How waterbottles792/cloudob works
Cloudob Security is an AI customer-support platform with an enterprise operations console. The principal live workflow is browser chat → FastAPI → stateful LangGraph turn: input checks and intent routing lead to retrieval, business-tool handling, or escalation; generated answers are validated and returned with citations and metrics. A separate ingestion path prepares and indexes knowledge documents. The console also presents dashboard, retrieval, guardrail, evaluation, trace, workflow, and administration views; README notes that analytics are illustrative/mock-backed. Some UI-to-backend detail and unsampled subsystems are documented rather than visible in source excerpts.
Main entry points: nodes.py Agent decisions and generation · workflow.py Turn orchestration · retriever.py Hybrid retrieval · ingest.py Document ingestion
Components 23
Operations console
- Assistant chat
frontend/app/(platform)/assistant/page.tsx - Console views
frontend/app/(platform) - Mock analytics
frontend/lib/mock.ts - Frontend API client
frontend/lib/api.ts
API services
- FastAPI endpoints
app/main.py - Live health and documents
app/services.py
Agent workflow
- Turn orchestration
graph/workflow.py - Conversation state
graph/state.py - Agent decisions and generation
graph/nodes.py - Input guardrails
guardrails/input_guardrails.py - Output validation
guardrails/output_guardrails.py - Model prompts
configs/prompts.py - Runtime configuration
configs/settings.py
Knowledge and tools
- Document ingestion
rag/ingest.py - Document loaders
rag/loaders.py - Cleaning and chunking
rag/processing.py - Hybrid retrieval
rag/retriever.py - Vector store factory
vectorstores/factory.py - Configured vector index
- Gemini model factory
llm/factory.py - Support and CRM tools
tools/support_tools.py
Other components
- Support operator
- Human support
Connections 26
- Support operator → Assistant chat: uses
- Assistant chat → Frontend API client: sends chat
- Frontend API client → Mock analytics: falls back to
frontend/lib/api.ts - Console views → Mock analytics: shows analytics
frontend/lib/mock.ts - Frontend API client → FastAPI endpoints: calls endpoints
- FastAPI endpoints → Turn orchestration: runs turn
app/main.py - FastAPI endpoints → Live health and documents: reads health and documents
app/main.py - Turn orchestration → Agent decisions and generation: dispatches nodes
graph/workflow.py - Turn orchestration → Conversation state: threads state
graph/workflow.py - Agent decisions and generation → Conversation state: reads and updates
graph/nodes.py - Agent decisions and generation → Input guardrails: checks input
graph/nodes.py - Agent decisions and generation → Output validation: validates answer
graph/nodes.py - Agent decisions and generation → Hybrid retrieval: retrieves context
graph/nodes.py - Agent decisions and generation → Support and CRM tools: dispatches tools
graph/nodes.py - Agent decisions and generation → Model prompts: uses prompts
graph/nodes.py - Agent decisions and generation → Gemini model factory: invokes models
- Turn orchestration → Human support: escalates to
- Document ingestion → Document loaders: loads files
rag/ingest.py - Document ingestion → Cleaning and chunking: cleans and chunks
rag/ingest.py - Document ingestion → Vector store factory: indexes chunks
rag/ingest.py - Hybrid retrieval → Vector store factory: searches index
rag/retriever.py - Hybrid retrieval → Document ingestion: loads parent store
rag/retriever.py - Vector store factory → Configured vector index: opens configured store
vectorstores/factory.py - Vector store factory → Gemini model factory: gets embeddings
vectorstores/factory.py - Hybrid retrieval → Gemini model factory: compresses results
rag/retriever.py - Live health and documents → Configured vector index: reads chunk counts
app/services.py