AI Perplexica - An AI-powered search engine

[Ethanator]

Overview and Core Functionality

Perplexica is an open-source, AI-powered search engine designed to provide in-depth answers to user queries by understanding the questions and leveraging advanced machine learning techniques. It is inspired by Perplexity AI and aims to deliver relevant, clear, and sourced information while prioritizing user privacy.As of May 14, 2025, Perplexica offers a range of features and capabilities, positioning itself as a comprehensive search tool.

Perplexica is equipped with a variety of features to enhance the search experience:

Key Features

Perplexica distinguishes itself by not just performing web searches but by comprehending user queries to deliver precise answers. It utilizes sophisticated machine learning algorithms, including similarity searching and embeddings, to refine search results and present them with cited sources.A key component of its architecture is SearxNG, an open-source metasearch engine. The use of SearxNG allows Perplexica to fetch the most current information available on the internet, ensuring that users receive up-to-date results without the system needing constant, large-scale data updates or indexing common in some other search tools. This approach also underpins Perplexica's commitment to privacy, as SearxNG, combined with Perplexica's open-source nature, means user search history is not tracked. The project was built with LangChain, highlighting its use of modern AI development frameworks. For those interested in the technical specifics, Perplexica provides documentation detailing its architecture and operational mechanics.

Installation

• Local LLM Integration: Users can integrate and utilize local Large Language Models (LLMs) such as Llama3 and Mixtral. This is facilitated through Ollama, allowing for more personalized or specialized search processing.
• Two Main Search Modes:
  • Normal Mode: This mode processes user queries and conducts a standard web search to retrieve information.
  • Copilot Mode: Currently in development, this mode aims to enhance search effectiveness. It works by generating multiple, varied queries based on the initial input to find a broader range of relevant internet sources. Unlike the Normal Mode, which primarily relies on the context provided by SearxNG, Copilot Mode actively visits the top search matches to extract pertinent information directly from the web pages
• Focus Modes: Perplexica offers specialized search modes tailored to specific types of queries. As of now, there are six focus modes:
  • All Mode: Conducts a comprehensive search across the entire web.
  • Writing Assistant Mode: Designed to assist with writing tasks that do not require real-time web searching.
  • Academic Search Mode: Specifically targets academic articles and research papers, ideal for scholarly research.
  • YouTube Search Mode: Finds videos on YouTube relevant to the search query.
  • Wolfram Alpha Search Mode: Leverages Wolfram Alpha to answer queries requiring calculations, data analysis, or factual data.
  • Reddit Search Mode: Searches for discussions, opinions, and community feedback on Reddit related to the query.
• Current Information: By using SearxNG, Perplexica ensures that the information provided is current. It achieves this by fetching and re-ranking results from various sources in real-time, thus avoiding the issue of outdated information that can occur with search engines relying on periodic web crawls and static indexes.
• API Access: Perplexica provides an Application Programming Interface (API), enabling developers to integrate its powerful search capabilities into their own applications. The API allows for running searches, utilizing different models, and retrieving answers programmatically.
• Image and Video Search: The platform includes capabilities for image and video search, among other features.
• Advanced Searching Capabilities: It employs sophisticated AI techniques such as similarity search and embeddings to enhance the accuracy and relevance of search results.
• Privacy-Focused: Being open-source and not tracking user search history are core tenets of Perplexica, ensuring user privacy.

Perplexica offers two primary methods for installation, with Docker being the highly recommended approach due to its simplified setup process for managing dependencies and environment variables.

• Installation with Docker:
  1. Prerequisites: Ensure Docker is installed and running on your system.
  2. Clone Repository: Clone the Perplexica GitHub repository using the command: git clone https://github.com/ItzCrazyKns/Perplexica.git.
  3. Navigate to Directory: Change to the newly cloned project directory.
  4. Configuration File: Rename the sample.config.toml file to config.toml.
  5. API Keys (Optional but Recommended): Populate the config.toml file with necessary API keys if you intend to use proprietary models. These can also be configured later via the settings dialog in the application:
     • OPENAI: Your OpenAI API key for using OpenAI models.
     • OLLAMA: Your Ollama API URL, typically http://host.docker.internalORT_NUMBER (e.g., http://host.docker.internal:11434 if Ollama is on port 11434). This is for using local Ollama models.
     • GROQ: Your Groq API key for using Groq's hosted models.
     • ANTHROPIC: Your Anthropic API key for using Anthropic models.
  6. Similarity Measure: The SIMILARITY_MEASURE field is pre-filled by default and can be left as is if unsure.
  7. Start Perplexica: Ensure you are in the directory containing the docker-compose.yaml file and execute docker compose up -d.
  8. Access: After a few minutes for setup, Perplexica will be accessible at http://localhost:3000 in a web browser. Once containers are built, Perplexica can be started directly from Docker without needing a terminal.
• Installation Without Docker:
  1. Install SearXNG: Install SearXNG and ensure that JSON format is enabled in its settings.
  2. Clone and Configure: Clone the Perplexica repository, navigate to the root directory, rename sample.config.toml to config.toml, and fill in all required fields in this configuration file.
  3. Install Dependencies: Run npm i to install project dependencies.
  4. Build Project: Execute npm run build to build the application.
  5. Start Application: Finally, run npm run start to launch Perplexica.[1]Note: The documentation emphasizes that using Docker is recommended as it simplifies the setup, especially concerning environment variables and dependencies. Further details on installation, including updates, can be found in the official documentation.

