How khenryaegis/vulnbot works
VulnBot is an LLM-driven penetration-testing framework with a multi-agent assessment workflow and a separately exposed knowledge-base/chat capability. The documented user flow starts a VulnBot run; its role layer plans and executes assessment work, including remote shell operations, while chat and knowledge-base interfaces support retrieval and administration. The graph preserves the major documented RAG and agent subsystems. Some implementation files were not sampled, so uncertain links are omitted; cited source-index references and README statements support the relationships shown.
Main entry points: role.py Role orchestrator · milvus_kb_service.py Milvus KB service · chat.py Chat service · planner.py Assessment planner
Components 33
User Interfaces
- CLI
cli.py - Chat service
server/chat/chat.py - KB API routes
server/api/kb_route.py - Knowledge UI
web/knowledge_base/knowledge_base.py
Pentest Agents
- Role orchestrator
roles/role.py - Collector agent
roles/collector.py - Scanner agent
roles/scanner.py - Exploiter agent
roles/exploiter.py - Assessment planner
actions/planner.py - Plan updates
actions/write_plan.py - Plan summary
actions/plan_summary.py - Command generation
actions/write_code.py - Task execution
actions/execute_task.py - Remote shell
actions/remote_shell.py - Agent prompts
prompts/prompt.py
Knowledge Retrieval
- Document API
rag/kb/api/kb_doc_api.py - KB service layer
rag/kb/base.py - Milvus KB service
rag/kb/service/milvus_kb_service.py - Load and split files
rag/kb/utils/kb_utils.py - Embedding provider
rag/embedding/embedding.py - Vector retriever
rag/retriever/milvus_vectorstore.py - Document reranker
rag/reranker/reranker.py
State and Configuration
- MySQL database
- DB session layer
utils/session.py - Message repository
db/repository/message_repository.py - Plan repository
db/repository/plan_repository.py - Plan model
db/models/plan_model.py - Task model
db/models/task_model.py - Runtime configuration
config/config.py
Other components
- Operator
- LLM provider
- Assessment target
- Milvus vector store
Connections 36
- Operator → CLI: starts run
- Operator → Chat service: submits query
- Operator → Knowledge UI: manages knowledge
- CLI → Role orchestrator: runs assessment
- Role orchestrator → Assessment planner: plans tasks
roles/role.py - Role orchestrator → Plan summary: summarizes plan
roles/role.py - Role orchestrator → Command generation: requests commands
roles/role.py - Role orchestrator → Plan repository: loads and saves plans
roles/role.py - Role orchestrator → Agent prompts: uses prompts
roles/role.py - Assessment planner → Plan updates: writes plan
actions/planner.py - Assessment planner → Agent prompts: uses prompts
actions/planner.py - Assessment planner → Task model: creates tasks
actions/planner.py - Command generation → Task execution: dispatches code
actions/write_code.py - Command generation → Agent prompts: uses prompts
actions/write_code.py - Task execution → Assessment target: executes assessment
- Remote shell → Assessment target: connects remotely
- Chat service → LLM provider: requests completion
server/chat/chat.py - Chat service → Document API: searches documents
server/chat/chat.py - Chat service → Document reranker: uses reranker
server/chat/chat.py - Chat service → Message repository: loads and saves messages
server/chat/chat.py - KB API routes → Document API: routes requests
server/api/kb_route.py - Knowledge UI → KB service layer: uses KB service
web/knowledge_base/knowledge_base.py - Document API → KB service layer: manages documents
rag/kb/api/kb_doc_api.py - Document API → Load and split files: uses file utilities
rag/kb/api/kb_doc_api.py - KB service layer → Milvus KB service: delegates vector work
rag/kb/base.py - Milvus KB service → Embedding provider: creates embeddings
rag/kb/service/milvus_kb_service.py - Milvus KB service → Vector retriever: searches vectors
rag/kb/service/milvus_kb_service.py - Milvus KB service → Milvus vector store: reads and writes vectors
rag/kb/service/milvus_kb_service.py - Milvus KB service → Load and split files: uses file metadata
rag/kb/service/milvus_kb_service.py - Message repository → DB session layer: opens DB session
db/repository/message_repository.py - DB session layer → MySQL database: commits transactions
utils/session.py - Plan repository → Plan model: loads and stores plans
db/repository/plan_repository.py - Plan model → Task model: contains tasks
db/models/plan_model.py - Plan repository → DB session layer: uses DB session
db/repository/plan_repository.py - Runtime configuration → DB session layer: configures database
utils/session.py - Runtime configuration → Milvus KB service: configures vector store
rag/kb/service/milvus_kb_service.py