aicaudit 0.2.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (84) hide show
  1. aicaudit-0.2.0/LICENSE +21 -0
  2. aicaudit-0.2.0/PKG-INFO +461 -0
  3. aicaudit-0.2.0/README.md +430 -0
  4. aicaudit-0.2.0/aicaudit/__init__.py +1 -0
  5. aicaudit-0.2.0/aicaudit/__main__.py +3 -0
  6. aicaudit-0.2.0/aicaudit/audit.py +167 -0
  7. aicaudit-0.2.0/aicaudit/cli.py +256 -0
  8. aicaudit-0.2.0/aicaudit/config.py +120 -0
  9. aicaudit-0.2.0/aicaudit/fix.py +213 -0
  10. aicaudit-0.2.0/aicaudit/graph.py +168 -0
  11. aicaudit-0.2.0/aicaudit/llm/__init__.py +1 -0
  12. aicaudit-0.2.0/aicaudit/llm/client.py +249 -0
  13. aicaudit-0.2.0/aicaudit/llm/prompts.py +195 -0
  14. aicaudit-0.2.0/aicaudit/output/__init__.py +1 -0
  15. aicaudit-0.2.0/aicaudit/output/json_output.py +34 -0
  16. aicaudit-0.2.0/aicaudit/output/markdown_output.py +52 -0
  17. aicaudit-0.2.0/aicaudit/output/sarif_output.py +121 -0
  18. aicaudit-0.2.0/aicaudit/rules/__init__.py +1 -0
  19. aicaudit-0.2.0/aicaudit/rules/base.py +90 -0
  20. aicaudit-0.2.0/aicaudit/rules/performance/__init__.py +1 -0
  21. aicaudit-0.2.0/aicaudit/rules/performance/complexity.py +57 -0
  22. aicaudit-0.2.0/aicaudit/rules/performance/nesting_depth.py +49 -0
  23. aicaudit-0.2.0/aicaudit/rules/quality/__init__.py +1 -0
  24. aicaudit-0.2.0/aicaudit/rules/quality/bare_except.py +30 -0
  25. aicaudit-0.2.0/aicaudit/rules/quality/magic_numbers.py +85 -0
  26. aicaudit-0.2.0/aicaudit/rules/quality/todo_comment.py +29 -0
  27. aicaudit-0.2.0/aicaudit/rules/quality/undefined_name.py +80 -0
  28. aicaudit-0.2.0/aicaudit/rules/quality/unused_variable.py +55 -0
  29. aicaudit-0.2.0/aicaudit/rules/security/__init__.py +1 -0
  30. aicaudit-0.2.0/aicaudit/rules/security/dangerous_functions.py +89 -0
  31. aicaudit-0.2.0/aicaudit/rules/security/insecure_random.py +58 -0
  32. aicaudit-0.2.0/aicaudit/rules/security/path_traversal.py +107 -0
  33. aicaudit-0.2.0/aicaudit/rules/security/secret_leak.py +71 -0
  34. aicaudit-0.2.0/aicaudit/rules/security/sql_injection.py +114 -0
  35. aicaudit-0.2.0/aicaudit/rules/security/ssrf.py +66 -0
  36. aicaudit-0.2.0/aicaudit/rules/security/weak_crypto.py +57 -0
  37. aicaudit-0.2.0/aicaudit/rules/security/xml_xxe.py +132 -0
  38. aicaudit-0.2.0/aicaudit/scan.py +266 -0
  39. aicaudit-0.2.0/aicaudit/web/__init__.py +5 -0
  40. aicaudit-0.2.0/aicaudit/web/server.py +495 -0
  41. aicaudit-0.2.0/aicaudit/web/static/app.css +416 -0
  42. aicaudit-0.2.0/aicaudit/web/static/app.js +440 -0
  43. aicaudit-0.2.0/aicaudit/web/templates/base.html +52 -0
