Intelligent Sprint Analysis Using Agentic System for Startup Projects

About

This project is a local-first sprint intelligence system for GitHub-based software teams. It combines feature extraction, retrieval-augmented generation, and multi-agent LLM reasoning to analyze sprint activity, identify risks, and generate evidence-backed recommendations.

The application runs against local Ollama models and can use GitHub data for ingestion and analysis workflows.

Key Features

• Data Ingestion: Automatically collect and process GitHub repositories (issues, pull requests, commits, discussions)
• Multi-Agent Analysis: 11-agent LLM pipeline for feature extraction, risk assessment, sprint forecasting, and recommendations
• Retrieval-Augmented Generation: Context-aware analysis powered by ChromaDB vector store and local Ollama LLMs
• Interactive UI: Configure analysis parameters, select local Ollama models, view sprint insights and recommendations
• Local-First: All processing happens locally—no cloud dependencies, no external API calls to LLM providers

Quick Start Workflow

1. Install & Setup (5 mins)

   • Install dependencies and configure environment variables
   • Start Ollama server locally

2. Ingest Data (5–10 mins)

   • Navigate to the Ingestion tab
   • Provide GitHub organization URL and repository names
   • Configure ingestion parameters (synthetic data, data generation settings)
   • Click "Run Pipeline"

3. Analyze Sprints (5–15 mins)

   • Go to the Sprint Analysis tab
   • Select repository and upload sprint JSON data, or use demo data
   • Choose evaluation mode (Resilient, Synthetic, Real)
   • Select desired local Ollama model
   • Click "Run Inference" to trigger the agent pipeline
   • View results and recommendations

System Architecture

Ingestion Pipeline

Sprint Analysis Pipeline

Methodology

The sprint intelligence system follows a multi-stage pipeline:

graph TD
    A["GitHub Repository<br/>(Issues, PRs, Commits)"] -->|Data Ingestion| B["Data Collection & Processing<br/>(ChromaDB Vector Store)"]
    
    B -->|Feature Extraction| C["Data Pipeline<br/>(Metrics, Activity, Dependencies)"]
    
    C -->|Context Preparation| D["Retrieval-Augmented Generation<br/>(RAG)"]
    
    D -->|Input to 11-Agent Pipeline| E["Multi-Agent Analysis"]
    
    E -->|Router Agent| F["Data Collector"]
    E -->|Feature Extraction Agent| G["Feature Engineer"]
    E -->|Risk Detection Agent| H["Risk Assessor"]
    E -->|Planning Agent| I["Sprint Planner"]
    E -->|Recommendation Agent| J["Recommender"]
    
    F --> K["Consolidated Analysis"]
    G --> K
    H --> K
    I --> K
    J --> K
    
    K -->|Sprint Health Score| L["Sprint Intelligence Report"]
    K -->|Risk Metrics| L
    K -->|Recommendations| L
    
    L --> M["Interactive UI<br/>(Sprint Analysis Tab)"]
    
    style A fill:#e1f5ff
    style B fill:#fff3e0
    style C fill:#fff3e0
    style D fill:#f3e5f5
    style E fill:#f3e5f5
    style K fill:#e8f5e9
    style L fill:#fff9c4
    style M fill:#c8e6c9

Loading

Key Stages:

1. Data Ingestion: Collect GitHub repository data (issues, PRs, commits, discussions)
2. Feature Extraction: Process raw data into actionable sprint metrics
3. Retrieval-Augmented Generation (RAG): Use ChromaDB vector store to provide context-aware prompts
4. Multi-Agent Analysis: 11 specialized agents analyze different aspects in parallel
5. Report Generation: Consolidate findings into sprint health scores, risk assessments, and actionable recommendations

Setup

Prerequisites

• Python 3.11+
• Git
• Ollama (install from ollama.ai)
  • Local Ollama server must be running at http://localhost:11434
  • Default model: qwen3:0.6b (auto-installed during setup)
• GitHub Token (required for GitHub API ingestion)
  • Create a personal access token at github.com/settings/tokens

Step 1: Clone & Environment

git clone https://github.com/bibekgupta3333/repo-sprint.git
cd repo-sprint

Step 2: Python Environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Step 3: Configure Environment Variables

Create a .env file or export these variables:

# GitHub API access (required for ingestion)
export GITHUB_TOKEN=your_github_personal_access_token

# Ollama LLM configuration (optional—defaults shown)
export OLLAMA_BASE_URL=http://localhost:11434
export OLLAMA_MODEL=qwen3:0.6b

Note: The app reads either GITHUB_TOKEN or GH_TOKEN for GitHub authentication.

Step 4: Start Ollama

In a separate terminal:

ollama serve
ollama pull qwen3:0.6b

This starts the Ollama server at http://localhost:11434 and downloads the default model.

Step 5: Run the Application

uvicorn src.app:app --host 0.0.0.0 --port 8000 --reload

Or use the npm script:

npm run fastapi

Open your browser to http://localhost:8000

Navigation Guide

Ingestion Tab

• Provide GitHub organization URL (e.g., https://github.com/mintplex-labs/)
• List repositories to ingest (comma-separated)
• Adjust ingestion parameters as needed
• Click Run Pipeline to start data collection and processing

Sprint Analysis Tab

• Select repository owner and name
• Upload sprint JSON file, or use Demo (Mock Data) for testing
• Choose evaluation mode:
  • Resilient (Graceful fallbacks): Recommended for most use cases
• Select a local Ollama model from the dropdown
• Click Run Inference
• View analysis results and sprint recommendations

About System Tab

• View system status and configuration details

Customizing the Ollama Model

You can change the Ollama model at runtime:

1. In the Sprint Analysis tab, select a different model from the Model Selector dropdown
2. Or update the OLLAMA_MODEL environment variable before starting the app
3. Ensure the selected model is installed locally: ollama pull <model-name>

Popular lightweight models:

• qwen3:0.6b (default, ~400MB)
• llama2:7b (~4GB)
• mistral:7b (~4GB)

Notes

• GitHub token required: The app needs a GitHub personal access token for API-based data ingestion. You can ingest local repositories without a token, but GitHub API workflows require authentication.
• Ollama must be running: Always start ollama serve in a separate terminal before starting the application.
• Model selection UI: The Sprint Analysis tab includes a dropdown to select from installed Ollama models at runtime. Your selection applies only to that analysis run.
• Troubleshooting: If models don't appear in the dropdown, check that Ollama is running at http://localhost:11434 and that models are installed locally.