I maintain the viero-audio repository on my GitHub, an HTML project that implements an end‑to‑end pipeline for support‑quality, operations‑intelligence, root‑cause analytics, forecast quarantine, and fairness auditing.
The repo was started in September 2026 and last updated in September 2026.
I run the pipeline with Python 3.11 or higher in a virtual environment and install pinned dependencies from requirements.txt.
- Clone the repo:
git clone https://github.com/injetiharsha/viero-audio.git
The pipeline produces an interactive Plotly dashboard, memo numbers JSON, driver analysis, early‑warning simulations, a fairness audit with 1,000 bootstrap iterations, a cost model, submission answers, and hand‑over notes.
- CSAT stability outside Pulse 2: 3.41 – 3.57 across six quarters.
I chose the --no-llm flag to keep execution deterministic and avoid reliance on external API keys, and I selected a 14‑day trailing replacement alert because it would have captured the defect on Dec 1 2025.
- Excess replacements: 598 units.
- Excess cost: Rs 1,088,025 (unit cost Rs 1,820).
- Finance’s erroneous estimate: Rs 2,500.
- Early‑warning detection date: December 1 2025.
- Defective units prevented: 1,741.
- Savings: Rs 8.75 Lakhs to Rs 8.88 Lakhs.
- Fairness audit: 7 of 10 bottom agents exit bottom; 27 of 38 Tier 1 agents indistinguishable from average.
- Create and activate a venv:
python -m venv venvsource venv/bin/activate - Install dependencies:
pip install -r requirements.txt - Place the five CSV files (tickets, agents/agents4, orders, customers, products) into the untracked
data/folder. - Run the deterministic offline pipeline:
python -m src.run_all --no-llm