  44. aicaudit-0.2.0/aicaudit/web/templates/config.html +48 -0
  45. aicaudit-0.2.0/aicaudit/web/templates/index.html +33 -0
  46. aicaudit-0.2.0/aicaudit/web/templates/results.html +34 -0
  47. aicaudit-0.2.0/aicaudit/web/templates/rules.html +29 -0
  48. aicaudit-0.2.0/aicaudit/web/templates/scan.html +48 -0
  49. aicaudit-0.2.0/aicaudit.egg-info/PKG-INFO +461 -0
  50. aicaudit-0.2.0/aicaudit.egg-info/SOURCES.txt +82 -0
  51. aicaudit-0.2.0/aicaudit.egg-info/dependency_links.txt +1 -0
  52. aicaudit-0.2.0/aicaudit.egg-info/entry_points.txt +2 -0
  53. aicaudit-0.2.0/aicaudit.egg-info/requires.txt +9 -0
  54. aicaudit-0.2.0/aicaudit.egg-info/top_level.txt +1 -0
  55. aicaudit-0.2.0/pyproject.toml +59 -0
  56. aicaudit-0.2.0/setup.cfg +4 -0
  57. aicaudit-0.2.0/tests/test_ai.py +89 -0
  58. aicaudit-0.2.0/tests/test_ai_harness.py +137 -0
  59. aicaudit-0.2.0/tests/test_ai_integration.py +116 -0
  60. aicaudit-0.2.0/tests/test_audit.py +50 -0
  61. aicaudit-0.2.0/tests/test_audit_ext.py +102 -0
  62. aicaudit-0.2.0/tests/test_collab.py +117 -0
  63. aicaudit-0.2.0/tests/test_config.py +74 -0
  64. aicaudit-0.2.0/tests/test_cov_edge.py +116 -0
  65. aicaudit-0.2.0/tests/test_coverage.py +167 -0
  66. aicaudit-0.2.0/tests/test_cwe_rules.py +142 -0
  67. aicaudit-0.2.0/tests/test_dx.py +173 -0
  68. aicaudit-0.2.0/tests/test_extended_coverage.py +243 -0
  69. aicaudit-0.2.0/tests/test_fix.py +112 -0
  70. aicaudit-0.2.0/tests/test_fix_cov.py +64 -0
  71. aicaudit-0.2.0/tests/test_graph.py +82 -0
  72. aicaudit-0.2.0/tests/test_insecure_random.py +63 -0
  73. aicaudit-0.2.0/tests/test_integration.py +87 -0
  74. aicaudit-0.2.0/tests/test_integration_edge.py +87 -0
  75. aicaudit-0.2.0/tests/test_llm_mock.py +117 -0
  76. aicaudit-0.2.0/tests/test_performance_rules.py +65 -0
  77. aicaudit-0.2.0/tests/test_quality_rules.py +123 -0
  78. aicaudit-0.2.0/tests/test_relay.py +59 -0
  79. aicaudit-0.2.0/tests/test_sarif_output.py +72 -0
  80. aicaudit-0.2.0/tests/test_security_rules.py +131 -0
  81. aicaudit-0.2.0/tests/test_suppression.py +72 -0
  82. aicaudit-0.2.0/tests/test_taint_engine.py +179 -0
  83. aicaudit-0.2.0/tests/test_taint_rules.py +114 -0
  84. aicaudit-0.2.0/tests/test_web.py +235 -0
aicaudit-0.2.0/LICENSE ADDED
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 IhateStomachProblems
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -0,0 +1,461 @@
1
+ Metadata-Version: 2.4
2
+ Name: aicaudit
3
+ Version: 0.2.0
4
+ Summary: AI-powered code audit CLI for Python projects. Static analysis + AI verification with evidence chains.
