Vane — Privacy-Focused AI Answer Engine
Self-hosted AI search and Q&A engine with multi-LLM support and cited answers.
- Source repo
- ItzCrazyKns/Vane
- Stars
- ★ 37k
- Last updated
- 1mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 42/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Privacy-conscious users: use local LLMs and SearxNG to avoid sending data to cloud services.
- Not a fit if
- Users who want a zero-config experience without SearxNG or local model setup
- Teams without Docker or Node.js who don't want to self-host
- Anyone planning to expose the instance publicly, since there is no authentication
- Source review
- 42/100 · Major gaps 2 safety controls not found
What does this agent do, and when should you use it?
Vane is a fully self-hostable AI answering engine designed with privacy at its core, handling all search processing on your own hardware. It combines internet knowledge with support for local LLMs (via Ollama) and cloud providers (OpenAI, Anthropic Claude, Google Gemini, Groq) to deliver accurate, cited answers while keeping searches private. It includes a built-in SearxNG aggregator for private web search, multiple search modes, source selection (web, discussions, academic), widgets (weather, calculations), image/video search, file uploads, and domain-specific search. All search history is stored locally. Installation is Docker-first (one command) with a manual fallback. Vane also offers an API for developers.
Accepts user queries, dispatches based on chosen mode (Speed, Balanced, Quality), gathers results from SearxNG (requires JSON format and Wolfram Alpha enabled), and generates cited answers via configured LLM providers (Ollama, OpenAI, Claude, Gemini, Groq). Supports file uploads (PDF, text, images) for document Q&A. Features domain-limited search, smart suggestions, Discover page, and local search history. Deployment via Docker (docker run ... itzcrazykns1337/vane:latest) or manual (npm i && npm run build && npm run start), serving on port 3000 with an HTTP API.
- Privacy-conscious users: use local LLMs and SearxNG to avoid sending data to cloud services.
- Researchers: leverage academic sources and Quality mode for well-cited answers.
- Technical documentation: restrict search to specific domains for quick API lookup.
- Offline environments: run without internet using local models and SearxNG.
- Productivity: set Vane as default search engine in browser.
- Developers: integrate Vane's API into custom applications.
How do you install or deploy this agent?
Quick start with Docker: run docker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latest and open http://localhost:3000 to configure API keys and models. For external SearxNG, use the slim image and set SEARXNG_API_URL environment variable. Manual install: clone repo, set up SearxNG (JSON format enabled, Wolfram Alpha enabled), then npm i, npm run build, npm run start, and access localhost:3000.
How do you use this agent?
After installation, open http://localhost:3000, enter API keys (for Ollama, put any value in the key field) and select models. Type a query, choose mode (Speed/Balanced/Quality) and source (web/discussions/academic). Use file upload for document Q&A and domain filter for site-specific searches. To use as a search engine, add http://localhost:3000/?q=%s in browser settings. For API integration, refer to docs/API/SEARCH.md.
What are this agent's strengths and limitations?
- Completely self-hosted, privacy is under your control, all data stays local
- Supports multiple LLM providers, including local Ollama and major cloud APIs, for flexible combinations
- Bundled SearxNG aggregator obfuscates search queries
- Offers Speed/Balanced/Quality modes and source selection
- Supports file uploads (PDF, text, images) for document Q&A
- Active community and Discord support
- Requires Docker or Node.js environment, technical setup barrier
- Common configuration issues with Ollama or SearxNG connections
- Local models demand significant hardware resources
- Depends on SearxNG instance; JSON interface must be enabled
- Some features (e.g., Tavily/Exa) are incomplete and on roadmap
- No authentication mechanism, exposing publicly is risky
How does this agent compare with similar options?
