This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend I built AcademAI for someone very close to my heart: my wife. She is currently finishing her undergraduate thesis in S1 PAUD (Pendidikan Anak Usia Dini — Early Childhood Education) in Indonesia while also managing our
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend I built AcademAI for someone very close to my heart: my wife. She is currently finishing her undergraduate thesis in S1 PAUD (Pendidikan Anak Usia Dini — Early Childhood Education) in Indonesia while also managing our home and daily family life. Every evening, I would sit beside her at the kitchen table and watch her struggle with the overwhelming friction of academic writing: AI Hallucinations: General AI chatbots repeatedly hallucinated scientific literature — inventing non-existent papers attributed to Jean Piaget, Lev Vygotsky, and Maria Montessori. Painful Citation Cross-Checking: Verifying whether citations were genuine and formatted properly in APA 7th took hours of manual cross-referencing across journals. Lack of Local Pedagogical Context: Commercial AI tools had zero understanding of Indonesian national early childhood education frameworks, such as STPPA (Permendikbudristek No. 5/2022) or the standard developmental assessment rubrics (BB / MB / BSH / BSB). Classroom Action Research (PTK) Statistics: In her action research, computing paired sample t-tests and Hake's Normalized Gain (N-Gain), then translating numeric output into formal Indonesian academic prose for Chapter IV (Bab IV) was a constant source of stress. Turnitin Anxiety: Meeting strict university plagiarism thresholds (<15%) required exhausting manual structural paraphrasing. Once AcademAI solved these hurdles for my wife, I asked myself: Why stop at Early Childhood Education? Every undergraduate and master's student in Indonesia faces this exact same friction, whether they are writing a thesis in Economics, Law, Engineering, or Health Sciences. I redesigned the system around an open DISCIPLINES Registry that dynamically tailors the AI's scientific persona, canonical theories, and research methodology across 12 distinct faculties: Academic Faculty Embedded Canonical Theories & Standards Supported Methodologies 🧸 PAUD (Early Childhood) Piaget, Vygotsky, Montessori, Ki Hajar Dewantara, STPPA Permendikbud No. 5/2022 PTK (Kemmis & McTaggart), Rubrik BB/MB/BSH/BSB 📚 Education & Teaching Behaviorism (Skinner), Constructivism, Kurikulum Merdeka, Bloom's Taxonomy R&D (ADDIE / 4D), Quasi-Experiment, PTK 💼 Economics & Business Kotler & Keller, Porter's Five Forces, Jensen & Meckling (Agency Theory) SEM / SmartPLS, Multiple Linear Regression, Classical Assumption ⚖️ Law & Jurisprudence Theories of Justice, Legal Certainty, Utilitarianism (Bentham) Normative Juridical, Empirical Juridical, Statutory Approach 💻 Computer Science & IT IEEE / ACM Guidelines, Software Architecture, Algorithmic Complexity SDLC (Agile / Waterfall), Usability Testing (SUS), Black-box 🏥 Health & Nursing Evidence-Based Practice (EBP), Bioethics, Epidemiological Models Cross-Sectional, Case-Control, Cohort, Clinical Observational 🧠 Psychology Psychometric Validity, Social Cognitive Theory, Big Five Personality Scale Development (Likert), Factor Analysis, Experimental Design ⚙️ Engineering Technical Engineering Standards, SNI / ISO Specifications, Quality Control Design & Prototyping, Finite Element, Laboratory Testing 🌾 Agriculture & Agrotech Agronomy, Soil Fertility, Integrated Pest Management (IPM) Completely Randomized Design (RAL / RAK), ANOVA, Duncan Test 💬 Communication Science Agenda Setting, Framing Analysis, Cultural Semiotics (Barthes) Framing Analysis (Entman), Critical Discourse (Fairclough) 🏛️ Social & Political Science Critical Theory (Habermas), Social Capital (Bourdieu), Public Policy Phenomenology, Grounded Theory, Evaluative Policy Research 🌐 General Academic Philosophy of Science (Ontology, Epistemology, Axiology), PRISMA Protocol Universal IMRaD, Systematic Literature Reviews (SLR) Grounding in Real Scientific Literature: Injects real-time queries against Google Scholar, grounding every paragraph in verified, indexed papers. Automated Citation Validator: Parses in-text citations and DOI links, cross-checking them against scholarly indices and outputting validation badges (VALID, PARTIAL, or INVALID). Zero-Dependency Statistics Engine: Computes Paired t-Tests and Hake (1999) N-Gain categories (Tinggi, Sedang, Rendah) from raw pretest/posttest scores and automatically generates publishable Indonesian academic analysis ready to insert into Bab IV. Academic Paraphrasing & Plagiarism Auditor: Employs syntactic nominalization and passive-voice academic transformations to safely lower Turnitin similarity (<15%) while preserving original scientific meaning. Full Thesis Generator: Produces an end-to-end 5-chapter draft with a bilingual abstract (Indonesian + English), research matrix, and APA 7th bibliography in a single pass. 