5
+ Author: IhateStomachProblems
6
+ License: MIT
7
+ Project-URL: Homepage, https://github.com/IhateStomachProblems/aicaudit
8
+ Project-URL: Repository, https://github.com/IhateStomachProblems/aicaudit
9
+ Project-URL: Documentation, https://github.com/IhateStomachProblems/aicaudit
10
+ Keywords: code-audit,security,static-analysis,code-quality,ai,vulnerability-scanning
11
+ Classifier: Development Status :: 3 - Alpha
12
+ Classifier: Intended Audience :: Developers
13
+ Classifier: License :: OSI Approved :: MIT License
14
+ Classifier: Programming Language :: Python :: 3.10
15
+ Classifier: Programming Language :: Python :: 3.11
16
+ Classifier: Programming Language :: Python :: 3.12
17
+ Classifier: Programming Language :: Python :: 3.13
18
+ Classifier: Topic :: Security
19
+ Classifier: Topic :: Software Development :: Quality Assurance
20
+ Requires-Python: >=3.10
21
+ Description-Content-Type: text/markdown
22
+ License-File: LICENSE
23
+ Requires-Dist: click>=8.0
24
+ Requires-Dist: tomli>=2.0; python_version < "3.11"
25
+ Requires-Dist: anthropic>=0.40.0
26
+ Requires-Dist: fastapi>=0.115.0
27
+ Requires-Dist: uvicorn>=0.30.0
28
+ Requires-Dist: jinja2>=3.0
29
+ Requires-Dist: pygments>=2.16
30
+ Dynamic: license-file
31
+
32
+ # AICAudit 🛡️
33
+
34
+ > Evidence-driven AI code audit for Python — every verdict ships with a machine-checkable taint path.
35
+ >
36
+ > 证据链驱动的 Python 代码审计:每个判定都附带可复核的污点传播路径(source → … → sink)
37
+
38
+ <p align="center">
39
+ <img src="https://img.shields.io/badge/Python-3.10+-3776AB?style=flat&logo=python&logoColor=white" alt="Python"/>
40
+ <img src="https://img.shields.io/badge/tests-305%20passed-brightgreen" alt="Tests"/>
41
+ <img src="https://img.shields.io/badge/coverage-92%25-brightgreen" alt="Coverage"/>
42
+ <img src="https://img.shields.io/badge/license-MIT-blue" alt="License"/>
43
+ <img src="https://img.shields.io/badge/rules-15-brightgreen" alt="Rules"/>
44
+ <img src="https://img.shields.io/badge/SARIF-2.1-blue" alt="SARIF"/>
45
+ <img src="https://img.shields.io/github/stars/IhateStomachProblems/aicaudit?style=social" alt="Stars"/>
46
+ </p>
47
+
48
+ ---
49
+
50
+ ## Why AICAudit
51
+
52
+ Other tools give you a verdict. AICAudit gives you the **evidence**:
53
+
54
+ ```text
55
+ S001 app.py:16 SQL injection risk
56
+ taint path: app.py:14 request.args (Flask user input)
57
+ -> app.py:14 assigned to 'uid'
58
+ -> app.py:15 assigned to 'query'
59
+ -> app.py:16 conn.execute()
60
+ ```
61
+
62
+ - **Taint engine** (pure Python, zero plugins): tracks user input from source
63
+ (Flask/Django/FastAPI request, `input()`, `os.environ`, `sys.argv`) through
64
+ assignments, f-strings, concatenation, and **across function boundaries** to
65
+ the sink — and proves constants safe, killing the classic false-positive
66
+ classes (`open(BASE_DIR/"x")`, constant SQL variables, `eval("1+1")`)
67
+ - **AI verdicts with receipts**: the LLM sees the full function and the taint
68
