracketEnd + 1);
+    try {
+      return normalizeRecordPayload(parseJsonLikePayload(candidate));
+    } catch {
+      // Continue with object extraction.
+    }
+  }
+
+  const braceStart = trimmed.indexOf("{");
+  const braceEnd = trimmed.lastIndexOf("}");
+  if (braceStart !== -1 && braceEnd > braceStart) {
+    const candidate = trimmed.slice(braceStart, braceEnd + 1);
+    try {
+      return normalizeRecordPayload(parseJsonLikePayload(candidate));
+    } catch {
+      return null;
+    }
+  }
+
+  return null;
 };
 
 const normalizeRecordPayload = (payload: unknown): DatasetRecord[] => {
   if (typeof payload === "string") {
-    try {
+    const parsed = parseRecordStringPayload(payload);
-      return normalizeRecordPayload(JSON.parse(payload));
+    if (parsed) {
-    } catch {
+      return parsed;
-      throw new Error("Corpus endpoint returned a non-JSON string payload.");
     }
+
+    const preview = payload.trim().slice(0, 120).replace(/\s+/g, " ");
+    throw new Error(
+      `Corpus endpoint returned a non-JSON string payload.${
+        preview ? ` Response preview: ${preview}` : ""
+      }`,
+    );
   }
 
   if (
@@ -265,9 +308,7 @@ const StatPage = () => {
       },
     );
 
-    const normalizedRecords = dedupeExplorerRecords(
+    const normalizedRecords = normalizeRecordPayload(response.data);
-      normalizeRecordPayload(response.data),
-    );
 
     setAllRecords(normalizedRecords);
     setAllRecordsKey(filterKey);
@@ -288,9 +329,7 @@ const StatPage = () => {
     try {
       const records = await ensureFilteredRecords();
       const context = buildExplorerContext(records);
-      const matched = dedupeExplorerRecords(
+      const matched = records.filter((record) => spec.matcher(record, context));
-        records.filter((record) => spec.matcher(record, context)),
-      );
       matched.sort((a, b) => {
         const aValue = String(a.dt ?? a.date ?? a.timestamp ?? "");
         const bValue = String(b.dt ?? b.date ?? b.timestamp ?? "");

frontend/src/utils/documentTitle.ts

@@ -3,7 +3,7 @@ const DEFAULT_TITLE = "Ethnograph View";
 const STATIC_TITLES: Record<string, string> = {
   "/login": "Sign In",
   "/upload": "Upload Dataset",
-  "/auto-scrape": "Auto Scrape Dataset",
+  "/auto-fetch": "Auto Fetch Dataset",
   "/datasets": "My Datasets",
 };
 

server/analysis/linguistic.py

@@ -1,17 +1,30 @@
-import pandas as pd
 import re
-
 from collections import Counter
-from itertools import islice
+from dataclasses import dataclass
+
+import pandas as pd
+
+
+@dataclass(frozen=True)
+class NGramConfig:
+    min_token_length: int = 3
+    min_count: int = 2
+    max_results: int = 100
 
 
 class LinguisticAnalysis:
     def __init__(self, word_exclusions: set[str]):
         self.word_exclusions = word_exclusions
+        self.ngram_config = NGramConfig()
 
-    def _tokenize(self, text: str):
+    def _tokenize(self, text: str, *, include_exclusions: bool = False) -> list[str]:
-        tokens = re.findall(r"\b[a-z]{3,}\b", text)
+        pattern = rf"\b[a-z]{{{self.ngram_config.min_token_length},}}\b"
-        return [t for t in tokens if t not in self.word_exclusions]
+        tokens = re.findall(pattern, text)
+
+        if include_exclusions:
+            return tokens
+
+        return [token for token in tokens if token not in self.word_exclusions]
 
     def _clean_text(self, text: str) -> str:
         text = re.sub(r"http\S+", "", text)  # remove URLs
@@ -21,13 +34,24 @@ class LinguisticAnalysis:
         text = re.sub(r"\S+\.(jpg|jpeg|png|webp|gif)", "", text)
         return text
 
+    def _content_texts(self, df: pd.DataFrame) -> pd.Series:
+        return df["content"].dropna().astype(str).apply(self._clean_text).str.lower()
+
+    def _valid_ngram(self, tokens: tuple[str, ...]) -> bool:
+        if any(token in self.word_exclusions for token in tokens):
+            return False
+
+        if len(set(tokens)) == 1:
+            return False
+
+        return True
+
     def word_frequencies(self, df: pd.DataFrame, limit: int = 100) -> list[dict]:
-        texts = df["content"].dropna().astype(str).str.lower()
+        texts = self._content_texts(df)
 
         words = []
         for text in texts:
-            tokens = re.findall(r"\b[a-z]{3,}\b", text)
+            words.extend(self._tokenize(text))
-            words.extend(w for w in tokens if w not in self.word_exclusions)
 
         counts = Counter(words)
 
