codejury 0.16.0__tar.gz → 0.16.1__tar.gz

{codejury-0.16.0 → codejury-0.16.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codejury
-Version: 0.16.0
+Version: 0.16.1
 Summary: AI code security review: an adversarial diff-audit engine and an agent-driven whole-repo review methodology, with security knowledge as rich rules
 Author: AISecLabs
 License-Expression: MIT

{codejury-0.16.0 → codejury-0.16.1}/codejury/diff/engine.py

@@ -12,6 +12,7 @@ from codejury.diff.prompts import SYSTEM, standard_audit_prompt
 from codejury.diff.rules import rules_for_diff
 from codejury.json_parse import extract_json_object
 from codejury.providers.base import Message, Provider
+from codejury.guides import guides_for_diff
 
 
 class AuditError(RuntimeError):
@@ -33,9 +34,10 @@ class AuditRunner:
     def run(self, diff: str, *, rules: str = "", context: str = "") -> list[Finding]:
         if not rules:
             rules = rules_for_diff(diff)  # inject the rules relevant to this diff
+        stack = guides_for_diff(diff)     # inject the target's language/framework conventions
         result = self._provider.complete(
             system=SYSTEM,
-            messages=[Message(role="user", content=standard_audit_prompt(diff, rules=rules, context=context))],
+            messages=[Message(role="user", content=standard_audit_prompt(diff, rules=rules, context=context, stack=stack))],
             model=self._model,
             max_tokens=self._max_tokens,
         )

{codejury-0.16.0 → codejury-0.16.1}/codejury/diff/prompts.py

@@ -57,7 +57,8 @@ def category_block() -> str:
     )
 
 
-def standard_audit_prompt(diff: str, *, rules: str = "", context: str = "") -> str:
+def standard_audit_prompt(diff: str, *, rules: str = "", context: str = "", stack: str = "") -> str:
+    stack_block = f"Conventions of the target's language/framework:\n{stack}\n\n" if stack else ""
     rules_block = f"Relevant security rules for reference:\n{rules}\n\n" if rules else ""
     context_block = (
         f"Surrounding code for tracing where values come from (not under review):\n"
@@ -69,6 +70,7 @@ def standard_audit_prompt(diff: str, *, rules: str = "", context: str = "") -> s
         "Review the following code change for security vulnerabilities.\n\n"
         f"{FOCUS}\n{DO_NOT_REPORT}\n"
         f"{category_block()}"
+        f"{stack_block}"
         f"{rules_block}"
         f"Code change (unified diff):\n```diff\n{diff}\n```\n\n"
         f"{context_block}"

{codejury-0.16.0/codejury/repo → codejury-0.16.1/codejury}/guides.py

@@ -13,6 +13,7 @@ code change, which keeps the unbounded language/framework axis out of code.
 from __future__ import annotations
 
 import fnmatch
+import re
 from dataclasses import dataclass
 from pathlib import Path
 
@@ -20,6 +21,8 @@ import yaml
 
 from codejury.resources import FRAMEWORKS_DIR, LANGUAGES_DIR
 
+_DIFF_PATH = re.compile(r"^(?:\+\+\+ b/|diff --git a/\S+ b/)(\S+)", re.MULTILINE)
+
 
 @dataclass(frozen=True, kw_only=True)
 class Guide:
@@ -62,18 +65,33 @@ def load_guides(languages_dir=LANGUAGES_DIR, frameworks_dir=FRAMEWORKS_DIR) -> l
     return out
 
 
-def _matches(guide: Guide, files: list[str], manifest: str) -> bool:
+def _matches(guide: Guide, files: list[str], text: str) -> bool:
     if any(fnmatch.fnmatch(f, pat) for pat in guide.detect_files for f in files):
         return True
-    if any(m in manifest for m in guide.detect_manifest):
+    if any(m in text for m in guide.detect_manifest):
+        return True
+    if any(i in text for i in guide.detect_imports):
         return True
     return False
 
 
-def select_guides(files, *, manifest_text: str = "", guides: list[Guide] | None = None) -> list[Guide]:
-    """The guides whose detect signals fire on the repo (its file paths and the
-    concatenated text of its dependency manifests), languages first then frameworks."""
+def select_guides(files, *, text: str = "", guides: list[Guide] | None = None) -> list[Guide]:
+    """The guides whose detect signals fire on the target, languages first then
+    frameworks. `files` are the target's file paths; `text` is content to scan for
+    manifest substrings and import markers (a repo's manifests, or a diff body)."""
     pool = load_guides() if guides is None else guides
     file_list = list(files)
-    manifest = manifest_text.lower()
-    return [g for g in pool if _matches(g, file_list, manifest)]
+    blob = text.lower()
+    return [g for g in pool if _matches(g, file_list, blob)]
+
+
+def _changed_paths(diff: str) -> list[str]:
+    """The file paths touched by a unified diff (from its `+++ b/` / `diff --git` headers)."""
+    return _DIFF_PATH.findall(diff)
+
+
+def guides_for_diff(diff: str, *, guides: list[Guide] | None = None) -> str:
+    """Concatenated bodies of the language/framework guides relevant to a diff, by
+    its changed paths and its content. Empty when nothing matches."""
+    selected = select_guides(_changed_paths(diff), text=diff, guides=guides)
+    return "\n\n---\n\n".join(g.body for g in selected)