+ path, not a one-line snippet — hallucination-resistant by construction
69
+ - **SARIF codeFlows**: taint paths render natively in the GitHub Security tab
70
+ - **Zero config**: `pip install` and scan; works offline, local LLMs supported
71
+
72
+ ---
73
+
74
+ ## Quick Start
75
+
76
+ ```bash
77
+ # From source (PyPI package arrives with v0.2.0)
78
+ pip install git+https://github.com/IhateStomachProblems/aicaudit.git
79
+ ```
80
+
81
+ ```bash
82
+ # Scan a file or directory
83
+ aicaudit scan ./src
84
+
85
+ # Markdown report (default)
86
+ aicaudit scan ./src --output markdown
87
+
88
+ # JSON output (for CI / scripts / AI agent integration)
89
+ aicaudit scan ./src --output json
90
+
91
+ # SARIF 2.1 output (GitHub Code Scanning compatible)
92
+ aicaudit scan ./src --output sarif
93
+
94
+ # Chinese language
95
+ aicaudit scan ./src --lang zh
96
+
97
+ # CI gate: exit 1 when any finding is error or worse (0 clean / 2 usage error)
98
+ aicaudit scan ./src --fail-on error
99
+ ```
100
+
101
+ ---
102
+
103
+ ## CI & Git Integration
104
+
105
+ **One-line GitHub Action** (this repo ships a composite action at its root):
106
+
107
+ ```yaml
108
+ - uses: IhateStomachProblems/aicaudit@main
109
+ with:
110
+ path: .
111
+ fail-on: error # optional: fail the job at this severity
112
+ sarif: "true" # optional: upload SARIF to Code Scanning
113
+ ```
114
+
115
+ **pre-commit** (local hook until the PyPI release):
116
+
117
+ ```yaml
118
+ repos:
119
+ - repo: local
120
+ hooks:
121
+ - id: aicaudit
122
+ name: aicaudit
123
+ entry: aicaudit scan --fail-on error
124
+ language: system
125
+ types: [python]
126
+ ```
127
+
128
+ **Project config** — generate it interactively:
129
+
130
+ ```bash
131
+ aicaudit init # writes [tool.aicaudit] into pyproject.toml
132
+ ```
133
+
134
+ ---
135
+
136
+ ## Custom Rules (Python files, no DSL)
137
+
138
+ Drop rule files into `.aicaudit/rules/` (auto-discovered) or list extra dirs
139
+ under `[tool.aicaudit] rule-dirs`. Rules are ordinary Python using the public
140
+ API — a later registration with the same id overrides the builtin:
141
+
142
+ ```python
143
+ # .aicaudit/rules/no_print.py
144
+ import ast
145
+ from aicaudit.rules.base import Finding, Rule, Severity, register
146
+
147
+ @register
148
+ class NoPrint(Rule):
149
+ id = "Q100"
150
+ name = "no-print"
151
+ severity = Severity.INFO
152
+ description = "Detect print() calls"
153
+
154
+ def check(self, tree, context):
155
+ return [Finding(rule_id=self.id, message="print() — prefer logging",
156
+ message_zh="print()——建议改用 logging",
157
+ file=str(context.file_path), line=n.lineno,
158
+ severity=self.severity)
159
+ for n in ast.walk(tree)
160
+ if isinstance(n, ast.Call) and isinstance(n.func, ast.Name)
161
+ and n.func.id == "print"]
162
+ ```
163
+
164
+ A ready-to-copy example lives at
165
+ [examples/custom_rule_example.py](examples/custom_rule_example.py).
166
+ (A declarative YAML rule format is planned for v0.3 — the loader is ready.)
