Quick Start Guide

Relevant source files

The following files were used as context for generating this wiki page:

• README.md
• app.py
• orchestrator.py
• scenarios.py
• reset_demo.py
• agents/triage_agent.md
• agents/manager_briefing_agent.md

The BrightPath AI Operations System is a lightweight multi-agent pipeline designed for training companies to automate inbound sales leads and customer service requests. Powered by Claude (Anthropic), the system processes messages through a three-stage pipeline to classify intent, draft professional responses, and generate manager briefings. Sources: README.md:1-12

The system utilizes a hybrid storage approach, combining a SQLite database for structured event logging and task management with Markdown-based "memory" files that provide long-term context for the AI agents. Sources: README.md:14-20, app.py:114-142

System Architecture

The core of BrightPath is a sequential pipeline where data flows through specialized AI agents. Each agent is defined by a Markdown prompt file in the agents/ directory, allowing for behavioral customization without modifying Python code. Sources: README.md:22-42, README.md:113-116

The following diagram illustrates the standard processing flow for an inbound message:

The pipeline ensures that every message is classified by the Triage Agent before being routed to a specialist for action. Sources: orchestrator.py:236-285, app.py:321-380

Setup and Installation

Prerequisites

• Python 3.x
• Anthropic API Key (Claude Sonnet and Haiku models)
• Dependencies: anthropic, rich, streamlit

Sources: README.md:126-130

Installation Steps

1. Install Dependencies:

   pip install -r requirements.txt
   

2. Configure Environment:

   export ANTHROPIC_API_KEY=your_key_here
   

3. Initialize System:
   The system automatically creates the SQLite database (brightpath.db) upon its first run. Sources: README.md:73-81, app.py:114-124

Running the System

BrightPath offers two primary interfaces: a web-based dashboard for visual interaction and a terminal-based orchestrator for automated or CLI-driven processing.

1. Streamlit Web UI

The recommended method for demonstrations and manual oversight.

streamlit run app.py

This interface provides four tabs: Run Pipeline, Tasks & History, Memory Files, and CRM Data. Sources: README.md:83-93

2. Terminal Orchestrator

A lightweight version suitable for CLI use or integration.

python orchestrator.py

The terminal version includes a menu to select from 18 pre-defined test scenarios or input custom messages. Sources: orchestrator.py:326-340, scenarios.py:10-100

Core Components

AI Agent Pipeline

The system utilizes three distinct agent roles to process data:

Agent	File	Primary Responsibility
Triage	agents/triage_agent.md	Classifies message, assesses urgency/sentiment, and routes to specialists.
Sales	agents/sales_agent.md	Qualifies leads, identifies upsell opportunities, and updates CRM.
CS	agents/customer_service_agent.md	Handles booking changes, refunds, and complaints based on company policy.
Manager Briefing	agents/manager_briefing_agent.md	Summarizes agent actions into a punchy, Telegram-ready format.

Sources: agents/triage_agent.md:3-9, agents/sales_agent.md:3-9, agents/customer_service_agent.md:3-10, agents/manager_briefing_agent.md:3-9

Database Schema

The SQLite database stores persistent operational data across three main tables:

Development and Maintenance

Testing with Scenarios

The system includes 18 pre-defined scenarios in scenarios.py to test various edge cases, such as "Angry customer complaint," "VIP upsell," and "Duplicate lead." Sources: scenarios.py:12-70

Resetting the Environment

To restore the system to a clean state for a new demo or after extensive testing, use the reset script:

python reset_demo.py

This script performs two main actions:

• Restores the .md files in the memory/ directory from the memory_backup/ folder.
• Clears the events, tasks, and crm_log tables in the SQLite database while resetting auto-increment counters.

Sources: reset_demo.py:19-55

Memory Persistence

Agents read context from five Markdown files located in the memory/ directory:

• company_context.md: Internal policies and pricing.
• customer_history.md: Interaction history and sentiment tracking.
• sales_notes.md: Pipeline status and activity logs.
• open_tasks.md: Outstanding actions.
• decision_log.md: Log of prior AI-driven decisions.

Sources: README.md:102-111, app.py:228-235

Conclusion

The BrightPath AI Operations System provides a robust framework for automating small business operations through specialized AI agents. By combining structured database logging with flexible Markdown-based memory, the system maintains context across interactions while providing a clear audit trail for human managers. Professional oversight is facilitated through the Manager Briefing Agent and the unified Streamlit dashboard. Sources: README.md:1-12, app.py:535-645