Agentic data intelligence using LangChain & Pandas for dataset cleaning, governance, and quality analysis
Tip
Vesper transforms raw messy datasets into governed, analysis-ready data using only simple steps.
Vesper is an autonomous agentic data analyst built with:
- LangChain agent orchestration
- Pandas DataFrame Agent
- Python execution tool
- Dataset governance layer
- Quality measurement engine
It performs real operations on real data:
load → clean → analyze → validate → score → explain
Key Capabilities
- True Pandas execution
- Automated cleaning workflows
- Measurable quality scoring
- Explainable transformations
- Reproducible lineage
- Automated data cleaning
- Dataset quality governance
- Exploratory analysis
- Pre-ML preparation
- BI readiness
Prerequisites
- Python 3.10+
- LLM provider key
# download source code and access directory
git clone https://github.com/TamerDotWork/vesper
cd vesper
# create virtual environment
python -m venv venv
# activation
source venv/bin/activate
# install dependencies
pip install -r requirements.txt
# run vesper
python app.py
# copy template
cp .env.example .env
# edit env
GOOGLE_AoPI_KEY=your_google_api_key_here
Note
Results are stored in full logs.
- Pandas runtime execution
- Safe Python sandbox
- Profiling engine
- Rule validation
- Audit logs
- Missing values
- Duplicates
- Type conflicts
- Outliers
- Schema drift
- Inconsistent categories
- Planner → strategy
- Pandas → execution
- Validator → quality
- Reporter → insights
Recommended models:
- OpenAI GPT-4o
- Claude Sonnet
- Gemini Pro
- Local Llama 3
See the documentation website for full guides.
Give us a star on GitHub if Vesper helps you.
Built with:
- LangChain
- Pandas
- Gemeni
Live testing link on Render:
https://vesper-y3bz.onrender.com/
Warning
Always review AI-applied transformations before production use.