@@ -40,25 +64,39 @@ class LinguisticAnalysis:
 
         return word_frequencies.to_dict(orient="records")
 
-    def ngrams(self, df: pd.DataFrame, n=2, limit=100):
+    def ngrams(self, df: pd.DataFrame, n: int = 2, limit: int | None = None) -> list[dict]:
-        texts = df["content"].dropna().astype(str).apply(self._clean_text).str.lower()
+        if n < 2:
+            raise ValueError("n must be at least 2")
+
+        texts = self._content_texts(df)
         all_ngrams = []
+        result_limit = limit or self.ngram_config.max_results
 
         for text in texts:
-            tokens = re.findall(r"\b[a-z]{3,}\b", text)
+            tokens = self._tokenize(text, include_exclusions=True)
 
-            # stop word removal causes strange behaviors in ngrams
+            if len(tokens) < n:
-            # tokens = [w for w in tokens if w not in self.word_exclusions]
+                continue
 
-            ngrams = zip(*(islice(tokens, i, None) for i in range(n)))
+            for index in range(len(tokens) - n + 1):
-            all_ngrams.extend([" ".join(ng) for ng in ngrams])
+                ngram_tokens = tuple(tokens[index : index + n])
+                if self._valid_ngram(ngram_tokens):
+                    all_ngrams.append(" ".join(ngram_tokens))
 
         counts = Counter(all_ngrams)
+        filtered_counts = [
+            (ngram, count)
+            for ngram, count in counts.items()
+            if count >= self.ngram_config.min_count
+        ]
+
+        if not filtered_counts:
+            return []
 
         return (
-            pd.DataFrame(counts.items(), columns=["ngram", "count"])
+            pd.DataFrame(filtered_counts, columns=["ngram", "count"])
-            .sort_values("count", ascending=False)
+            .sort_values(["count", "ngram"], ascending=[False, True])
-            .head(limit)
+            .head(result_limit)
             .to_dict(orient="records")
         )
 

server/analysis/stat_gen.py

@@ -89,39 +89,17 @@ class StatGen:
             df.to_json(orient="records", date_format="iso", date_unit="s")
         )
 
-    def _dedupe_records(self, records: list[dict]) -> list[dict]:
-        unique_records = []
-        seen = set()
-
-        for record in records:
-            key_data = {
-                "post_id": record.get("post_id"),
-                "parent_id": record.get("parent_id"),
-                "reply_to": record.get("reply_to"),
-                "author": record.get("author"),
-                "type": record.get("type"),
-                "timestamp": record.get("timestamp"),
-                "dt": record.get("dt"),
-                "title": record.get("title"),
-                "content": record.get("content"),
-                "source": record.get("source"),
-                "topic": record.get("topic"),
-            }
-            key = json.dumps(key_data, sort_keys=True, separators=(",", ":"))
-            if key in seen:
-                continue
-
-            seen.add(key)
-            unique_records.append(record)
-
-        return unique_records
-
     ## Public Methods
     def filter_dataset(self, df: pd.DataFrame, filters: dict | None = None) -> list[dict]:
         filtered_df = self._prepare_filtered_df(df, filters)
-        return self._dedupe_records(self._json_ready_records(filtered_df))
+        return self._json_ready_records(filtered_df)
 
-    def temporal(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def temporal(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return {
@@ -129,7 +107,12 @@ class StatGen:
             "weekday_hour_heatmap": self.temporal_analysis.heatmap(filtered_df),
         }
 
-    def linguistic(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def linguistic(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return {
@@ -139,7 +122,12 @@ class StatGen:
             "lexical_diversity": self.linguistic_analysis.lexical_diversity(filtered_df)
         }
 
-    def emotional(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def emotional(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return {
@@ -149,7 +137,12 @@ class StatGen:
             "emotion_by_source": self.emotional_analysis.emotion_by_source(filtered_df)
         }
 
-    def user(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def user(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return {
@@ -157,7 +150,12 @@ class StatGen:
             "users": self.user_analysis.per_user_analysis(filtered_df)
         }
 
-    def interactional(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def interactional(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return {
@@ -166,7 +164,12 @@ class StatGen:
             "conversation_concentration": self.interaction_analysis.conversation_concentration(filtered_df)
         }
 
