Research Proposal CLI Evaluator

A command-line tool that uses AI to evaluate research proposals and provide detailed scoring across multiple criteria.

Features

• PDF and DOCX text extraction and analysis
• AI-powered evaluation of 10 research criteria
• Automated scoring system (100 points total)
• Detailed feedback generation
• JSON output format

Installation

1. Clone the repository:

git clone https://github.com/unibeninterns/proposal-cli.git
cd proposal-cli

2. Install Node.js dependencies:

npm install
npm link

(This will also build the TypeScript code)

3. Install Python dependencies:

pip install transformers torch pdfplumber python-docx

Usage

Evaluate a proposal from a file:

proposal-reviewer -f path/to/your/proposal.pdf
# or
proposal-reviewer -f path/to/your/proposal.docx

Evaluate a proposal from text content:

proposal-reviewer -t "This is the text of my research proposal..."

Usage as a Module

This package includes TypeScript support. You can import and use the reviewProposal function directly in your Node.js or TypeScript projects.

1. Install the package:

npm install uniben-ai-proposal-review-cli

2. Import and use the reviewProposal function:

Using CommonJS require:

Evaluating a file:

const { reviewProposal } = require('uniben-ai-proposal-review-cli');

async function evaluateFile() {
  try {
    const filePath = 'path/to/your/proposal.pdf'; // or .docx
    const evaluation = await reviewProposal(filePath, '-f');
    console.log(JSON.stringify(evaluation, null, 2));
  } catch (error) {
    console.error('Evaluation failed:', error);
  }
}

evaluateFile();

Evaluating text content:

const { reviewProposal } = require('uniben-ai-proposal-review-cli');

async function evaluateText() {
  try {
    const textContent = `This is the text of my research proposal...`;
    const evaluation = await reviewProposal(textContent, '-t');
    console.log(JSON.stringify(evaluation, null, 2));
  } catch (error) {
    console.error('Evaluation failed:', error);
  }
}

evaluateText();

Using ES Modules import:

Evaluating a file:

import { reviewProposal } from 'uniben-ai-proposal-review-cli';

async function evaluateFile() {
  try {
    const filePath = 'path/to/your/proposal.pdf'; // or .docx
    const evaluation = await reviewProposal(filePath, '-f');
    console.log(JSON.stringify(evaluation, null, 2));
  } catch (error) {
    console.error('Evaluation failed:', error);
  }
}

evaluateFile();

Evaluating text content:

import { reviewProposal } from 'uniben-ai-proposal-review-cli';

async function evaluateText() {
  try {
    const textContent = `This is the text of my research proposal...`;
    const evaluation = await reviewProposal(textContent, '-t');
    console.log(JSON.stringify(evaluation, null, 2));
  } catch (error) {
    console.error('Evaluation failed:', error);
  }
}

evaluateText();

Example output:

{
  "scores": {
    "relevance": 8,
    "originality": 12,
    "clarity": 7,
    "methodology": 13,
    "literature": 8,
    "team": 9,
    "feasibility": 8,
    "budget": 7,
    "outcomes": 4,
    "sustainability": 4,
    "total": 80
  },
  "comment": "Evaluation Summary:
    
    Strengths:
    - Strong methodology
    - High originality
    - Effective team composition
    
    Areas for Improvement:
    - Could improve outcomes
    - Consider enhancing sustainability"
}

Scoring Criteria

The tool evaluates proposals across these dimensions:

| Criterion | Max Points | Description | |-----------|------------|-------------| | Relevance | 10 | Alignment with priorities | | Originality | 15 | Innovation level | | Clarity | 10 | Problem definition | | Methodology | 15 | Research approach | | Literature | 10 | Literature review | | Team | 10 | Team expertise | | Feasibility | 10 | Implementation viability | | Budget | 10 | Cost justification | | Outcomes | 5 | Expected impact | | Sustainability | 5 | Long-term viability |

Total possible score: 100 points

Requirements

• Node.js 14+
• Python 3.8+
• Text-searchable PDF files or DOCX files
• Maximum file size: 50MB
• Internet connection (for AI model downloads)

Error Handling

The tool will show clear error messages for:

• Missing files
• Invalid file formats
• PDF reading errors
• Missing dependencies

Limitations

• English language only
• Text-based PDFs and DOCX files only
• Requires internet for first run (model download)
• May take 1-2 minutes for initial evaluation

Contributing

Feel free to submit issues and enhancement requests.

License

MIT License - see LICENSE file for details1