IndiGrader is an autograding platform designed for closed-network university programming labs, focusing on isolated execution and secure network configuration.
Important
Evaluation Boundary: The grading engine supports stdin, command-line arguments, and file-based inputs. However, default evaluations are restricted to comparing stdout. Evaluating file outputs requires a custom evaluator script (refer to the Templates section).
Comprehensive documentation is available in the docs/ directory:
IndiGrader utilizes a template-based architecture to support different assignment structures:
- Out-of-the-Box Template: The standard environment for evaluating single-file submissions (C, C++, Python, AWK) and multi-file Makefile projects.
- Custom Evaluator Template: Boilerplate for implementing custom evaluation logic (e.g., partial marking, tolerance thresholds, or format parsing).
IndiGrader's architecture is guided by the following principles:
- Isolated Execution: Grading scripts execute within Firejail sandboxes. This prevents untrusted code from accessing the host network or unauthorized files, ensuring the stability and security of the server environment.
- Offline Deployment: Designed for closed networks, the platform operates independently of external services. It relies on local state management and static assets to minimize dependencies and external points of failure.
- Asynchronous Grading: To handle burst traffic (e.g., hundreds of concurrent submissions at a lab deadline), the system decouples the web server from the grading engine using a Celery task queue and a Redis broker.
- CLI-Centric Workflow: The student interface is provided via a command-line tool (
ig) rather than a web GUI. This design encourages familiarity with terminal environments and integrates directly into the standard development workflow.
- Decoupled Evaluation: A unified evaluation script (
grade.sh) is utilized for both local student testing and server-side grading, ensuring parity between local results and server verdicts. - Input Handling: The engine detects and processes test cases involving standard input streams, command-line arguments, and auxiliary files.
- Language Support: Supports concurrent evaluation of multiple programming languages (C, C++, Python, AWK) within a single lab instance, identifying the target language via file extensions.
- Subnet Binding: Implements IP binding middleware to restrict access. Initial connections bind a student's roll number to a specific IP address for the duration of the session.
- Migration Auditing: Provides a mechanism for machine migration during a lab session, logging rebind events for administrative review.
- Path Isolation: Submission histories are indexed numerically, avoiding direct file path transmission and preventing path-traversal vulnerabilities.
- Global Resource Injection: Administrators can configure shared static resources (e.g., header files, databases) that are automatically mounted into all evaluation sandboxes.
- Data Packaging: Submissions involving multiple files or directories are compressed automatically for transmission.
- Static Leaderboard: Generates an offline, static HTML leaderboard reflecting the current grading state without requiring a secondary web framework.
IndiGrader has been utilized across multiple courses at IIT Madras:
- CS6150 Advanced Programming (Jul-Nov 2025): The core engine was originally developed for this course. Instructors Meghana Nasre and Anantha Padmanabha supported its use as the primary evaluation tool in post-midsem labs. Thanks to the CS6150 TA team, specifically Raj Kumar, for proposing the IP Binding mechanism.
- CS2810 Object Oriented Algorithms, Implementation and Analysis Lab (Jan-May 2026): Adopted for use in a portion of the course curriculum.
- CS1234 Small-Scale Application Development (Jan-May 2026): The platform underwent revisions guided by course instructor Rupesh Nasre. Feedback from CS1234 students (CSE BTech25 batch) contributed to current feature refinement.
- CSE DCF Staff (IIT Madras): Thanks to the CSE Department Computing Facility (DCF) Staff for their assistance with system administration and environment configuration.
This project is licensed under the MIT License.
For suggestions, feature requests, or issue reporting, contact cs24s018 [at] cse [dot] iitm [dot] ac [dot] in or open an issue in the repository.
This project was authored and is maintained by Sanket Tarafder. The documentation for this repository was generated using the Gemini 3.1 Pro (High) LLM model and has been manually reviewed and verified.