ETL + Python Analysis

Extracts data from 3 sources in parallel (CSV, JSON, mock API), merges them, runs Python statistical analysis, then generates an LLM executive narrative report.

Pipeline Graph

parallel (3 extracts simultaneously):
  ├── extract-sales    (CSV → structured rows)
  ├── extract-users    (JSON → user records)
  └── extract-metrics  (JSON → system metrics)
    ↓
merge-data
    ↓
analyze-python (Python bridge — statistical aggregations, pure stdlib, no pandas required)
    ↓
generate-narrative (LLM → executive report)

Features

• 3 parallel extracts with no shared state
• Python analysis via the @flomatai/bridge-python subprocess bridge
• Pure Python stdlib (no pandas dependency — works in any environment)
• Synthetic sample data bundled in data/ — works out of the box
• LLM generates executive summary, highlights, recommendations, and risk flags

Setup

cp .env.example .env
# Fill in ANTHROPIC_API_KEY (or OPENAI_API_KEY)
pnpm install
pnpm build

Usage

# Run with bundled synthetic data (Q1 2024 sales/users/metrics)
node dist/src/run.js

# Save report to file
node dist/src/run.js --output ./report.md

# Use your own data files
node dist/src/run.js \
  --sales ./my-data/sales.csv \
  --users ./my-data/users.json \
  --metrics ./my-data/metrics.json \
  --period "Q2 2024"

Data Formats

sales.csv columns: date, product, region, quantity, unit_price, revenue, salesperson_id

users.json: Array of { id, name, plan, signup_date, region, active, sessions_last_30d, spend_lifetime }

metrics.json: { period, api_calls: {...}, performance: {...}, growth: {...} }

Environment Variables

| Variable | Description | |----------|-------------| | ANTHROPIC_API_KEY | Anthropic API key | | OPENAI_API_KEY | OpenAI API key (alternative) | | OPENCODE_BASE_URL | Use local OpenCode proxy | | PYTHON_PATH | Python executable (default: python3) | | SALES_CSV | Path to sales CSV (default: bundled) | | USERS_JSON | Path to users JSON (default: bundled) | | METRICS_JSON | Path to metrics JSON (default: bundled) |