~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.
Python academiaanthropicarxivbravedeep-researchencryptionhome-automationhomeserverlocallocal-deep-researchlocal-llmmistralollamaopenaipubmedresearchresearch-toolretrieval-augmented-generationsearxngself-hosted
"""Direct unit tests for ``validated_uuid_path_param`` (web/utils/path_ids.py).
Mirrors ``tests/security/test_collection_id_xss.py:50-60``, the direct
unit test for ``_validated_collection_id`` -- the sibling helper this
one's docstring is modelled on. ``test_page_route_uuid_fences.py``
already exercises this function indirectly through the two page routes
over HTTP; these tests drive it directly, including the property that
distinguishes it from a bare shape check: the return value is the
*canonical* form of the UUID, not an echo of whatever alias spelling
was submitted (path_ids.py: ``canonical = str(uuid.UUID(str(value)))``).
"""
import uuid
import pytest
from fastapi import HTTPException
from local_deep_research.web.utils.path_ids import validated_uuid_path_param
_CANON = "00000000-0000-0000-0000-000000000000"
_HEX32 = "0" * 32
# Malformed / hostile shapes: none of these is a well-formed UUID under
# any spelling, so every one must 404 -- whatever the payload.
REJECTED_VALUES = [
"<script>alert(1)</script>",
"1';alert(document.cookie);var x='",
'1";alert(1);var y="',
"../../etc/passwd",
"1 OR 1=1",
"",
"not-a-uuid",
7,
None,
# ``uuid.UUID`` strips only braces and hyphens before its 32-character
# length guard, so padding a full 32-digit hex string leaves 33
# characters and is rejected -- ``int(hex, 16)`` would tolerate the
# whitespace, but it never runs. (Whitespace plus 31 hex digits does
# parse, as a *different* value, so it is not an alias of this one.)
(" " + _HEX32),
(_HEX32 + "\n"),
]
@pytest.mark.parametrize("bad", REJECTED_VALUES, ids=lambda v: repr(v)[:40])
def test_malformed_value_is_rejected(bad):
"""Every non-UUID shape 404s, carrying the fence's own status code."""
with pytest.raises(HTTPException) as exc_info:
validated_uuid_path_param(bad, "Note")
assert exc_info.value.status_code == 404
# Alias spellings of the SAME UUID value: must be accepted, and the
# return value must be that UUID's canonical form -- not the alias
# verbatim. (Pre-tidy, ``validated_uuid_path_param`` returned
# ``str(value)``, so every one of these reached the template exactly
# as submitted; see gate report R5.)
ALIAS_VALUES = [
("urn_uuid_prefix", "urn:uuid:" + _CANON),
("braces", "{" + _CANON + "}"),
("hyphen_free_hex", _HEX32),
("sixteen_kb_hyphens_plus_hex", "-" * 16384 + _HEX32),
("unicode_digits_arabic_indic", "٠" * 32),
("uppercase_hex", "A" * 32),
]
@pytest.mark.parametrize(
"value", [v for _, v in ALIAS_VALUES], ids=[n for n, _ in ALIAS_VALUES]
)
def test_alias_shape_is_accepted_and_canonicalised(value):
result = validated_uuid_path_param(value, "Note")
# Same UUID value as the alias input...
assert uuid.UUID(result) == uuid.UUID(str(value))
# ...returned in canonical form: idempotent under re-parsing, so it
# is not just "a" valid spelling but *the* canonical one.
assert result == str(uuid.UUID(result))
def test_real_uuid4_round_trips_unchanged():
"""Stored ids are already ``str(uuid.uuid4())``: canonicalising a
real id must be a no-op, or every existing note/document link
breaks.
"""
real = str(uuid.uuid4())
assert validated_uuid_path_param(real, "Note") == real