{"id":26356103,"url":"https://github.com/webessentz/parrot","last_synced_at":"2025-07-30T09:34:58.108Z","repository":{"id":280473251,"uuid":"882340541","full_name":"WebEssentz/Parrot","owner":"WebEssentz","description":null,"archived":false,"fork":false,"pushed_at":"2025-03-08T16:30:21.000Z","size":408,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-03-08T17:20:07.555Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"https://orbe-ai.vercel.app","language":"TypeScript","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/WebEssentz.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}},"created_at":"2024-11-02T15:10:42.000Z","updated_at":"2025-03-08T16:30:24.000Z","dependencies_parsed_at":"2025-03-08T17:20:14.546Z","dependency_job_id":null,"html_url":"https://github.com/WebEssentz/Parrot","commit_stats":null,"previous_names":["webessentz/orbe-ai","webessentz/parrot"],"tags_count":3,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/WebEssentz%2FParrot","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/WebEssentz%2FParrot/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/WebEssentz%2FParrot/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/WebEssentz%2FParrot/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/WebEssentz","download_url":"https://codeload.github.com/WebEssentz/Parrot/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":243871862,"owners_count":20361380,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","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":[],"created_at":"2025-03-16T13:19:17.718Z","updated_at":"2025-07-30T09:34:58.092Z","avatar_url":"https://github.com/WebEssentz.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003ch1 align=\"center\"\u003eAvurna AI Agent\u003c/h1\u003e\n\n\u003cp align=\"center\"\u003e\n  Avurna is an open-source, AI-powered web agent built with Next.js and the Vercel AI SDK. It goes beyond a simple chatbot, using a multi-tool, recursive architecture to understand user goals, navigate the web, and synthesize information for you.\n\u003c/p\u003e\n\n---\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"#-latest-major-update-exa-pro-deep-research-2025-06-17\"\u003e\u003cstrong\u003eLatest Update\u003c/strong\u003e\u003c/a\u003e ·\n  \u003ca href=\"#features\"\u003e\u003cstrong\u003eFeatures\u003c/strong\u003e\u003c/a\u003e ·\n  \u003ca href=\"#deploy-your-own\"\u003e\u003cstrong\u003eDeploy\u003c/strong\u003e\u003c/a\u003e ·\n  \u003ca href=\"#running-locally\"\u003e\u003cstrong\u003eRun Locally\u003c/strong\u003e\u003c/a\u003e ·\n  \u003ca href=\"#authors\"\u003e\u003cstrong\u003eAuthors\u003c/strong\u003e\u003c/a\u003e\n\u003c/p\u003e\n\u003cbr/\u003e\n\n## 🚀 Latest Major Update: v2.0.0 - Enhanced UI/UX \u0026 Input Experience (2025-07-21)\n\nThis release significantly refines the user interface and interaction, especially around the chat input.\n\n-   **Immediate Input Clearing:** The chat input textarea now clears instantly upon message submission, providing immediate visual feedback and a smoother interaction flow.\n-   **Uninterrupted Typing during AI Response:** The textarea remains fully enabled for user input even while the AI is streaming its response. This allows for continuous conversation and follow-up questions without interruption.\n-   **Improved Textarea Visuals \u0026 Expansion:** Resolved a UI bug where typed text was obscured by the bottom button bar. The textarea now smoothly expands and maintains correct padding, ensuring all text is visible and the input area dynamically adjusts to content.\n\n---\n\n## 🚀 2025-06-17: Exa Pro Deep Research Integration\n\n-   **Exa Pro Deep Research:** Avurna now uses Exa Pro for web research, providing research-grade, multi-source, reasoned answers with direct citations. This is a leap beyond traditional search—it's automated deep research.\n-   **Record-Breaking Speed:** Achieved a new record: 9.8 seconds for a full deep research answer (measured end-to-end, streaming enabled). No URL fails in testing.\n-   **Streaming Results:** Answers stream in real time for the fastest possible experience.\n-   **More Grounded Than Google or ChatGPT:** Every claim is directly cited, with no hallucinated URLs. Exa Pro is more grounded and up-to-date than Google or ChatGPT search.\n-   **Note:** The standard Exa model is even faster than Exa Pro for pure search.\n\n---\n\n## 🚀 2025-06-12: Major Architectural Upgrade: The Semantic Web Agent\n\nThis release marks a fundamental shift from a chatbot to a true AI agent.\n\n-   **Semantic Navigation:** Avurna no longer just follows links; it uses a fast AI model to understand the *meaning* of links before navigating. This allows it to make human-like decisions, avoiding irrelevant rabbit holes and finding information more accurately.\n-   **Multi-Modal Understanding:** Avurna can see images and watch videos on webpages using `gemma-3-27b-it` to analyze content, not just text.\n-   **Recursive Web Exploration:** When information isn't on the first page, Avurna intelligently explores deeper, following the most promising paths to fulfill your request. This is all done with strict safety guards (domain-locking, page limits, timeouts).\n-   **Task-Specific Processing:** Understands if you want a picture, a video, or a summary, and adapts its process to be as efficient as possible.