Usage

• Using Perplexica as a Search Engine in Your Browser:To use Perplexica like traditional search engines directly from your browser's search bar:
  1. Open your browser's settings.
  2. Navigate to the 'Search Engines' or 'Site Search' section.
  3. Add a new site search entry. Use the URL http://localhost:3000/?q=%s. If Perplexica is hosted on a different address or port, replace localhost and 3000 accordingly.
• Using Perplexica's API:Developers can integrate Perplexica's search functionalities into their applications using its API. This allows for programmatic searching, model selection, and answer retrieval. Detailed API documentation is available through a link on the Perplexica GitHub page.
• Exposing Perplexica to a Network:Perplexica runs on Next.js and handles all API requests. It is designed to work immediately on the same network where it's hosted and can remain accessible via port forwarding if needed.

If you encounter issues connecting to Ollama's API, it usually means the Perplexica backend cannot reach the Ollama service. Here are some troubleshooting steps:

Troubleshooting Ollama Connection Errors

Upcoming Features

• Check Ollama API URL: Verify that the API URL is correctly set in Perplexica's settings menu.
• Update API URL Based on Operating System:
  • Windows: Use http://host.docker.internal:11434.
  • Mac: Use http://host.docker.internal:11434.
  • Linux: Use http://<YOUR_INTERNAL_IP>:11434 (replace <YOUR_INTERNAL_IP> with your machine's internal IP address accessible by the Docker container). Adjust the port number if Ollama is running on a port other than the default 11434.
• For Linux Users - Expose Ollama to Network: If Ollama is running directly on the Linux host, you may need to configure it to accept connections from other network interfaces (like the Docker bridge). To do this, edit the Ollama service file at /etc/systemd/system/ollama.service and add the line Environment="OLLAMA_HOST=0.0.0.0". After saving the change, restart the Ollama service using systemctl restart ollama. More details can be found in the Ollama documentation
• Firewall: Ensure that the port Ollama uses (default 11434) is not blocked by your firewall.

Perplexica's development roadmap includes several enhancements and new functionalities. According to the provided source, planned features are:

Perplexica is an open-source project that welcomes community support and contributions:

Support and Contribution

• Add settings page (though some settings for API keys are already available).
• Adding support for local LLMs (this feature is also listed as currently available).
• History Saving features.
• Introducing various Focus Modes (this feature with six modes is also listed as currently available).
• Adding API support (this feature is also listed as currently available).
• Adding Discover (a new feature for content discovery).
• Finalizing Copilot Mode (currently listed as in development).
• Support Us: Users who find Perplexica useful can support the project by giving it a star on its GitHub repository: https://github.com/ItzCrazyKns/Perplexica. This helps increase visibility and supports development. Donations are also accepted through various options listed on their GitHub page to help sustain the project.
• Contribution: The project encourages users to report bugs or share ideas for new features via GitHub Issues. Perplexica is built on the principle of making AI and LLMs easily accessible. Detailed information on how to contribute can be found in the CONTRIBUTING.md file on their GitHub repository.

For questions, feedback, or assistance, users can:

Community, Help, and Support

• Create an issue on the Perplexica GitHub page.
• Join the official Perplexica Discord server. A link to join is provided on the GitHub page. The Discord server is a place to connect with other users, share experiences and reviews, and receive more personalized help.
• For discussions outside of regular support channels, users can contact the developer itzcrazykns on Discord.

Perplexica aims to continuously improve and expand its capabilities, valuing user feedback and contributions to make the AI-powered search experience even better.

---

Authoritative Sources

1. Overview, Features, Installation, Using as a Search Engine, Using Perplexica's API, Expose Perplexica to a network, One-Click Deployment, Upcoming Features, Support Us, Contribution, Help and Support. [https://github.com/ItzCrazyKns/Perplexica]↩
2. Perplexica is an AI-powered open source search engine. [https://www.skool.com/data-alchemy/perplexica-is-an-ai-powered-open-source-search-engine]↩
3. Perplexica - An AI-powered search engine. [https://www.linkedin.com/posts/lang...arch-engine-activity-7190144088405340162-G7tH]↩
4. Perplexica AI: The Open-Source, Privacy-Focused Search Engine. [https://www.toolify.ai/ai-news/perplexica-ai-the-open-source-privacyfocused-search-engine-3318440]↩
5. Perplexica. [https://repocloud.io/details/?app_id=267]↩

Answer Provided by www.iAsk.ai – Ask AI.