Why AuditPilot
很多 Agent 项目停留在聊天框或 Demo。AuditPilot 选择一个更“硬”的落地场景：企业审计交付。它需要证据、控制、风险、复核、报告和整改闭环，也天然要求可追溯、可回归、可解释。
AuditPilot 的目标不是替代审计师，而是把审计师反复执行的取证、映射、检查、补证和交付动作，组织成一套可治理的 Agent 工作流。
What it does
| 模块 | 能力 |
|---|---|
| Audit Workspace | 审计立项、控制矩阵、审计程序、抽样计划、发现、整改和交付包。 |
| Agent Runtime | Plan / Execute / Reflect、任务产物、失败恢复、安全门和运行指标。 |
| Agentic RAG | 知识写入、切块、检索、来源引用、证据质量门和缺证提示。 |
| Skills / MCP-style Tools | 工具 Schema、权限声明、TTL 缓存、熔断器、调用日志和工具指标。 |
| Memory | Working / Episodic / Profile Memory，保留多轮审计上下文。 |
| Evaluation Harness | Agent / RAG / Research 评测、基线对比、release gate 和 badcase 沉淀。 |
| Interview-driven Diagnostics | 将大厂 Agent 面经/JD 中常问的 Runtime、RAG、Tool、Memory、评测和生产化问题转成可执行诊断。 |
Screenshots
| Audit workspace | Agent runtime |
|---|---|
 |
 |
| Agent collaboration | Mobile overview |
|---|---|
 |
 |
Architecture
Audit request
-> Hybrid Intent Router
-> Working + Episodic + Profile Memory
-> Planner / Evidence / Control / Risk / Compliance / Remediation Agents
-> Agentic RAG + Knowledge Graph + Skills / MCP-style Tools
-> Safety Gate + Reflection + Human Review
-> Audit Repository + Evaluation Baseline + Delivery Package
Design boundaries:
- LLMs help with understanding, summarization and explanation.
- Evidence gaps, quality gates, permissions, risk signals and delivery state stay auditable.
- High-risk or low-confidence outputs are routed to evidence补充 and human review.
Quick start
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
Copy-Item config.env.example config.env
python start.py
The system can run in deterministic fallback mode without model keys. Add an OpenAI-compatible provider only when you want LLM-enhanced analysis.
Security
Do not commit real API keys.
- Put local secrets in
config.env. - Use platform secrets for deployment.
config.env,.env*, runtime data, logs, model artifacts and local databases are ignored by Git.- Example config files use placeholders only.
Validation
.\.venv\Scripts\python.exe -m compileall -q agents services rag web tests scripts
.\.venv\Scripts\python.exe -m unittest discover -s tests -p "test_*.py"
Current tests cover intent routing, memory compaction, skill validation/cache, runtime reflection, evaluation regression, self-evolution harness, quality diagnostics and global search.
Project layout
agents/ audit agent chain and control library
services/ runtime, memory, router, skills, safety, evaluation, delivery
rag/ agentic RAG and persisted knowledge base
knowledge_graph/ optional graph construction and Neo4j adapter
training/ evaluation and offline training entry points
web/ FastAPI application and APIs
templates/ + static/ product UI
tests/ regression tests
docs/ project archive, interview material and screenshots
data/ local runtime data, mostly ignored by Git
Reality notes
- Runtime records, tool calls, evaluations, memory and audit cases are generated by real code and persisted locally.
- Untested throughput, accuracy or availability claims are intentionally not presented as project facts.
- SFT / RLHF-style model training is treated as an offline extension; the Web app focuses on workflow, evaluation and data preparation.