167
+
168
+ ---
169
+
170
+ ## Rules
171
+
172
+ ### Security
173
+
174
+ | ID | Rule | Severity |
175
+ |----|------|----------|
176
+ | S001 | SQL injection detection | CRITICAL |
177
+ | S002 | Hardcoded secret detection | CRITICAL |
178
+ | S003 | Dangerous functions (eval, exec, pickle, os.system) | ERROR |
179
+ | S004 | Path traversal detection | ERROR |
180
+ | S005 | SSRF detection | ERROR |
181
+ | S006 | Weak cryptography detection | WARNING |
182
+ | S007 | XML External Entity (XXE) detection | ERROR |
183
+ | S008 | Insecure random (non-crypto PRNG) | WARNING |
184
+
185
+ ### Quality
186
+
187
+ | ID | Rule | Severity |
188
+ |----|------|----------|
189
+ | Q001 | Bare except detection | WARNING |
190
+ | Q002 | Magic number detection | INFO |
191
+ | Q003 | Undefined name detection | ERROR |
192
+ | Q004 | TODO/FIXME comment detection | INFO |
193
+ | Q005 | Unused variable detection | WARNING |
194
+
195
+ ### Performance
196
+
197
+ | ID | Rule | Severity |
198
+ |----|------|----------|
199
+ | P001 | Cyclomatic complexity | WARNING |
200
+ | P002 | Nesting depth | WARNING |
201
+
202
+ ---
203
+
204
+ ## Inline Suppression
205
+
206
+ Suppress specific findings with inline comments:
207
+
208
+ ```python
209
+ # Ignore a specific rule on this line
210
+ query = f"SELECT * FROM users WHERE id={user_id}" # aicaudit: ignore S001
211
+
212
+ # Ignore all rules on this line
213
+ eval(user_input) # aicaudit: ignore
214
+ ```
215
+
216
+ ---
217
+
218
+ ## Web UI
219
+
220
+ Prefer a browser over the terminal? Start the built-in web interface:
221
+
222
+ ```bash
223
+ aicaudit web # http://127.0.0.1:8080
224
+ ```
225
+
226
+ The UI is vendored and offline-first — no CDN, no build step, works air-gapped:
227
+
228
+ - **Results browser**: code context with syntax highlighting (server-side
229
+ pygments), the **taint path as a visual stepper** (source → hops → sink),
230
+ AI verdict card with confidence, and diff-preview/apply/rollback for fixes
231
+ - **Live scan progress**: per-file streaming (SSE), no fake progress bars
232
+ - **Scan history**: sessions persist under `.aicaudit/web/`, dashboard shows
233
+ trends across runs
234
+ - Rules browser and AI provider config (relay / OpenAI / Claude / OpenRouter /
235
+ local ollama)
236
+
237
+ API docs at `/docs` (Swagger UI). Found something rough? [Open an issue](https://github.com/IhateStomachProblems/aicaudit/issues).
238
+
239
+ ---
240
+
241
+ ## GitHub Code Scanning Integration
242
+
243
+ AICAudit produces [SARIF 2.1](https://sarifweb.azurewebsites.net/) output compatible with GitHub Code Scanning:
244
+
245
+ ```bash
246
+ aicaudit scan ./src --output sarif > aicaudit.sarif
247
+ ```
248
+
249
+ Upload the result in a GitHub Actions workflow:
250
+
251
+ ```yaml
252
+ - name: Run AICAudit
253
+ run: aicaudit scan . --output sarif > aicaudit.sarif
254
+
255
+ - name: Upload SARIF
256
+ uses: github/codeql-action/upload-sarif@v3
257
+ with:
258
+ sarif_file: aicaudit.sarif
259
+ ```
260
+
261
+ ---
262
+
263
+ ## Caveats
264
+
265
+ AICAudit is a young project. Here are some honest limitations:
266
+
267
+ - **Python only** for now — other languages are planned
268
+ - **Static analysis** — not a runtime security tool
269
+ - **Best-effort** — no tool catches every bug ; please review findings critically
270
+ - The rules reflect common patterns but may not fit every codebase
271
+
272
+ ---
273
+
274
+ ## Performance
275
+
276
+ | Scenario | Time |
277
+ |----------|------|
278
+ | 35 files directory | ~0.13s |
279
+ | Single file | ~0.01s |
280
+
281
+ ---
282
+
283
+ ## Benchmarks
284
+
285
+ Labeled corpus (104 cases: function-level samples + a realistic Flask app with
286
+ planted vulnerabilities), scored against ground truth with an honest
287
+ methodology — including the categories where aicaudit is deliberately
288
+ conservative or deliberately silent. Bandit runs on the same corpus for
289
+ reference; buckets Bandit does not cover are marked.