Vane is functionally similar to Perplexica (both are AI search tools), but Vane emphasizes privacy and self-hosting, whereas Perplexica may focus more on immediate usability. Compared to traditional search engines, Vane provides AI-generated answers with citations. Specific benchmark comparisons are not provided.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Vane — Privacy-Focused AI Answer Engine This agent | 42 · Major gaps | Self-hosted serviceFree + model costs | ★ 37k | 1mo ago | TypeScript | — |
| AgenticSeek | 40 · Major gaps | Self-hosted serviceFree + model costs | ★ 28k | 3d ago | Python | OpenAI API |
| Sidekick | 39 · Major gaps | Desktop appFree + model costs | ★ 3.3k | 19d ago | Swift | OpenAI API |
| VectorVein | 49 · Major gaps | Desktop appFree + model costs | ★ 940 | 14d ago | Vue | OpenAI API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README emphasizes privacy and local data storage, but lacks specific mechanisms for least privilege, user confirmation, data flow transparency. Many dependencies but no security audit or vulnerability scanning evidence. External effects (web search) exist but no user confirmation or revocation mechanism. Rollback not mentioned. Source attribution mentioned but not detailed. Deductions: lack of implementation details and evidence.
Evidence shows: README and package.json are consistent, feature descriptions match dependencies. Many dependencies but no availability guarantee or mirror. Failure messages partially present (e.g., Ollama connection errors) but not comprehensive. Deductions: dependency availability unverified, failure messages incomplete.
Evidence shows: README describes multiple use cases (local, cloud, search modes), capability boundaries described (multiple providers and sources), but trigger precision (e.g., search mode selection) not detailed. Environment fit has Docker and non-Docker installation instructions. Deductions: trigger precision and boundary descriptions not precise.
Evidence shows: README structure clear, installation notes detailed, examples provided (e.g., Docker commands), but missing known limitations and version changelog. License is MIT, but maintenance responsibility not clear. Deductions: missing known limitations, version changelog, and maintenance responsibility.
Evidence shows: Output usability (cited sources, multiple search modes) described, marginal value (privacy, local running) evident, but cost-benefit (resource consumption, API costs) not discussed. Deductions: cost-benefit not addressed.
Evidence shows: README claims features but no testing or verification evidence. Source citation mentioned but fact/inference separation not clear. Deductions: lack of verifiable evidence and fact/inference separation.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- No security audit or vulnerability scanning evidence; many dependencies, supply chain risk needs assessment.
- No user confirmation mechanism; external searches may execute automatically, privacy and compliance concerns.
- No rollback mechanism; upgrades or misconfigurations may be hard to recover.
- No version changelog; upgrade risks are opaque.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
What's Changed * Fixed history double-conversion bugs to ensure smoother suggestion generation. * Added validation and timeouts to prevent hung search requests. * Improved widget execution handling; errors no longer crash the entire research pipeline. * Resolved build-time errors by addressing missing package dependencies. * Fixed edge cases involving non-array search query formats. * Integrated serverUtils and updated the reverse geolocation API for better accuracy. * Enhanced error handling for file upload workflows. …View on GitHub ↗ -
Release notes excerpt
What's changed - Added LM Studio provider - Fixed issues with function calling with OpenAI compatible providers. - Fixed issues with JSON parsing and extraction. Full Changelog: https://github.com/ItzCrazyKns/Perplexica/compare/v1.12.0...v1.12.1View on GitHub ↗ -
Release notes excerpt
What's Changed This is a major release featuring a complete architectural overhaul. We've removed Langchain in favor of a custom implementation for low-level control over streaming, generation, and provider-specific features. This refactor lays the groundwork for new capabilities in upcoming releases. General Updates - Renamed providers to connections for better UX. Added switch config field, separated personalization and preferences. - Fixed SearxNG build and startup errors. - Added new agents DIR, migrated suggestions, video and audio agents. …View on GitHub ↗
FAQ
Does Vane support Chinese language queries?
What hardware is required to run Vane?
Can Vane be integrated into existing applications?
How to resolve Ollama connection errors?
http://host.docker.internal:11434, Linux use host's private IP; ensure Ollama listens on 0.0.0.0.