Gsoc 2025 proposal

Show & Tell
1

Hello everyone, I am applying for GSOC 2025 with the “AI- Powered Insights for Django Admin Interface” project. I just drafted my proposal and would appreciate any feedback from the community on further improvement.

GSoC 2025 Proposal: Smart Admin
AI-Powered Insights for Django Admin Interface

  1. Overview
    Problem : Django Admin is powerful but lacks modern AI-driven features to help users analyze data, generate insights, or interact with the interface using natural language. Non-technical users often struggle to:
  • Write complex database queries for reports.
  • Identify trends or anomalies in data.
  • Use the Admin interface efficiently without SQL/detailed ORM knowledge.

Solution: Integrate lightweight, privacy-focused AI tools into Django Admin to enable:

  • Natural Language Queries(e.g., “Show top 5 customers in California”).
  • Automated Insights (e.g., anomaly detection, trend summaries).
  • Predictive Actions (e.g., “Flag users likely to churn”).
  1. Goals & Deliverables
    Primary Goals
  2. Add NLP-driven query parsing to translate natural language into ORM queries.
  3. Implement automated report generation and anomaly detection.
  4. Ensure privacy by prioritizing on-device processing (avoid external APIs unless opt-in).

Deliverables

  • Feature 1: Natural Language Query Interface
    • Users type questions (e.g., “Users with >10 orders in 2024”) → system converts to ORM queries.
    • Tools: SpaCy (NLP), Django’s Q objects for query construction.
  • Feature 2: Automated Insights Dashboard
    • Auto-generate charts/statistics for common patterns (e.g., monthly sales trends).
    • Tools: Pandas for data analysis, Plotly for visualizations.
  • Feature 3: Anomaly Detection Hooks
    • Flag outliers (e.g., suspicious login attempts, inventory shortages).
    • Tools: Scikit-learn (Isolation Forest), Django signals.
  • Documentation : Tutorials, security guidelines, and extensibility docs.
  1. Technical Approach
    Phase 1: NLP-to-ORM Query Engine (Weeks 1–4)
  • Build a parser to convert natural language into database queries.
    • Example: “Customers from New York with >5 orders” →
      User.objects.filter(location="New York", order_count__gt=5)
  • Use SpaCy for entity recognition and dependency parsing.
  • Add a secure REST endpoint (optional OpenAI API fallback for complex queries).

Phase 2: Insights & Anomaly Detection (Weeks 5–8)

  • Create an AdminInsights class to auto-generate reports:
    • Time-series analysis (e.g., weekly sales).
    • Integration with Django Admin’s list_display.
  • Add hooks for developers to customize anomaly detection models.

Phase 3: Privacy & Extensibility (Weeks 9–12)

  • Ensure all processing happens locally unless users opt into cloud APIs.
  • Publish a django-ai-admin package (decoupled from Django core for experimentation).
  • Write tests covering edge cases (e.g., ambiguous queries, large datasets).
  1. Timeline

Week 1–4 - NLP query parser, basic ORM translation, proof-of-concept demo.
Week 5–8 -Automated reports, anomaly detection hooks, UI integration.
Week 9–12 - Privacy features, extensibility API, documentation, performance tuning.

  1. Impact
  • Democratize Data Analysis : Non-technical users gain self-service analytics.
  • Modernize Admin: Align Django with tools like Retool/Hex.
  • Privacy First : No data leaves the server unless explicitly allowed.
  • Extensible : Developers can plug in custom models (e.g., Llama 2, GPT-4).