Apppublic

sankardhanushkodi/call_center_summary_agent

sourceHugging Faceupdated 7mo agoView on Hugging Face

0likes

App README

Call Center Summary Agent

A multi-agent LangGraph pipeline that transcribes, summarizes, and quality-scores call center recordings with a Streamlit frontend.

Pipeline

Upload audio / paste transcript
          ↓
  [Intake Agent]      — validates format, extracts metadata
          ↓
  [Transcription Agent] — OpenAI Whisper API → text
          ↓
  [Summarization Agent] — summary, key points, action items, sentiment
          ↓
  [Quality Agent]     — 5-dimension rubric score (0–100)
          ↓
  [Routing Agent]     — complete / flag for review / escalate / retry

Agents

Agent	Type	Responsibility
Intake	Pure function	Validate input format/size, extract file metadata
Transcription	Pure function	OpenAI Whisper API (audio) or text passthrough
Summarization	LLM + Pydantic	Summary, key points, action items, category, sentiment
Quality Scoring	LLM + Pydantic	Tone · Professionalism · Resolution · Empathy · Clarity (each 0–20)
Routing	Pure function	Deterministic rules → complete / flag / escalate / retry

Quality Rubric

Dimension	Max	Description
Tone	20	Friendliness, calmness, warmth
Professionalism	20	Procedure adherence, accurate info, appropriate language
Resolution	20	Issue effectively identified and resolved
Empathy	20	Customer feelings acknowledged
Clarity	20	Clear communication, no jargon
Overall	100	Sum of all dimensions

Routing thresholds: ≥70 → Complete · ≥50 → Flag for Review · <50 → Escalate/Retry

Deploy to Hugging Face Spaces

1.Create a new Space → Docker SDK at huggingface.co/new-space
2.Push the repo:

bash

   git init
   git add .
   git commit -m "Initial commit"
   git remote add hf https://huggingface.co/spaces/<your-username>/call-center-agent
   git push hf main

1.Add your API key under Settings → Repository Secrets:

Secret name	Value
`OPENAI_API_KEY`	`sk-...`
`LLM_MODEL`	`gpt-4o-mini`

HF Spaces will build the Docker image automatically and expose the app on port 7860.

Local Docker build

bash

docker build -t call-center-agent .
docker run -p 7860:7860 \
    -e OPENAI_API_KEY=sk-... \
    -e LLM_MODEL=gpt-4o-mini \
    call-center-agent
# → open http://localhost:7860

Local development

bash

bash setup.sh          # creates .venv, installs deps, copies .env
# edit .env — add OPENAI_API_KEY
source .venv/bin/activate
streamlit run app.py

Disclaimer

Transcriptions and AI-generated summaries are for informational purposes only. Always verify quality scores with human review before using them for agent evaluation.