economist-agents

Economist Agents

Multi-Agent Content Generation Pipeline

A sophisticated multi-agent system (6 AI personas collaborate) that produces publication-quality blog posts in The Economist's signature style. This project orchestrates specialized AI agents for topic discovery, editorial voting, research, writing, editing, and chart generation.

🚀 Project Status

Current Phase: Integration and Production Deployment (Sprint 15 - Day 4) Current Quality Score: 30/100 (red) - Test suite needs stabilization Sprint 15 Progress: Story 9 COMPLETE ✅, Story 10 IN PROGRESS (5/13 pts, 38%)

Recent Achievement (Sprint 14 - Complete ✅):

• Flow-Based Orchestration: Deterministic state-machine (@start/@listen/@router) - 9/9 tests passing
• Style Memory RAG: ChromaDB integration with <200ms query latency
• ROI Telemetry: Business value tracking (8,333x efficiency gain validated)
• Sprint Velocity: 82% faster than estimates (5h vs 28.5h) with 10/10 quality score

Foundation:

• Defect Prevention: Automated system catching 83% of historical bug patterns
• Green Software: Self-validating agents reducing token waste by 30%
• Quality Gates: 4-layer validation (automated checkpoints enforce standards)

Test Suite Status: 2112 passed, 84 skipped

🛠️ Installation

Python Version Requirement: Python ≥3.11 (tested with 3.12)

1. Clone the repository:

   git clone https://github.com/your-org/economist-agents.git
   cd economist-agents
   

2. Create virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   pip install --upgrade pip
   

3. Install dependencies:

   pip install -r requirements.txt
   pip install -r requirements-dev.txt
   

   Note: Includes arXiv integration for fresh academic research sources

4. Configure environment:

   cp .env.example .env
   # Edit .env with your ANTHROPIC_API_KEY
   ```1 total)
   

Core Agents:

Agent	Role	Responsibility
code-quality-specialist	Code Quality	TDD-based refactoring and quality standards enforcement
devops	Infrastructure	CI/CD automation and deployment
git-operator	Version Control	Git operations and repo management
po-agent	Product	Backlog refinement and story creation
product-research-agent	Research	Market analysis and competitive intelligence
scrum-master	Orchestration	Sprint coordination and ceremonies
test-specialist	Testing	Comprehensive test writing and quality assurance
visual-qa-agent	Visual Quality	Chart and design validation

Pipeline Runners (Anthropic Agent SDK)

Stage 3 (src/agent_sdk/stage3_runner.py) — Content Generation

• Purpose: Research, writing, and chart creation pipeline
• Research: Deterministic web search (arXiv + Google Scholar) — no LLM in research path
• Agents: Writer Agent → Graphics Agent → Editor Agent (all Claude/Anthropic)
• Stat Audit: Strips sentences with stats not found in the research brief
• Output: Article with verified sources and chart data
• Integration: Via the sequential pipeline in src/agent_sdk/pipeline.py

Stage 4 (src/agent_sdk/stage4_runner.py) — Editorial Review

• Purpose: Deterministic post-processing quality gates
• Output: JSON with quality assessment and edited article