adeelahmad · Oct 7, 2026 · Oct 7, 2026 · Oct 7, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -4,6 +4,10 @@ The backend (`fastapi_backend`) and the frontend (`nextjs-frontend`) are version
 
 ## Unreleased
 
+- **anytopdf can find objects.** Settings › Video › Object detector › anytopdf runs Lens's YOLOX model with
+  anytopdf's objects plugin (sandboxed, no network), so a lean server or a Raspberry Pi finds objects without ONNX
+  Runtime. Lens fetches anytopdf and the model on first use. YOLOX on ONNX Runtime stays the default
+  ([Objects](docs/configuration.md#objects)).
 - **Places and dates from anytopdf.** Turn on Settings › Analysis › Find places and dates with anytopdf (off by
   default) and analysis also asks anytopdf, offline, for the places a text names, the dates it gives (with their ISO
   value) and where a photo was taken (from its GPS fix, with `exiftool` installed). They become places and dates in the

diff --git a/docs/components.md b/docs/components.md
@@ -15,7 +15,7 @@ archive that only uses cloud models never downloads them, which keeps a small ma
 | Whisper model | `transcribe.engine` is `whisper` | the faster-whisper model named in `transcribe.whisper.model` |
 | Voice IDs | speakers are separated by clustering (the default) | SpeechBrain and its ECAPA voiceprint model |
 | Face models | `video.face_engine` is `opencv` (the default) | OpenCV, YuNet and SFace |
-| Object model | `video.object_engine` is `yolox` (the default) | ONNX Runtime and YOLOX-s |
+| Object model | `video.object_engine` is `yolox` (the default) or `anytopdf` | YOLOX-s, and ONNX Runtime for `yolox` |
 | Chat model, embedding model | the LLM provider or embeddings server is Ollama | pulled on that server (`/api/pull`) |
 | Outlook .msg emails | listed in `components.also` | extract-msg (GPL-3.0, so only when asked for) |
 

diff --git a/docs/configuration.md b/docs/configuration.md
@@ -303,13 +303,16 @@ video's sampled frames and on a document's or an image's pages ([API](api.md#obj
   a job needs them ([Components](components.md)), and also finds a model put in `/opt/lens/models`.
 * `ultralytics`: Ultralytics YOLO (`pip install ultralytics`). It's AGPL-3.0: a server that lets others use it must
   offer them its source, so it's in no image and no extra.
+* `anytopdf`: the same YOLOX model, run by [anytopdf](#anytopdf)'s objects plugin, sandboxed with no network. It
+  needs no ONNX Runtime in Lens, for a lean server or a Raspberry Pi; Lens fetches anytopdf and YOLOX-s the first
+  time a job needs them. anytopdf runs here even when documents go to a conversion node.
 * `off`.
 
 Without one the step is skipped, and its job says why. Settings → Video, OCR, faces and objects:
 
 | Setting | Default | |
 |---|---|---|
-| `video.object_engine` | `yolox` | `yolox`, `ultralytics` or `off` |
+| `video.object_engine` | `yolox` | `yolox`, `ultralytics`, `anytopdf` or `off` |
 | `video.object_min_score` | 0.4 | how sure the detector must be to keep what it found, 0.05–0.95 |
 
 Set at startup only:

diff --git a/fastapi_backend/app/domain/components.py b/fastapi_backend/app/domain/components.py
@@ -389,39 +389,44 @@ def fetch(self, cfg, m, say):
 
 class Objects(Component):
     def needed(self, cfg, m):
-        return cfg["video"].get("object_engine") == "yolox"
+        return cfg["video"].get("object_engine") in ("yolox", "anytopdf")
 
     def present(self, cfg):
         from . import objects
 
-        return importable("onnxruntime") and bool(objects.yolox_model(cfg))
+         == "anytopdf" or importable("onnxruntime")  # anytopdf runs the model itself
+        return onnx and bool(objects.yolox_model(cfg))
 
     def fetch(self, cfg, m, say):
-        if not importable("onnxruntime"):
+        if cfg["video"].get("object_engine") != "anytopdf" and not importable("onnxruntime"):
             pip_install(cfg, "objects", say)
         fetch_file(cfg, "yolox_s.onnx", say)
 
 
 class Anytopdf(Component):
     """The anytopdf program (anytopdf.py), fetched when Settings → Documents asks for it to run here, or Settings →
-    Analysis asks it for places and dates (enrich.py)."""
+    Analysis asks it for places and dates (enrich.py), or Settings → Video for objects (objects.py)."""
 
