diff --git a/bincio/reextract_cmd.py b/bincio/reextract_cmd.py
index ff73287..8c68231 100644
--- a/bincio/reextract_cmd.py
+++ b/bincio/reextract_cmd.py
@@ -33,7 +33,9 @@ _GC_EVERY = 50  # call gc.collect() + malloc_trim every N activities
 @click.option("--data-dir", required=True, type=click.Path(), help="BAS data directory")
 @click.option("--handle", required=True, help="User handle to re-extract for")
 @click.option("--force", is_flag=True, default=False, help="Re-extract even if activity JSON already exists")
-def reextract_originals(data_dir: str, handle: str, force: bool) -> None:
+@click.option("--offset", default=0, type=int, help="Skip first N originals (for batch processing)")
+@click.option("--limit", default=0, type=int, help="Process at most N originals then stop (0 = all)")
+def reextract_originals(data_dir: str, handle: str, force: bool, offset: int, limit: int) -> None:
     """Re-extract activities from stored Strava originals (originals/strava/*.json).
 
     Prints one JSON object per line to stdout for streaming progress:
@@ -57,13 +59,22 @@ def reextract_originals(data_dir: str, handle: str, force: bool) -> None:
         _emit({"type": "error", "message": f"No Strava originals directory at {originals_dir}"})
         sys.exit(1)
 
-    original_files = sorted(originals_dir.glob("*.json"))
-    total = len(original_files)
-    if total == 0:
+    all_files = sorted(originals_dir.glob("*.json"))
+    if not all_files:
         _emit({"type": "error", "message": "No Strava originals found"})
         sys.exit(1)
 
-    _emit({"type": "status", "message": f"Found {total} originals, starting extraction…"})
+    # Apply offset/limit for batch processing
+    batch = all_files[offset:] if not limit else all_files[offset: offset + limit]
+    total_all = len(all_files)
+    total = len(batch)
+    original_files = batch
+
+    _emit({"type": "status", "message": (
+        f"Batch {offset + 1}–{offset + total} of {total_all}, starting extraction…"
+        if offset or limit else
+        f"Found {total_all} originals, starting extraction…"
+    )})
 
     # Load existing index to get owner info and existing summaries
     index_path = user_dir / "index.json"
diff --git a/bincio/serve/server.py b/bincio/serve/server.py
index 052d008..723f641 100644
--- a/bincio/serve/server.py
+++ b/bincio/serve/server.py
@@ -642,37 +642,62 @@ async def admin_reextract_originals(
     import sys as _sys
     bincio_exe = str(Path(_sys.executable).parent / "bincio")
     data_dir = str(_get_data_dir())
-    log.info("reextract[%s]: spawning subprocess via %s", handle, bincio_exe)
+
+    # Count originals so we can split into memory-safe batches.
+    total_originals = len(list(originals_dir.glob("*.json")))
+    # Each activity can briefly peak at ~10–30 MB; 100 per batch keeps RSS
+    # well under 3 GB even on a cheap VPS.
+    _BATCH = 100
+    log.info("reextract[%s]: %d originals, batch size %d, via %s",
+             handle, total_originals, _BATCH, bincio_exe)
 
     async def event_stream():
-        proc = await asyncio.create_subprocess_exec(
-            bincio_exe, "reextract-originals",
-            "--data-dir", data_dir,
-            "--handle", handle,
-            stdout=asyncio.subprocess.PIPE,
-            stderr=asyncio.subprocess.PIPE,
-        )
-        assert proc.stdout is not None
+        total_imported = total_skipped = total_errors = 0
+        offset = 0
 
-        async for raw_line in proc.stdout:
-            line = raw_line.decode(errors="replace").strip()
-            if not line:
-                continue
-            # Forward the JSON line as an SSE event
-            yield f"data: {line}\n\n"
+        while offset < total_originals:
+            limit = min(_BATCH, total_originals - offset)
+            proc = await asyncio.create_subprocess_exec(
+                bincio_exe, "reextract-originals",
+                "--data-dir", data_dir,
+                "--handle", handle,
+                "--offset", str(offset),
+                "--limit", str(limit),
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE,
+            )
+            assert proc.stdout is not None
 
-        await proc.wait()
-        stderr_out = b""
-        if proc.stderr:
-            stderr_out = await proc.stderr.read()
+            async for raw_line in proc.stdout:
+                line = raw_line.decode(errors="replace").strip()
+                if not line:
+                    continue
+                yield f"data: {line}\n\n"
+                try:
+                    evt = json.loads(line)
+                    if evt.get("type") == "done":
+                        total_imported += evt.get("imported", 0)
+                        total_skipped += evt.get("skipped", 0)
+                        total_errors += evt.get("errors", 0)
+                except Exception:
+                    pass
 
-        if proc.returncode != 0:
-            log.error("reextract[%s]: subprocess exited %d — stderr: %s",
-                      handle, proc.returncode, stderr_out.decode(errors="replace")[:500])
-            yield f"data: {json.dumps({'type': 'error', 'message': f'Process exited with code {proc.returncode}'})}\n\n"
-        else:
-            log.info("reextract[%s]: subprocess done, triggering rebuild", handle)
-            _trigger_rebuild(handle)
+            await proc.wait()
+            if proc.returncode != 0:
+                stderr_out = await proc.stderr.read() if proc.stderr else b""
+                log.error("reextract[%s]: batch offset=%d exited %d — stderr: %s",
+                          handle, offset, proc.returncode,
+                          stderr_out.decode(errors="replace")[:500])
+                yield f"data: {json.dumps({'type': 'error', 'message': f'Batch {offset}–{offset+limit} exited with code {proc.returncode}'})}\n\n"
+                return  # stop on batch failure
+
+            offset += limit
+
+        # All batches complete
+        log.info("reextract[%s]: all batches done — imported=%d skipped=%d errors=%d; triggering rebuild",
+                 handle, total_imported, total_skipped, total_errors)
+        _trigger_rebuild(handle)
+        yield f"data: {json.dumps({'type': 'done', 'imported': total_imported, 'skipped': total_skipped, 'errors': total_errors})}\n\n"
 
     return StreamingResponse(
         event_stream(),