{codejury-0.16.0 → codejury-0.16.1}/codejury/repo/scaffold.py

@@ -13,7 +13,7 @@ from __future__ import annotations
 from dataclasses import dataclass, field
 from pathlib import Path
 
-from codejury.repo.guides import Guide, select_guides
+from codejury.guides import Guide, select_guides
 from codejury.repo.model import build_repo_model_from_dir
 from codejury.resources import AGENT_DIR
 
@@ -105,7 +105,7 @@ def scaffold(target: str | Path, workspace: str | Path) -> ScaffoldResult:
     (ws / "entrypoints" / "_entrypoints.md").write_text(_entrypoints_md(model), encoding="utf-8")
 
     # detect the stack and seed its review guides (languages + frameworks)
-    guides = select_guides(model.files, manifest_text=_read_manifests(target))
+    guides = select_guides(model.files, text=_read_manifests(target))
     (ws / "_stack.md").write_text(_stack_md(guides), encoding="utf-8")
 
     return ScaffoldResult(

{codejury-0.16.0 → codejury-0.16.1}/codejury.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codejury
-Version: 0.16.0
+Version: 0.16.1
 Summary: AI code security review: an adversarial diff-audit engine and an agent-driven whole-repo review methodology, with security knowledge as rich rules
 Author: AISecLabs
 License-Expression: MIT

{codejury-0.16.0 → codejury-0.16.1}/codejury.egg-info/SOURCES.txt

@@ -3,6 +3,7 @@ README.md
 pyproject.toml
 codejury/__init__.py
 codejury/cli.py
+codejury/guides.py
 codejury/json_parse.py
 codejury/report.py
 codejury/resources.py
@@ -63,7 +64,6 @@ codejury/providers/openai.py
 codejury/providers/openai_format.py
 codejury/providers/retry.py
 codejury/repo/__init__.py
-codejury/repo/guides.py
 codejury/repo/model.py
 codejury/repo/scaffold.py
 tests/test_anthropic_provider.py

{codejury-0.16.0 → codejury-0.16.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "codejury"
-version = "0.16.0"
+version = "0.16.1"
 description = "AI code security review: an adversarial diff-audit engine and an agent-driven whole-repo review methodology, with security knowledge as rich rules"
 readme = "README.md"
 requires-python = ">=3.12"

{codejury-0.16.0 → codejury-0.16.1}/tests/test_diff_engine.py

@@ -85,11 +85,12 @@ def test_engine_raises_on_wrong_shape_json():
 
 
 def test_prompt_carries_diff_focus_and_do_not_report():
-    p = standard_audit_prompt(_DIFF, rules="RULE-X", context="def caller(): ...")
+    p = standard_audit_prompt(_DIFF, rules="RULE-X", context="def caller(): ...", stack="STACK-NOTE")
     assert "SELECT * FROM u" in p          # the diff
     assert "Do NOT report" in p            # the noise-control list
     assert "IDOR" in p                     # the focus
     assert "RULE-X" in p                   # rules excerpt
+    assert "STACK-NOTE" in p               # language/framework conventions block
     assert "def caller()" in p             # context block
 
 

{codejury-0.16.0 → codejury-0.16.1}/tests/test_guides.py

@@ -1,7 +1,7 @@
 """Language/framework review guides load and are selected by detection signals
 (file globs + dependency-manifest substrings), so adding one is a drop-in file."""
 
-from codejury.repo.guides import Guide, load_guides, select_guides
+from codejury.guides import Guide, guides_for_diff, load_guides, select_guides
 
 
 def test_shipped_guides_load():
@@ -19,7 +19,7 @@ def test_select_by_file_glob():
 
 
 def test_select_by_manifest_substring():
-    matched = {g.id for g in select_guides(["main.py"], manifest_text="Django==4.2\nrequests\n")}
+    matched = {g.id for g in select_guides(["main.py"], text="Django==4.2\nrequests\n")}
     assert "django" in matched and "python" in matched
 
 
@@ -32,3 +32,14 @@ def test_select_respects_injected_pool():
                   detect_manifest=(), detect_imports=(), body="b")]
     assert [g.id for g in select_guides(["a.xyz"], guides=only)] == ["x"]
     assert select_guides(["a.py"], guides=only) == []
+
+
+def test_guides_for_diff_by_path_and_content():
+    diff = ("diff --git a/app/urls.py b/app/urls.py\n"
+            "+from django.urls import path\n+urlpatterns = []\n")
+    notes = guides_for_diff(diff)
+    assert "Django" in notes and "Python" in notes   # urls.py + .py + the django import
+
+
+def test_guides_for_diff_empty_when_irrelevant():
+    assert guides_for_diff("+++ b/README.md\n+hello\n") == ""