-    def needed(self, cfg, m):
+    def _serves(self, cfg):
         from . import anytopdf
 
-        here = anytopdf.mode(cfg) == "anytopdf" and not anytopdf.node(cfg)
-        return (here or bool((cfg.get("analysis") or {}).get("anytopdf"))) and bool(anytopdf.archive())
+        converts = anytopdf.mode(cfg) == "anytopdf" and not anytopdf.node(cfg)
+        detects = cfg["video"].get("object_engine") == "anytopdf"  # its objects plugin runs here, node or not
+        enriches = bool((cfg.get("analysis") or {}).get("anytopdf"))
+        return ({"transcribe"} if converts else set()) | ({"objects"} if detects else set()) | ({"analyze"} if enriches else set())
 
-    def present(self, cfg):
+    def needed(self, cfg, m):
         from . import anytopdf
 
-        return bool(anytopdf.binary(cfg))
+        return bool(self._serves(cfg)) and bool(anytopdf.archive())
 
     def serves(self, cfg, m):
+        return self._serves(cfg)
+
+    def present(self, cfg):
         from . import anytopdf
 
-        here = anytopdf.mode(cfg) == "anytopdf" and not anytopdf.node(cfg)
-        return ({"transcribe"} if here else set()) | ({"analyze"} if (cfg.get("analysis") or {}).get("anytopdf") else set())
+        return bool(anytopdf.binary(cfg))
 
     def fetch(self, cfg, m, say):
         from . import anytopdf
@@ -556,8 +561,8 @@ def detail(self, cfg):
     Anytopdf(
         "anytopdf",
         "anytopdf",
-        "makes documents and photographed pages into searchable PDFs (documents.converter), finds places and dates",
-        steps={"transcribe", "analyze"},
+        "makes documents and photographed pages into searchable PDFs (documents.converter), finds places and dates, and objects (video.object_engine)",
+        steps={"transcribe", "analyze", "objects"},
         size_mb=50,
         license="MIT OR Apache-2.0",
     ),

diff --git a/fastapi_backend/app/domain/objects.py b/fastapi_backend/app/domain/objects.py
@@ -6,6 +6,8 @@
   one in MODELS or the data folder's models (components.py fetches yolox_s there on first use); `pip install "lens[objects]"` brings ONNX Runtime.
 - ultralytics: Ultralytics YOLO (video.ultralytics_model, such as yolov8n.pt). It's AGPL-3.0: a server that lets
   others use it must offer them its source, so it's in no image or extra; `pip install ultralytics` to use it.
+- anytopdf: the same YOLOX model, run by anytopdf's objects plugin (anytopdf.analyze, sandboxed with no network), so a
+  lean server or a Raspberry Pi needs no ONNX Runtime in Python. components.py fetches anytopdf and the model.
 Without one the objects step is skipped, saying why.
 
 A resource's objects are rows of object_track, one per kind: the spans it's in (time, or pages for a document, as
@@ -167,11 +169,39 @@ def detect(self, path):
         return sorted(out, key=lambda o: -o["score"])
 
 
+class AnytopdfObjects:
+    """YOLOX run by anytopdf's objects plugin: the program here (documents.anytopdf, or the one fetched) and a YOLOX
+    .onnx model, without ONNX Runtime in Lens."""
+
+    name = "anytopdf"
+
+    def __init__(self, cfg):
+        from . import anytopdf
+
+        if not anytopdf.binary(cfg):
+            raise RuntimeError("anytopdf isn't installed here (it's fetched the first time a job needs it)")
+        model = yolox_model(cfg)
+        if not model or not pathlib.Path(model).is_file():
+            raise RuntimeError("no YOLOX model: set video.yolox_model to a YOLOX .onnx file (the lens:full image has one)")
+        self.cfg, self.models = cfg, {"ANYTOPDF_OBJECTS_MODEL": model}
+        self.min_score = float(cfg["video"].get("object_min_score") or 0.4)
+
+    def detect(self, path):
+        from . import anytopdf
+
+        out = []
+        for o in anytopdf.analyze(self.cfg, path, self.models):
+            if o["kind"] == "object" and o["score"] >= self.min_score:
+                label = str(o["label"]).replace("_", " ").casefold()
+                out.append({"label": label, "score": o["score"], "box": [round(float(v), 4) for v in o["box"]]})
+        return sorted(out, key=lambda o: -o["score"])
+
+
 def _fractions(x1, y1, x2, y2, W, H):
     return [round(x1 / W, 4), round(y1 / H, 4), round((x2 - x1) / W, 4), round((y2 - y1) / H, 4)]
 
