GPT-based Call QC App (Configurable Rubrics)

A lightweight Streamlit app that scores and audits call transcripts using OpenAI GPT with your own API key. You can define your own QC metrics/rubrics in YAML and run single-call or batch evaluations.

Features

• Paste or upload transcripts (or upload an audio file to transcribe).
• Your OpenAI API key stays local (entered in the UI; not stored on disk by default).
• Fully configurable metrics via YAML (weights, rubrics, thresholds).
• Structured JSON output (per-metric scores, rationales, quotes, red flags, summary).
• Batch mode: upload CSV with columns: call_id, transcript (optional audio_path).
• Download results as JSON or CSV.

Quickstart

python -m venv .venv && source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -r requirements.txt
streamlit run app.py

Open the local URL shown by Streamlit.

OpenAI Models

By default the app suggests a model like gpt-4o-mini. You may change it in Settings.
See OpenAI docs for the latest models & pricing. (Ref: OpenAI API docs).

Files

• app.py – Streamlit UI.
• qc_evaluator.py – Prompt builder and evaluator.
• config/qc_metrics.example.yaml – Example Zomato-style QC rubric.
• samples/batch_sample.csv – Example batch input.

Notes

• Audio transcription uses OpenAI's transcription endpoint. If unavailable in your region or account, paste transcripts directly.
• This is a reference app. Review prompts/rubrics for your policies and compliance.