{"id":30432255,"url":"https://github.com/vero-code/castanea","last_synced_at":"2026-04-10T11:32:31.438Z","repository":{"id":309093385,"uuid":"1034992212","full_name":"vero-code/castanea","owner":"vero-code","description":"Multi-agent AI research assistant for students. Powered by Google ADK, Perplexity, and Gemini models.","archived":false,"fork":false,"pushed_at":"2025-08-09T20:42:36.000Z","size":93,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"master","last_synced_at":"2025-08-22T21:27:26.560Z","etag":null,"topics":["agents","ai","analysis","automation","edtech","fastapi","gemini","google-adk","hackathon","perplexity","rag","react","research","vite","writing"],"latest_commit_sha":null,"homepage":"https://devpost.com/software/castanea","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/vero-code.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-08-09T12:28:06.000Z","updated_at":"2025-08-09T20:42:39.000Z","dependencies_parsed_at":"2025-08-09T20:43:00.187Z","dependency_job_id":"cc122685-bc70-44fa-81f7-c1826437ede7","html_url":"https://github.com/vero-code/castanea","commit_stats":null,"previous_names":["vero-code/castanea"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/vero-code/castanea","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vero-code%2Fcastanea","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vero-code%2Fcastanea/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vero-code%2Fcastanea/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vero-code%2Fcastanea/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/vero-code","download_url":"https://codeload.github.com/vero-code/castanea/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vero-code%2Fcastanea/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31641114,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-10T07:40:12.752Z","status":"ssl_error","status_checked_at":"2026-04-10T07:40:11.664Z","response_time":98,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agents","ai","analysis","automation","edtech","fastapi","gemini","google-adk","hackathon","perplexity","rag","react","research","vite","writing"],"created_at":"2025-08-22T20:43:02.903Z","updated_at":"2026-04-10T11:32:31.433Z","avatar_url":"https://github.com/vero-code.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Castanea\n\n![React](https://img.shields.io/badge/React-61DAFB?style=for-the-badge\u0026logo=react\u0026logoColor=black)\n![Python](https://img.shields.io/badge/Python-3776AB?style=for-the-badge\u0026logo=python\u0026logoColor=white)\n![Google ADK](https://img.shields.io/badge/Google%20ADK-4285F4?style=for-the-badge\u0026logo=google\u0026logoColor=white)\n![Perplexity API](https://img.shields.io/badge/Perplexity-000000?style=for-the-badge\u0026logo=perplexity\u0026logoColor=white)\n![Winner](https://img.shields.io/badge/Winner-4th_Place-FFD700?style=for-the-badge\u0026logo=PANDA\u0026logoColor=black)\n\n\u003e **Three smart AI agents that find, analyze, and create — so you can focus on what matters.**\n\nCastanea is an AI-powered multi-agent platform for students and researchers. It automates time-consuming academic tasks by combining **real-time search**, **text analysis**, and **content generation** in one place. \n\nBuilt with the power of **Google ADK** and **Perplexity API**, it features three specialized AI agents coordinated to deliver accurate, well-structured, and useful results. \n\n🏆 **Winner: Certificate of Excellence (Fourth Place)** at [PANDA Hacks 2025 🐼🚀](https://devpost.com/software/castanea)\n### 📺 Demo Video\n[![Watch the Demo Video](https://img.youtube.com/vi/rnamBJCO2cs/maxresdefault.jpg)](https://youtu.be/rnamBJCO2cs)\n\n*Built in just 8 hours.* 🐼\n*Note: The second half of the video perfectly demonstrates the interaction between agents in ADK Web.*\n\n## Features\n\nCastanea operates through a team of specialized AI agents, each designed for a specific academic task:\n\n-   ### 🧠 ResearcherAgent\n    \n    Performs in-depth research on complex topics. It leverages the **Perplexity** search engine to gather real-time, factual data and then uses the power of **Gemini Pro** to analyze, synthesize, and structure this information into a comprehensive and coherent report. This Retrieval-Augmented Generation (RAG) approach ensures answers are both accurate and well-written.\n    \n-   ### ⚡ AnalystAgent\n    \n    Provides quick and efficient analysis of existing text. Built on the **Gemini Flash** model for maximum speed, this agent can instantly summarize articles, extract key points from lecture notes, or identify the main arguments in a difficult text, helping students grasp core concepts faster.\n    \n-   ### ✍️ WriterAgent\n    \n    Acts as a versatile writing assistant. It can generate original content like essays, reports, and formal emails based on a user's prompt. Additionally, it is equipped with a tool to save its output directly to a file (e.g., social_media_essay.txt), making it easy to store and edit the work.\n\n## 🏗️ Architecture\n\nCurious about how the agents work together? Check out our [Architecture Documentation](ARCHITECTURE.md) to see the system design and interaction flow.\n\n## Technologies\n\n- Vite, React, JavaScript\n- Python\n- Agent Development Kit (ADK)\n- Gemini Pro/Flash\n- Perplexity’s Sonar API\n\n## Installation\n\n1. Clone the repository:\n\n\t```bash\n\tgit clone https://github.com/vero-code/castanea.git\n\tcd castanea\n\t```\n2. Configure the `.env` file:\n\n\t- Create `.env` in the root of the project.\n\t- See an example in `.env.example`.\n\n3. Run the services:\n\n\t```bash\n\t# Frontend\n\tcd frontend\n\tnpm install\n\tnpm run dev\n\t```\n \n\t```bash\n\t# Backend\n\tcd backend\n\tpython -m venv .venv\n\t.venv\\Scripts\\Activate.ps1\n\tpip install -r requirements.txt\n\t```\n\n## Testing\n\n### Test for ResearcherAgent  \n\nThe CRISPR and GMO problem is a complex, topical issue that requires gathering information from different fields (ethics, biology, law).\nIt is an ideal stress test for perplexity_search and the agent's ability to synthesize a complex answer.\n\n```\nWrite a detailed report comparing the ethical implications of CRISPR gene editing with traditional GMO technology.\n```\n\n### Test for AnalystAgent\n\nAnalyzing a scientific abstract is something that students encounter all the time. \nThe task checks whether the agent (Flash model) can quickly and accurately extract the essence from a scientific, \"dry\" text.\n\n```\nSummarize the key finding of this scientific abstract in a single, clear sentence:\n\nThis study investigates the correlation between bilingualism and delayed onset of dementia. \nA longitudinal study was conducted over 10 years with a cohort of 500 monolingual and 500 bilingual participants. \nCognitive decline was measured using the Mini-Mental State Examination (MMSE) annually. \nResults indicated that the bilingual cohort exhibited first symptoms of dementia on average 4.5 years later \nthan the monolingual cohort, suggesting that the constant cognitive effort of managing two languages \nbuilds cognitive reserve, thereby providing a protective effect against neurodegeneration.\n```\n\n### Test for WriterAgent  \n\nWriting an argumentative essay is not just text generation, but a demonstration of the ability to build a logical structure (thesis, arguments, conclusion). \nUsing the save_report tool here is absolutely logical - the student will want to save his work.\n\n```\nWrite a 5-paragraph argumentative essay on whether social media has a net positive or negative impact on teenage mental health.\nSave the output to 'social_media_essay.txt'.\n```\n\n## 📜 License\n\nThis project is licensed under the [MIT License](LICENSE).","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvero-code%2Fcastanea","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fvero-code%2Fcastanea","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvero-code%2Fcastanea/lists"}