 
-ENGINES = {"yolox": YoloxOnnx, "ultralytics": UltralyticsYolo}
+ENGINES = {"yolox": YoloxOnnx, "ultralytics": UltralyticsYolo, "anytopdf": AnytopdfObjects}
 
 
 def engine(cfg):

diff --git a/fastapi_backend/app/domain/settings.py b/fastapi_backend/app/domain/settings.py
@@ -129,7 +129,7 @@
     ("reports", "audio"): {"link", "embed", "none"},
     ("video", "ocr_engine"): {"auto", "tesseract", "apple-vision", "rapidocr", "doctr", "none"},
     ("video", "face_engine"): {"opencv", "insightface", "none"},
-    ("video", "object_engine"): {"yolox", "ultralytics", "off"},
+    ("video", "object_engine"): {"yolox", "ultralytics", "anytopdf", "off"},
     ("decisions", "engine"): {"auto", "jev", "laya", "llm", "off"},
     ("voice", "input"): {"auto", "server", "browser"},
     ("voice", "stt"): {"same", "sensevoice", "whisper", "mlx-whisper", "openai", "elevenlabs", "assemblyai", "deepgram"},

diff --git a/fastapi_backend/tests/api/test_objects.py b/fastapi_backend/tests/api/test_objects.py
@@ -8,7 +8,8 @@
 
 import pytest
 
-from app.domain import ingest, jobs, objects, store
+from app.domain import components, ingest, jobs, objects, store
+from tests import fake_anytopdf
 from tests.helpers import drain, login, make_user, quiet, text_pdf
 from tests.video_helpers import make_video, needs_ffmpeg
 
@@ -146,7 +147,37 @@ def test_object_settings_are_checked(client, people):
     assert (r.status_code, r.json()["detail"]) == (400, "video.object_min_score is a number from 0.05 to 0.95")
     r = client.put("/api/v1/settings/video", json={"object_engine": "magic"}, headers=ha)
     assert r.status_code == 400 and "ultralytics" in r.json()["detail"]
+    assert client.put("/api/v1/settings/video", json={"object_engine": "anytopdf"}, headers=ha).status_code == 200
     assert client.put("/api/v1/settings/video", json={"object_engine": "off", "object_min_score": 0.5}, headers=ha).status_code == 200
     assert client.put("/api/v1/settings/video", json={"yolox_model": "/tmp/x.onnx"}, headers=ha).status_code == 400  # startup only
     assert client.put("/api/v1/settings/video", json={"object_min_score": 0.5}, headers=people["he"]).status_code == 403
     assert "yolox_model" in client.get("/api/v1/settings", headers=ha).json()["bootstrap"]
+
+
+def test_anytopdf_finds_objects_with_lens_yolox_model(cfg, tmp_path, monkeypatch):
+    cfg["video"].update(object_engine="anytopdf", yolox_model=None)
+    monkeypatch.setattr(objects, "MODELS", tmp_path / "no-models")
+    cfg["documents"]["anytopdf"] = str(tmp_path / "gone")
+    assert objects.engine(cfg) == (None, "anytopdf isn't installed here (it's fetched the first time a job needs it)")
+    log = tmp_path / "runs.jsonl"
+    cfg["documents"]["anytopdf"] = fake_anytopdf.make(tmp_path / "anytopdf", log)
+    found, why = objects.engine(cfg)
+    assert found is None and why.startswith("no YOLOX model")
+
+    model = tmp_path / "yolox_s.onnx"
+    model.write_bytes(b"a model")
+    cfg["video"]["yolox_model"] = str(model)
+    found, why = objects.engine(cfg)
+    assert (found.name, why) == ("anytopdf", None)
+    pic = tmp_path / "bus.jpg"
+    pic.write_bytes(b"a picture")
+    assert found.detect(pic) == [{"label": "bus", "score": 0.93, "box": [0.02, 0.2, 0.9, 0.5]}]  # its face is left out
+    assert fake_anytopdf.runs(log)[-1]["anytopdf_env"] == ["ANYTOPDF_DATA_DIR", "ANYTOPDF_OBJECTS_MODEL"]
+    cfg["video"]["object_min_score"] = 0.95
+    assert objects.engine(cfg)[0].detect(pic) == []  # not sure enough
+
+    # what it needs is fetched on first use: anytopdf and the model, not ONNX Runtime
+    assert components.BY_ID["anytopdf"].serves(cfg, None) == {"objects"}
+    assert components.BY_ID["objects"].needed(cfg, None) and components.BY_ID["objects"].present(cfg)
+    cfg["video"]["object_engine"] = "yolox"
+    assert components.BY_ID["anytopdf"].serves(cfg, None) == set()
diff --git a/nextjs-frontend/components/settings/model.ts b/nextjs-frontend/components/settings/model.ts
@@ -1496,6 +1496,7 @@ export const FIELDS: FieldSpec[] = [
     options: [
       { value: "yolox", label: "YOLOX on ONNX Runtime (Apache-2.0)" },
       { value: "ultralytics", label: "Ultralytics YOLO (AGPL-3.0)" },
5532      { value: "anytopdf", label: "YOLOX run by anytopdf (no ONNX Runtime)" },
       { value: "off", label: "Off" },
     ],
     hint: "Finds people, vehicles, animals and everyday things on frames and pages",