\n-   **Resilient \u0026 Transparent:** If one method fails (like direct page fetching), Avurna can pivot to another (like web search) and will explain its reasoning process to the user.\n\n## Features\n\n-   **Agentic Web Navigation:** A smart agent that can browse websites, understand content, and find answers for you.\n-   **Multi-Modal Analysis:** Processes text, images, and video to get the full picture.\n-   **Summarization Engine:** Can read long articles and provide concise, accurate summaries.\n-   **Tool-Using Framework:** Seamlessly switches between browsing, searching, and analyzing to get the job done.\n-   **Strategic Dual-Model AI Core:**\n    -   Uses **`gemma-3n-e4b-it`** for lightning-fast tasks (like understanding your initial request).\n    -   Uses **`gemma-3-27b-it`** for powerful, heavy-lifting tasks (like vision analysis and summarization).\n-   **Modern, Responsive UI:** Built with [shadcn/ui](https://ui.shadcn.com/), [Tailwind CSS](https://tailwindcss.com), and the [Next.js](https://nextjs.org) App Router.\n-   **Streaming Responses:** Text and results stream in real-time, powered by the [Vercel AI SDK](https://sdk.vercel.ai/docs).\n-   **Enhanced Chat Input Experience:** Immediate clearing, uninterrupted typing during AI responses, and smooth textarea expansion.\n\n\n## 🚀 2025-05-20: Features \u0026 Improvements\n\n-   **Mobile Safearea for Privacy Message:**\n    -   On mobile, the privacy policy message, textarea, and disclaimer are now wrapped in a \"safearea\". Tapping outside this area dismisses the privacy message, while taps inside do not.\n    -   This improves usability and prevents accidental dismissal when interacting with the input or privacy message.\n\n## 🚀 Major 2025 Updates (Prior to v2.0.0)\n\n-   **One-Shot Search Mode:** When the Search button is selected, the next message always uses real-time web search, then reverts to the default model. No more infinite search loops.\n-   **Privacy \u0026 Transparency:** Avurna never mentions its training data, internal tools, or creator details unless asked. Capabilities are described in plain language (e.g., \"I can search Google\").\n-   **Modern UI:** The header now displays \"Avurna\" in a semibold, modern font on desktop, and the original icon on mobile/tablet. The search button's lit state is user-controlled and visually consistent. The textarea input is cleared after sending a message.\n-   **Robust Model Handling:** Improved code structure for model selection and POST handling to avoid race conditions and ensure correct model usage.\n-   **Responsive Design:** Enhanced experience for both desktop and mobile users.\n\nSee [CHANGELOG.md](./CHANGELOG.md) for details.\n\n---\n\n## ⚡️ 2025-06-17: Exa Pro Deep Research Integration\n\n-   Exa Pro is now the default for web research, replacing Google Search.\n-   Achieved a record-breaking 9.8 seconds for full deep research answers (streaming, no URL fails).\n-   Streaming mode returns partial results in real time.\n-   Every claim is directly cited, with no hallucinated URLs.\n-   Exa Pro is more grounded and up-to-date than Google or ChatGPT search.\n-   According to Exa’s developer, the standard Exa model is even faster than Exa Pro for pure search.\n\n---\n\n## Deploy Your Own\n\nYou can deploy your own version to Vercel by clicking the button below:\n\n[![Deploy with Vercel](https://vercel.com/button)](https://vercel.com/new/clone?project-name=Vercel+x+Groq+Chatbot\u0026repository-name=ai-sdk-starter-groq\u0026repository-url=https%3A%2F%2Fgithub.com%2Fvercel-labs%2Fai-sdk-starter-groq\u0026demo-title=Vercel+x+Groq+Chatbot\u0026demo-url=https%3A%2F%2Fai-sdk-starter-groq.labs.vercel.dev%2F\u0026demo-description=A+simple+chatbot+application+built+with+Next.js+that+uses+Groq+via+the+AI+SDK+and+the+Vercel+Marketplace\u0026products=%5B%7B%22type%22%3A%22integration%22%2C%22protocol%22%3A%22ai%22%2C%22productSlug%22%3A%22api-key%22%2C%22integrationSlug%22%3A%22groq%22%7D%5D)\n\n## Running Locally\n\n1. Clone the repository and install dependencies:\n\n    ```bash\n    npm install\n    # or\n    yarn install\n    # or\n    pnpm install\n    ```\n\n2. Install the [Vercel CLI](https://vercel.com/docs/cli):\n\n    ```bash\n    npm i -g vercel\n    # or\n    yarn global add vercel\n    # or\n    pnpm install -g vercel\n    ```\n\n    Once installed, link your local project to your Vercel project:\n\n    ```bash\n    vercel link\n    ```\n\n    After linking, pull your environment variables:\n\n    ```bash\n    vercel env pull\n    ```\n\n    This will create a `.env.local` file with all the necessary environment variables.\n\n3. Run the development server:\n\n    ```bash\n    npm run dev\n    # or\n    yarn dev\n    # or\n    pnpm dev\n    ```\n\n4. Open [http://localhost:3000](http://localhost:3000) to view your new AI chatbot application.\n\n## Authors\n\nAvurna AI is maintained by:\n\n- Godwin (Lead Developer)\n- Charles (UI/UX \u0026 Training)\n- The open-source community\n\n\nContributions are welcome! Feel free to open issues or submit pull requests to enhance functionality or fix bugs.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fwebessentz%2Fparrot","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fwebessentz%2Fparrot","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fwebessentz%2Fparrot/lists"}