A production-shaped multi-agent system for sales training, deal execution, live assistance, and methodology-aware coaching.
LangGraph · FastAPI · React · Hybrid RAG · Redis · PostgreSQL · OpenTelemetry
SalesBoost is not a single-chatbot demo. It is an end-to-end AI application that separates orchestration, retrieval, memory, model access, evaluation, compliance, and product workflows so each layer can be tested and evolved independently.
The product covers the complete sales loop: training → battle preparation → live assistance → post-call review → pipeline / executive analytics.
| Surface | What it demonstrates |
|---|---|
| AI Sales Training | NPC simulation, coaching, strategy evaluation |
| Battle Prep | methodology state, gaps, talking points |
| Live Assist | real-time AI copilot workflow |
| Pipeline | deal lifecycle with MEDDPICC / SPIN / Challenger |
| Executive Cockpit | funnel and methodology-level analytics |
Implemented product pages and APIs are linked from the full technical README.
flowchart LR
UI[React / Vite] --> API[FastAPI]
API --> TENANT[Tenant Middleware]
TENANT --> COORD[Production Coordinator]
COORD --> GRAPH[Dynamic LangGraph Workflow]
GRAPH --> ROUTER[Intent Router]
GRAPH --> NPC[NPC Agent]
GRAPH --> COACH[Coach Agent]
GRAPH --> COMP[Compliance Agent]
ROUTER --> RAG[Hybrid RAG]
RAG --> VECTOR[Dense Retrieval]
RAG --> BM25[BM25]
RAG --> RERANK[BGE Reranker]
GRAPH --> MEM[Context / Memory]
GRAPH --> GW[Model Gateway]
MEM --> REDIS[(Redis)]
API --> PG[(PostgreSQL)]
GW --> MODELS[LLM Providers]
API --> OBS[OpenTelemetry / Prometheus]
| Capability | Repository evidence |
|---|---|
| Dynamic multi-agent workflow | dynamic_workflow.py |
| Production coordinator facade | production_coordinator.py |
| Hybrid retrieval | vector_store.py · bm25_retriever.py |
| Self-RAG / reflection | self_rag.py |
| HyDE retrieval | hyde_retriever.py |
| Multi-tier context / memory | memory.py |
| Self-correcting tool execution | reflection.py · executor.py |
| Multi-provider model gateway | model_gateway.py |
| Golden regression tests | test_golden_regression.py |
| Observability | otel_tracing.py · prometheus_exporter.py |
git clone https://github.com/Benjamindaoson/SalesBoost.git
cd SalesBoost
# backend
cd backend
pip install -r requirements.txt
python main.py
# frontend — new terminal
cd frontend
npm install
npm run devConfigure .env for the model provider and optional PostgreSQL / Redis services. The complete setup, configuration, architecture notes, feature matrix, and roadmap remain in the full README.
Agent orchestration · retrieval · memory · evaluation · observability · product delivery