Skip to content

STARTUP.md

Latest commit

 

History

History
179 lines (130 loc) · 4.66 KB

STARTUP.md

File metadata and controls

179 lines (130 loc) · 4.66 KB
Raw

🚀 Research2Code-GenAI — Startup Guide

Prerequisites

1. Clone & Enter the Project

git clone https://github.com/YOUR_USERNAME/Research2Text.git
cd Research2Text-main

2. Create a Virtual Environment

Windows (PowerShell)

python -m venv venv
.\venv\Scripts\activate

Windows (CMD)

python -m venv venv
venv\Scripts\activate.bat

macOS / Linux

python3 -m venv venv
source venv/bin/activate

You should see (venv) at the start of your terminal prompt.

3. Install All Dependencies (Covers All 5 Phases)

pip install --upgrade pip
pip install -r requirements.txt

This single command installs everything for Phases 1–5: PyMuPDF, ChromaDB, Streamlit, PyTorch, Tesseract bindings, Ollama, SymPy, NetworkX, and all dev tools.

Optional: Best-in-class PDF Extraction

These are large packages and only needed if you want the premium extraction pipeline:

# MinerU — preserves LaTeX equations, tables, and reading order
pip install "mineru[all]"

# olmOCR — VLM-based fallback for scanned/complex PDFs
pip install "olmocr[gpu]"

Without these, the pipeline auto-falls back to PyMuPDF (already installed).

4. Setup Ollama (Optional but Recommended)

If you want LLM-powered method extraction and code generation:

# Install a coder model (pick one)
ollama pull deepseek-coder-v2
ollama pull qwen2.5-coder
ollama pull codellama

# Verify it's running
ollama list

The pipeline auto-detects which model is available. If none are installed, it falls back to heuristic extraction.

5. Create Data Directories

mkdir -p data/raw_pdfs data/raw_texts outputs

On Windows CMD:

mkdir data\raw_pdfs data\raw_texts outputs

6. Run the App

Option A: Streamlit UI (Recommended)

cd src
!```

Open **http://localhost:8501** in your browser.

### Option B: Run Tests

```bash
python tests/test_multiagent_pipeline.py

Expected output:

TEST 1 PASSED: All agents imported successfully.
TEST 2 PASSED: Orchestrator initialized with all 10 agents.
TEST 3 PASSED: All agents process mock data correctly.
TEST 4 PASSED: Full pipeline completed successfully.
🎉 ALL TESTS PASSED!

Option C: CLI Pipeline

cd src
python paper_to_code_multiagent.py --pdf ../data/raw_pdfs/your_paper.pdf

7. Quick Test Workflow

  1. Start the app → streamlit run src/app_streamlit.py
  2. Upload a PDF → Drag any research paper PDF into the upload area
  3. Click "Run Multi-Agent Pipeline" → Watch the 9-stage progress tracker
  4. Review results → Check confidence scores, generated code, knowledge graph
  5. Download artifacts → Click "Download Artifacts" for the ZIP bundle

Troubleshooting

Problem Fix
ModuleNotFoundError Make sure venv is activated: .\venv\Scripts\activate
chromadb errors pip install chromadb --upgrade
Empty PDF extraction Install MinerU: pip install mineru[all]
No LLM answer Install Ollama + pull a model: ollama pull codellama
Sandbox validation fails Install PyTorch: pip install torch

Project Structure

Research2Text-main/
├── src/
│   ├── app_streamlit.py        ← Main UI (streamlit run this)
│   ├── agents/
│   │   ├── orchestrator.py     ← 10-agent pipeline coordinator
│   │   ├── ingest_agent.py     ← MinerU/olmOCR/PyMuPDF extraction
│   │   ├── vision_agent.py     ← Image/table batch processing
│   │   ├── method_extractor_agent.py  ← Conformal prediction + LLM
│   │   ├── validator_agent.py  ← Sandbox execution + self-heal
│   │   └── ...                 ← 5 more specialized agents
│   ├── schemas.py              ← Pydantic models + ConfidenceScore
│   ├── code_generator.py       ← Confidence-aware code generation
│   └── ...
├── tests/
│   └── test_multiagent_pipeline.py
├── data/                       ← PDFs and extracted text
├── outputs/                    ← Generated code + pipeline results
├── requirements.txt
├── start.bat                   ← Windows launcher
└── STARTUP.md                  ← This file