290
+
291
+ | Rule | aicaudit P/R | Bandit P/R (comparable bucket) |
292
+ |------|-------------|-------------------------------|
293
+ | S001 SQL injection | 100% / 100% | 100% / 86% |
294
+ | S003 dangerous functions | 100% / 100% | 92% / 75% |
295
+ | S004 path traversal | 100% / 100% | no coverage |
296
+ | S005 SSRF | 100% / 100% | no coverage |
297
+ | S007 XXE | 100% / 100% | 100% / 50% |
298
+
299
+ Full table with F1, timing, and the honest-notes section (design tiers,
300
+ conservative flags): [benchmarks/RESULTS.md](benchmarks/RESULTS.md).
301
+ Reproduce and challenge it yourself:
302
+
303
+ ```bash
304
+ python benchmarks/run.py
305
+ ```
306
+
307
+ ---
308
+
309
+ ## Testing
310
+
311
+ - 305 unit tests (pytest)
312
+ - 92% code coverage (pytest-cov)
313
+ - Taint engine: 17 dedicated tests (sources, constness, sanitizers, interprocedural)
314
+ - AI pipeline: fail-safe parsing, transport retry, category-aware prompts all tested
315
+ - Integration tests for CLI, JSON, Markdown, SARIF, Web UI, Chinese output
316
+ - Self-scan validation: we audit our own codebase
317
+ - CI: GitHub Actions on Python 3.10–3.13 (ruff + mypy + coverage + self-scan)
318
+
319
+ ---
320
+
321
+ ## AI Verdicts (evidence-grounded, fail-safe)
322
+
323
+ `aicaudit scan ./src --ai` attaches a verdict to every finding. The LLM sees
324
+ the full enclosing function, the file's imports, and the engine-traced taint
325
+ path — not a one-line snippet — and answers with a structured verdict
326
+ (`is_real`, `confidence`, `cwe`, `reason`, `suggested_fix`).
327
+
328
+ Three verdict states, by design:
329
+
330
+ | status | meaning |
331
+ |--------|---------|
332
+ | `confirmed` | AI agrees it is real, confidence ≥ 0.5 |
333
+ | `false_positive` | AI judged it noise (reason always included) |
334
+ | `unverified` | missing/failed/low-confidence response — **never** auto-confirmed |
335
+
336
+ Why not auto-filter? The best published LLM-verifier configurations still
337
+ wrongly suppress ~22% of true vulnerabilities ([arXiv 2601.22952](https://arxiv.org/abs/2601.22952)),
338
+ so verdicts mark findings instead of deleting them. If you want aggressive
339
+ filtering anyway, it is an explicit opt-in:
340
+
341
+ ```bash
342
+ aicaudit scan ./src --ai --ai-strict # keep only AI-confirmed findings
343
+ ```
344
+
345
+ Prompts are category-aware: injection-class findings (S001/S003/S004/S005/S007)
346
+ get dataflow verification instructions anchored on the taint path;
347
+ policy-class findings (S002/S006/S008) get usage-context instructions —
348
+ targeting the documented failure modes of LLM verifiers (surface pattern
349
+ matching, CWE mislabeling, crypto/policy dismissals).
350
+
351
+ ---
352
+
353
+ ## AI Configuration
354
+
355
+ AICAudit supports multiple LLM providers for AI-powered verification.
