Skip to content

Personal AI, On Personal Devices

Python

fix(cli): don't let a broken numpy crash CLI/server startup on Windows (#433)

Jon Saad-Falconad7c86495ff5
diff --git a/src/openjarvis/cli/__init__.py b/src/openjarvis/cli/__init__.py
index f42773a4..b1553595 100644
--- a/src/openjarvis/cli/__init__.py
+++ b/src/openjarvis/cli/__init__.py
@@ -17,7 +17,6 @@ from openjarvis.cli.compose_cmd import compose
from openjarvis.cli.config_cmd import config
from openjarvis.cli.connect_cmd import connect
from openjarvis.cli.daemon_cmd import restart, start, status, stop
-from openjarvis.cli.deep_research_setup_cmd import deep_research_setup
from openjarvis.cli.digest_cmd import digest
from openjarvis.cli.doctor_cmd import doctor
from openjarvis.cli.eval_cmd import eval_group
@@ -117,8 +116,22 @@ cli.add_command(config, "config")
cli.add_command(scan, "scan")
cli.add_command(connect, "connect")
cli.add_command(digest, "digest")
-cli.add_command(deep_research_setup, "deep-research-setup")
-cli.add_command(deep_research_setup, "research")
+# deep-research setup pulls the ingestion pipeline (embeddings/numpy). Guard it
+# so a broken or slow numpy on Windows — which can raise at IMPORT time, not
+# just ImportError (#404) — can never take down the whole CLI, including
+# `jarvis serve`. Invoking `jarvis deep-research-setup` without the deps still
+# errors clearly on demand.
+try:
+ from openjarvis.cli.deep_research_setup_cmd import deep_research_setup
+
+ cli.add_command(deep_research_setup, "deep-research-setup")
+ cli.add_command(deep_research_setup, "research")
+except Exception as _dr_exc:
+ import logging as _logging
+
+ _logging.getLogger(__name__).debug(
+ "deep-research command unavailable: %s", _dr_exc
+ )
cli.add_command(self_update, "self-update")
cli.add_command(bootstrap_cmd, "_bootstrap")
diff --git a/src/openjarvis/connectors/embeddings.py b/src/openjarvis/connectors/embeddings.py
index 45b35119..d7e51be8 100644
--- a/src/openjarvis/connectors/embeddings.py
+++ b/src/openjarvis/connectors/embeddings.py
@@ -13,11 +13,17 @@ so ingestion never fails because a sidecar service is down.
from __future__ import annotations
import logging
-from typing import List, Optional
+from typing import TYPE_CHECKING, List, Optional
-import numpy as np
import requests
+# numpy is imported lazily inside the functions that use it (not at module
+# load). The CLI imports this module eagerly via the deep-research command
+# chain, so a module-level `import numpy` makes a broken/slow numpy on Windows
+# crash every `jarvis` command — including `jarvis serve` (#404, #309).
+if TYPE_CHECKING:
+ import numpy as np
+
logger = logging.getLogger(__name__)
@@ -110,6 +116,8 @@ class OllamaEmbedder:
)
return None
+ import numpy as np
+
arr = np.asarray(vec, dtype=np.float32)
if self._dim is None:
self._dim = int(arr.shape[0])
@@ -136,16 +144,20 @@ class OllamaEmbedder:
def decode_embedding(
- blob: Optional[bytes], *, dtype: type = np.float32
+ blob: Optional[bytes], *, dtype=None
) -> Optional[np.ndarray]:
"""Reconstruct a 1-D vector from a BLOB written by ``OllamaEmbedder.embed``.
Returns ``None`` when the input is missing or zero-length so callers can
- treat absent embeddings uniformly.
+ treat absent embeddings uniformly. ``dtype`` defaults to ``np.float32``
+ (resolved lazily; passing ``np.float32`` as a default arg would import
+ numpy at module load).
"""
if not blob:
return None
- return np.frombuffer(blob, dtype=dtype)
+ import numpy as np
+
+ return np.frombuffer(blob, dtype=dtype if dtype is not None else np.float32)
__all__ = [
diff --git a/src/openjarvis/connectors/hybrid_search.py b/src/openjarvis/connectors/hybrid_search.py
index 2165db7e..b6e6210a 100644
--- a/src/openjarvis/connectors/hybrid_search.py
+++ b/src/openjarvis/connectors/hybrid_search.py
@@ -24,8 +24,8 @@ from dataclasses import dataclass, field
from datetime import datetime
from typing import Any, Dict, List, Optional, Sequence, Tuple
-import numpy as np
-
+# numpy imported lazily inside _vector_recall (see embeddings.py) so importing
+# this module never forces numpy at load time (#404, #309).
from openjarvis.connectors.embeddings import OllamaEmbedder, decode_embedding
from openjarvis.connectors.store import KnowledgeStore
@@ -254,6 +254,8 @@ class HybridSearch:
"""Return ``[(chunk_id, cosine_score), ...]`` from a brute-force scan."""
if self._embedder is None:
return []
+ import numpy as np
+
q_blob = self._embedder.embed(query)
q_vec = decode_embedding(q_blob)
if q_vec is None or q_vec.size == 0:
diff --git a/tests/cli/test_cli.py b/tests/cli/test_cli.py
index 63668b36..84e2daac 100644
--- a/tests/cli/test_cli.py
+++ b/tests/cli/test_cli.py
@@ -3,6 +3,7 @@
from __future__ import annotations
import io
+import subprocess
import sys
from pathlib import Path
from unittest import mock
@@ -136,3 +137,29 @@ class TestCLI:
assert config_path.exists()
content = config_path.read_text()
assert "[engine]" in content
+
+
+class TestStartupResilience:
+ """Importing the CLI must not force heavy/native deps (#404, #309).
+
+ A broken or slow numpy on Windows otherwise raises at import time and takes
+ down every `jarvis` command — including `jarvis serve` — because the CLI
+ eagerly pulls the deep-research command chain (-> embeddings -> numpy).
+ """
+
+ def test_importing_cli_does_not_import_numpy(self) -> None:
+ # Run in a fresh subprocess: the pytest session itself almost certainly
+ # has numpy loaded from other tests, so an in-process check is useless.
+ code = (
+ "import openjarvis.cli, sys; "
+ "leaked=[m for m in sys.modules if m=='numpy' or m.startswith('numpy.')]; "
+ "assert not leaked, leaked; "
+ "print('numpy-free')"
+ )
+ result = subprocess.run(
+ [sys.executable, "-c", code], capture_output=True, text=True
+ )
+ assert result.returncode == 0, (
+ "importing openjarvis.cli pulled in numpy (a broken numpy would then "
+ f"crash `jarvis serve`):\nstdout={result.stdout}\nstderr={result.stderr}"
+ )