🌐 Live Web Application: https://academicai-production-a41d.up.railway.app 📡 Live Health Endpoint: https://academicai-production-a41d.up.railway.app/healthz How to Explore the 5 Dedicated Academic Modules: Tab 1 — Chat & Drafting: Select any of the 12 disciplines (PAUD, Economics, Law, Health, etc.), choose your research workflow (Drafting, Systematic Literature Review, Proposal, Abstract, or Statistics), and ask a question. Notice how every output cites real, indexed papers retrieved on the fly. Tab 2 — Full Thesis Generator: Input a thesis topic (e.g., "Pengaruh Literasi Keuangan Terhadap Kinerja UMKM" or "Efektivitas Media Loose Parts Pada Perkembangan Kognitif Anak"). AcademAI produces a complete 5-chapter draft tailored to that discipline's research standards. Tab 3 — Citation Validator: Paste any academic paragraph containing in-text citations (such as (Piaget, 1976) or (Kotler, 2021)) to receive an instant line-by-line validation audit (VALID, PARTIAL, or INVALID). Tab 4 — Plagiarism Auditor: Paste existing text to inspect potential similarity hotspots and generate Turnitin-compliant academic paraphrasing. Tab 5 — Statistics Engine: Enter classroom or experimental pretest/posttest scores to compute Paired Sample t-Tests, degrees of freedom, and Hake (1999) N-Gain categories, accompanied by formal Indonesian academic prose ready for Bab IV. Health Check Verification: curl -s https://academicai-production-a41d.up.railway.app/healthz { "status": "ok", "server": "AcademAI Universal Academic Engine", "discipline": "Universal Academic Research (Multi-Disciplinary)", "features": [ "memory_context", "pdf_parser", "plagiarism_checker", "google_scholar", "zotero_sync", "full_generator", "citation_validator", "academic_stats", "dataviz_mcp" ], "models": [ "gemini-3.1-flash-lite", "gemini-3.5-flash-lite", "gemini-3.7-flash", "gemini-3.8-flash" ] } / academic_ai ============================================ AcademAI v2.0 — README ============================================ 🎓 AcademAI — AI Scientific Article & Thesis Generator Frontier Edition (Powered by Google Gemini 2.0 & n8n Agent Harness) 🎃 Hacktoberfest 2026 Weekend Challenge Submission: Build for a Friend Dedicated to: Helping my wife complete her undergraduate thesis in Early Childhood Education (S1 PAUD). Frontier AI: Google Gemini 3.8 Flash · n8n Open Agent Harness · Google Scholar · Zotero · Gotenberg Version: 2.0 | License: MIT 📁 Struktur Project academ-ai/ ├── docker-compose.yml # Stack: n8n + Gotenberg ├── .env.example # Template environment variables ├── start.ps1 # Script setup & run otomatis ├── DEV_SUBMISSION.md # Naskah submission resmi DEV.to ├── academic.md # Dokumen spesifikasi arsitektur lengkap ├── n8n/ │ ├── academ_ai_workflow_gemma.json # 💎 Workflow Gemma 2 (Open-Weight / Ollama) │ ├── academ_ai_workflow_gemini.json # Workflow Google Gemini API │ └── academ_ai_workflow_v2.json # Workflow Claude Sonnet ├── frontend/ # Web … View on GitHub ┌─────────────────────────────────────────────────────────────────┐ │ Client Layer: Vanilla HTML5 · CSS3 · ES2022 (Zero Build) │ │ 5 Academic Tabs: Chat · Generator · Validator · Plag · Stats │ └────────────────────────────────┬────────────────────────────────┘ │ REST API ┌────────────────────────────────▼────────────────────────────────┐ │ OPEN-SOURCE AGENT HARNESS (Standalone Node.js / Express) │ │ │ │ ┌────────────────────────────────────────────────────────┐ │ │ │ DISCIPLINES Registry (12 Faculties, MIT License) │ │ │ │ Dynamic Personas · Theory Guides · Methodology Models │ │ │ └────────────────────────────┬───────────────────────────┘ │ │ │ │ │ ┌───────────────────────┼────────────────────────┐ │ │ ▼ ▼ ▼ │ │ ┌───────────────┐ ┌─────────────────┐ ┌───────────────┐ │ │ │ Google Gemini │ │ Google Scholar │ │ Zotero REST │ │ │ │ 4-Model Chain │ │ & SerpApi Index │ │ Auto-Sync │ │ │ └───────────────┘ └─────────────────┘ └───────────────┘ │ │ │ │ ┌─────────────────────────────┐ ┌───────────────────────────┐ │ │ │ Pure-JS Statistics Engine │ │ RegEx Citation Validator │ │ │ │ Paired t-Test · Hake N-Gain │ │ APA7 / CrossRef Checker │ │ │ └─────────────────────────────┘ └───────────────────────────┘ │ │ │ │ ┌─────────────────────────────┐ ┌───────────────────────────┐ │ │ │ Persistent Document Memory │ │ DOCX Academic Exporter │ │ │ │ Local JSON Session Storage │ │ Formatted Word Packaging │ │ │ └─────────────────────────────┘ └───────────────────────────┘ │ └─────────────────────────────────────────────────────────────────┘ Deployed via Dockerfile → Railway Cloud The foundational open-source component in server.js is the modular DISCIPLINES registry, which switches scientific personas, theories, and research methodologies dynamically: // server.js (MIT Licensed) const DISCIPLINES = { paud: { name: 'Pendidikan Anak Usia Dini (PAUD)', searchSuffix: 'pendidikan anak usia dini jurnal PAUD', persona: 'Co-Pilot Riset Skripsi S1 PAUD dan Ilmu Keguruan Anak Usia Dini Terkemuka di Indonesia', theoryGuide: Teori Pokok: Piaget (Pra-operasional 2–7 tahun), Vygotsky (ZPD & Scaffolding), Montessori (Prepared environment, media sensorik), Ki Hajar Dewantara (Sistem Among), STPPA (Permendikbudristek No. 5/2022: 6 Aspek Perkembangan)., methodGuide: Metodologi: PTK model Kemmis & McTaggart (Planning, Acting, Observing, Reflecting), Rubrik Standar: BB(1), MB(2), BSH(3), BSB(4). Target ketuntasan klasikal >= 75-80%. }, economics: { name: 'Ilmu Ekonomi & Bisnis', searchSuffix: 'jurnal ekonomi manajemen akuntansi bisnis', persona: 'Co-Pilot Riset Bidang Ilmu Ekonomi, Manajemen, dan Akuntansi', theoryGuide: Teori Pokok: Agency Theory (Jensen & Meckling), Porter's Five Forces, Kotler & Keller, Signaling Theory., methodGuide: Metodologi: Kuantitatif Asosiatif, Structural Equation Modeling (SEM/SmartPLS), Regresi Linear Berganda. }, law: { name: 'Ilmu Hukum', searchSuffix: 'jurnal ilmu hukum perundang-undangan jurisprudensi', persona: 'Co-Pilot Riset Bidang Ilmu Hukum dan Perundang-undangan', theoryGuide: Teori Pokok: Teori Keadilan (John Rawls), Teori Kepastian Hukum (Gustav Radbruch), Teori Kemanfaatan., methodGuide: Metodologi: Penelitian Yuridis Normatif (Statute, Conceptual, Case Approach) dan Yuridis Empiris. }, // + 9 other faculties: education, computer_science, engineering, health, psychology, agriculture, etc. }; function buildMasterAcademicPrompt(discipline, citationFormat, extraContext) { const disc = DISCIPLINES[discipline] || DISCIPLINES['general_academic']; return Anda adalah AcademAI — ${disc.persona}. DISIPLIN ILMU AKTIF: ${disc.name} ${disc.theoryGuide} ${disc.methodGuide} Format Sitasi Wajib: ${citationFormat} ${extraContext}; } // server.js — Zero external statistical dependencies app.post('/api/stats/calculate', (req, res) => { const { pretest, posttest, maxScore = 100, variableName = 'Variabel Penelitian' } = req.body; const n = pretest.length; const diffs = pretest.map((pre, i) => posttest[i] - pre); const meanDiff = diffs.reduce((a, b) => a + b, 0) / n; const variance = diffs.reduce((sum, d) => sum + Math.pow(d - meanDiff, 2), 0) / (n - 1); const standardError = Math.sqrt(variance / n); const tStat = standardError === 0 ? 0 : meanDiff / standardError; const df = n - 1; const preAvg = pretest.reduce((a, b) => a + b, 0) / n; const postAvg = posttest.reduce((a, b) => a + b, 0) / n; const nGain = (maxScore - preAvg === 0) ? 0 : (postAvg - preAvg) / (maxScore - preAvg); const nGainCategory = nGain >= 0.7 ? 'Tinggi' : nGain >= 0.3 ? 'Sedang' : 'Rendah'; res.json({ success: true, tStat: tStat.toFixed(4), df, nGain: nGain.toFixed(4), nGainPercent: (nGain * 100).toFixed(2) + '%', nGainCategory, preAvg: preAvg.toFixed(2), postAvg: postAvg.toFixed(2), interpretation: generateAcademicProse({ variableName, tStat, df, nGain, nGainCategory, preAvg, postAvg }) }); }); To build a reliable academic co-pilot for my wife without forcing her to use clunky software or pay steep monthly subscriptions, I engineered AcademAI as a lightweight, standalone open-source agent harness built with Node.js and Express (server.js), powered by Google Gemini (with a 4-tier model fallback chain). server.js) Instead of relying on heavy visual workflow engines that consume massive server memory, AcademAI runs as a single, ultra-fast Node.js service (MIT License): Modular DISCIPLINES Registry: An open-source routing engine containing 12 distinct academic faculties. When a student chooses a discipline, the harness dynamically injects the appropriate scientific theories and empirical methodologies into the prompt context. Dynamic Research Grounding: Before sending any request to the LLM, the harness queries Google Scholar via SerpApi for recent peer-reviewed literature. It injects these authentic findings directly