356
+
357
+ ### Direct API
358
+
359
+ ```bash
360
+ # OpenAI
361
+ export AICAUDIT_AI_PROVIDER=openai
362
+ export AICAUDIT_AI_KEY=sk-xxx
363
+ aicaudit scan ./src --ai
364
+
365
+ # Claude
366
+ export AICAUDIT_AI_PROVIDER=claude
367
+ export ANTHROPIC_API_KEY=sk-ant-xxx
368
+ aicaudit scan ./src --ai
369
+ ```
370
+
371
+ ### Relay / Proxy Service (中转接口)
372
+
373
+ Any OpenAI-compatible relay service works. Set the provider to `relay` and point to your relay endpoint:
374
+
375
+ ```bash
376
+ # Example: API2D, OhMyGPT, NewAPI, OneAPI, etc.
377
+ export AICAUDIT_AI_PROVIDER=relay
378
+ export AICAUDIT_AI_BASE=https://your-relay.com/v1
379
+ export AICAUDIT_AI_KEY=sk-your-key
380
+ export AICAUDIT_AI_MODEL=gpt-4o-mini
381
+ aicaudit scan ./src --ai
382
+ ```
383
+
384
+ Also accepts `custom` or `proxy` as provider names for the same behavior.
385
+
386
+ ### Local Models
387
+
388
+ ```bash
389
+ export AICAUDIT_AI_PROVIDER=ollama
390
+ aicaudit scan ./src --ai
391
+ ```
392
+
393
+ ---
394
+
395
+ ## License
396
+
397
+ MIT © IhateStomachProblems
398
+
399
+ ---
400
+
401
+ <div align="center">
402
+
403
+ ---
404
+
405
+ # AICAudit 中文版
406
+
407
+ ## 快速开始
408
+
409
+ ```bash
410
+ # 源码安装(v0.2.0 将上架 PyPI)
411
+ pip install git+https://github.com/IhateStomachProblems/aicaudit.git
412
+ aicaudit scan ./项目目录 # 扫描项目
413
+ aicaudit scan ./src --lang zh # 使用中文输出
414
+ aicaudit scan ./src --output json # JSON 输出
415
+ aicaudit scan ./src --output sarif # SARIF 输出(GitHub Code Scanning 兼容)
416
+ ```
417
+
418
+ ## 规则列表
419
+
420
+ **安全**:SQL注入检测、硬编码密钥检测、危险函数检测、路径遍历、SSRF、弱加密、XXE、不安全随机数
421
+ **质量**:裸except、魔法数字、未定义变量、TODO注释、未使用变量
422
+ **性能**:圈复杂度、嵌套深度
423
+
424
+ ## 行内抑制
425
+
426
+ ```python
427
+ # 忽略特定规则
428
+ query = f"SELECT * FROM users WHERE id={user_id}" # aicaudit: ignore S001
429
+
430
+ # 忽略该行所有规则
431
+ eval(user_input) # aicaudit: ignore
432
+ ```
433
+
434
+ ## GitHub Code Scanning 集成
435
+
436
+ ```bash
437
+ aicaudit scan ./src --output sarif > aicaudit.sarif
438
+ ```
439
+
440
+ 在 GitHub Actions 中上传结果:
441
+
442
+ ```yaml
443
+ - name: Run AICAudit
444
+ run: aicaudit scan . --output sarif > aicaudit.sarif
445
+
446
+ - name: Upload SARIF
447
+ uses: github/codeql-action/upload-sarif@v3
448
+ with:
449
+ sarif_file: aicaudit.sarif
450
+ ```
451
+
452
+ ## 注意事项
453
+
454
+ - 目前仅支持 Python — 更多语言正在规划中
455
+ - 纯静态分析,不是运行时安全工具
456
+ - 没有工具能发现所有问题,请结合人工审查
457
+ - 规则反映常见模式,可能不适用于所有代码库
458
+
459
+ ## 测试
460
+
461
+ 305 个单元测试,92% 代码覆盖率:污点引擎专项、AI 判定管线(fail-safe/重试/类别感知 prompt)、Web UI(SSE/pygments/修复回滚/会话持久化)、CI 退出码/init 向导/外部规则目录、CLI/JSON/Markdown/SARIF/中文输出全覆盖。