Perplexideez offers a self-hosted, AI-powered search experience designed for users who want to integrate web search with their own applications and require multi-user support. It aims to provide a Perplexity-like interface with enhanced control over data sources and user access.

How It Works

The system leverages a PostgreSQL database for data storage and supports either Ollama or OpenAI-compatible endpoints for AI model interactions. Web searches are facilitated through a SearXNG instance. Its architecture emphasizes statelessness for containerized deployments, enabling easier scaling and rolling updates in environments like Kubernetes.

Quick Start & Requirements

• Installation: Clone the repository, install dependencies with pnpm install, create a .env file from .env.example, and run with pnpm dev:up.
• Prerequisites: Requires a PostgreSQL database, an OpenAI API token or OpenAI-compatible endpoint, or an Ollama instance. SearXNG is also required.
• Resources: Development setup includes Postgres and SearXNG. Full setup requires configuring AI endpoints and SearXNG.
• Links: Roadmap, Contributing

Highlighted Details

• Supports multi-user accounts with OpenID Connect (OIDC) for Single Sign-On (SSO).
• Allows sharing of search results with configurable access control and embeddable links.
• Provides source annotations for AI-generated conclusions, enabling users to verify information.
• Offers customization for different AI models for various tasks (speed, balanced, quality, embeddings).

Maintenance & Community

The project is actively developed by a single maintainer. Community interaction and contributions are encouraged via GitHub issues and pull requests.

Licensing & Compatibility

Distributed under the AGPL License. This license may have implications for commercial use or linking with closed-source software due to its strong copyleft provisions.

Limitations & Caveats

SearXNG instances may be rate-limited by Perplexideez requests. The project is still working on full statelessness and blocking exits during generation. Kubernetes deployment currently relies on example manifests, with a Helm chart still in development.