-    def cultural(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def cultural(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return {
@@ -175,7 +178,12 @@ class StatGen:
             "avg_emotion_per_entity": self.cultural_analysis.get_avg_emotions_per_entity(filtered_df)
         }
 
-    def summary(self, df: pd.DataFrame, filters: dict | None = None) -> dict:
+    def summary(
+        self,
+        df: pd.DataFrame,
+        filters: dict | None = None,
+        dataset_id: int | None = None,
+    ) -> dict:
         filtered_df = self._prepare_filtered_df(df, filters)
 
         return self.summary_analysis.summary(filtered_df)

server/app.py

@@ -152,9 +152,9 @@ def get_dataset_sources():
     return jsonify(list_metadata)
 
 
-@app.route("/datasets/scrape", methods=["POST"])
+@app.route("/datasets/fetch", methods=["POST"])
 @jwt_required()
-def scrape_data():
+def fetch_data():
     data = request.get_json()
     connector_metadata = get_connector_metadata()
 
@@ -424,7 +424,7 @@ def get_linguistic_analysis(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.linguistic(dataset_content, filters)), 200
+        return jsonify(stat_gen.linguistic(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:
@@ -448,7 +448,7 @@ def get_emotional_analysis(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.emotional(dataset_content, filters)), 200
+        return jsonify(stat_gen.emotional(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:
@@ -472,7 +472,7 @@ def get_summary(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.summary(dataset_content, filters)), 200
+        return jsonify(stat_gen.summary(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:
@@ -496,7 +496,7 @@ def get_temporal_analysis(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.temporal(dataset_content, filters)), 200
+        return jsonify(stat_gen.temporal(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:
@@ -520,7 +520,7 @@ def get_user_analysis(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.user(dataset_content, filters)), 200
+        return jsonify(stat_gen.user(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:
@@ -544,7 +544,7 @@ def get_cultural_analysis(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.cultural(dataset_content, filters)), 200
+        return jsonify(stat_gen.cultural(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:
@@ -568,7 +568,7 @@ def get_interaction_analysis(dataset_id):
 
         dataset_content = dataset_manager.get_dataset_content(dataset_id)
         filters = get_request_filters()
-        return jsonify(stat_gen.interactional(dataset_content, filters)), 200
+        return jsonify(stat_gen.interactional(dataset_content, filters, dataset_id=dataset_id)), 200
     except NotAuthorisedException:
         return jsonify({"error": "User is not authorised to access this content"}), 403
     except NonExistentDatasetException:

server/connectors/boards_api.py

@@ -11,7 +11,7 @@ from server.connectors.base import BaseConnector
 
 logger = logging.getLogger(__name__)
 
-HEADERS = {"User-Agent": "Mozilla/5.0 (compatible; ForumScraper/1.0)"}
+HEADERS = {"User-Agent": "Mozilla/5.0 (compatible; ForumFetcher/1.0)"}
 
 
 class BoardsAPI(BaseConnector):

server/core/datasets.py

@@ -26,7 +26,34 @@ class DatasetManager:
     def get_dataset_content(self, dataset_id: int) -> pd.DataFrame:
         query = "SELECT * FROM events WHERE dataset_id = %s"
         result = self.db.execute(query, (dataset_id,), fetch=True)
-        return pd.DataFrame(result)
+        df = pd.DataFrame(result)
+        if df.empty:
+            return df
+
+        dedupe_columns = [
+            column
+            for column in [
+                "post_id",
+                "parent_id",
+                "reply_to",
+                "author",
+                "type",
+                "timestamp",
+                "dt",
+                "title",
+                "content",
+                "source",
+                "topic",
+            ]
+            if column in df.columns
+        ]
+
+        if dedupe_columns:
+            df = df.drop_duplicates(subset=dedupe_columns, keep="first")
+        else:
+            df = df.drop_duplicates(keep="first")
+
+        return df.reset_index(drop=True)
 
     def get_dataset_info(self, dataset_id: int) -> dict:
         query = "SELECT * FROM datasets WHERE id = %s"
@@ -52,6 +79,16 @@ class DatasetManager:
         if event_data.empty:
             return
 
+        dedupe_columns = [
+            column for column in ["id", "type", "source"] if column in event_data.columns
+        ]
+        if dedupe_columns:
+            event_data = event_data.drop_duplicates(subset=dedupe_columns, keep="first")
+        else:
+            event_data = event_data.drop_duplicates(keep="first")
+
+        self.delete_dataset_content(dataset_id)
+
         query = """
             INSERT INTO events (
                 dataset_id,