into the prompt context so the AI never hallucinates citations. Automated Citation Validator: A custom RegEx engine that audits in-text citations (such as (Piaget, 1976) or (Sujiono, 2021)) and DOI links against Google Scholar results, labeling each reference as VALID, PARTIAL, or INVALID. Pure JavaScript Statistics Engine: Computes classroom action research (PTK) statistics — including Paired Sample t-Tests and Hake (1999) Normalized Gain (N-Gain) — using pure math with zero external libraries. It outputs ready-to-paste formal Indonesian academic prose for Chapter IV (Bab IV). Session Memory Persistence: Stores session context locally on disk (./data/memory.json) so students can iteratively build their thesis across multiple days without losing research state. Academic DOCX Exporter: Converts structured Markdown chapters into formatted Microsoft Word (.docx) files using the open-source docx npm library. We utilize the official Google Gemini API (an official Hacktoberfest challenge partner) for its state-of-the-art academic prose and generous free tier: gemini-3.1-flash-lite: Ultra-low latency for quick inquiries and definitions. gemini-3.5-flash-lite: Efficient reasoning for literature synthesis. gemini-3.7-flash: Balanced speed and academic writing depth. gemini-3.8-flash: Highest quality for comprehensive 5-chapter thesis generation. If one model encounters a rate limit or server load, the harness automatically cascades to the next tier without interrupting the student's writing flow. There are over 8 million university students across Indonesia. Most students in regional universities study on 5-to-8-year-old budget laptops or smartphones. Closed, proprietary academic AI tools charge $20 to $30 per month — completely prohibitive for students in developing nations. At the same time, forcing students to run massive multi-gigabyte models locally would exclude my wife and millions like her whose hardware cannot run heavy models without overheating. Open innovation bridged this divide: by combining a clean, open-source MIT-licensed agent harness, open academic search APIs (Google Scholar / SerpApi), and Gemini's generous free-tier API, we delivered enterprise-grade research tooling at zero cost to the end user. Because AcademAI's agent harness, prompt routing, and validation logic are 100% open-source in server.js, the project is completely decoupled from any single LLM provider: Anyone can clone the repository, install dependencies via npm install, and start it with npm start. The open architecture allows developers to easily swap or extend the LLM backend to any provider with minimal code adjustments. In academic research, black-box AI is dangerous because it fabricates truth. Open innovation allowed us to write transparent algorithms for citation validation and statistical analysis. Students and academic advisors can inspect every formula, prompt rule, and validation check directly on GitHub. AcademAI was engineered with the assistance of Antigravity IDE (Google DeepMind's agentic pair-programming system). Throughout the session, the agent and I: Consolidated the application into a single, clean, standalone server.js agent harness to maximize performance and simplify deployment. Refactored the core logic into a universal multi-disciplinary academic engine supporting 12 faculties. Built and tested the zero-dependency statistical calculation engine for Classroom Action Research (PTK) and Hake's N-Gain formulas. Implemented real-time citation extraction and verification algorithms. Packaged the application with Docker and verified the live cloud deployment on Railway with automated /healthz monitoring. Google Gemini: AcademAI is powered by the Google Gemini API, utilizing a 4-tier model fallback chain (gemini-3.1-flash-lite, gemini-3.5-flash-lite, gemini-3.7-flash, and gemini-3.8-flash) combined with an open-source prompt routing engine tailored for academic research. Best Open Source Tool: The entire agent harness (server.js), 12-discipline prompt registry, citation verification engine, statistics calculator, and vanilla web frontend are 100% MIT-licensed, completely standalone, and freely available for students and educators worldwide.
Key Takeaways #
- •This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend I built AcademAI for someone very close to my heart: my wife
- •This story was reported by Dev.to , covering developments in thedev space.
- •AI advancements continue to reshape industries — read the full article on Dev.to for complete coverage.
📖 Continue reading the full article: