{"id":39555476,"url":"https://github.com/cyberlife-coder/VelesDB","last_synced_at":"2026-01-26T15:00:45.997Z","repository":{"id":329166065,"uuid":"1118411810","full_name":"cyberlife-coder/VelesDB","owner":"cyberlife-coder","description":"VelesDB: A high-performance, local-first vector database written in Rust. The SQLite of vector search.","archived":false,"fork":false,"pushed_at":"2026-01-20T16:41:20.000Z","size":18295,"stargazers_count":10,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-01-20T16:44:06.436Z","etag":null,"topics":["ai","embeddings","hnsw","local-first","machine-learning","rag","rust","search-engine","vector-database"],"latest_commit_sha":null,"homepage":"https://velesdb.com","language":"Rust","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/cyberlife-coder.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"docs/SECURITY_AUDIT_2025_01_07.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-12-17T18:01:02.000Z","updated_at":"2026-01-20T11:16:40.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/cyberlife-coder/VelesDB","commit_stats":null,"previous_names":["cyberlife-coder/velesdb"],"tags_count":43,"template":false,"template_full_name":null,"purl":"pkg:github/cyberlife-coder/VelesDB","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cyberlife-coder%2FVelesDB","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cyberlife-coder%2FVelesDB/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cyberlife-coder%2FVelesDB/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cyberlife-coder%2FVelesDB/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/cyberlife-coder","download_url":"https://codeload.github.com/cyberlife-coder/VelesDB/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cyberlife-coder%2FVelesDB/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":28781308,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-01-26T13:55:28.044Z","status":"ssl_error","status_checked_at":"2026-01-26T13:55:26.068Z","response_time":59,"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":["ai","embeddings","hnsw","local-first","machine-learning","rag","rust","search-engine","vector-database"],"created_at":"2026-01-18T07:00:33.418Z","updated_at":"2026-01-26T15:00:45.988Z","avatar_url":"https://github.com/cyberlife-coder.png","language":"Rust","funding_links":["https://buymeacoffee.com/wiscale"],"categories":["Multidimensional data / Vectors","⚙️ Development Tools \u0026 Libraries","Development"],"sub_categories":["Database Solutions","Plugins"],"readme":"\u003cp align=\"center\"\u003e\n  \u003cimg src=\"velesdb_icon_pack/favicon/android-chrome-512x512.png\" alt=\"VelesDB Logo\" width=\"200\"/\u003e\n\u003c/p\u003e\n\n\u003ch1 align=\"center\"\u003e\n  \u003cimg src=\"velesdb_icon_pack/favicon/favicon-32x32.png\" alt=\"VelesDB\" width=\"32\" height=\"32\" style=\"vertical-align: middle;\"/\u003e\n\u003c/h1\u003e\n\n\u003ch3 align=\"center\"\u003e\n  🧠 \u003cstrong\u003eThe Local Knowledge Engine for AI Agents\u003c/strong\u003e 🧠\u003cbr/\u003e\n  \u003cem\u003eVector + Graph + ColumnStore Fusion • 57µs Search • Single Binary • Privacy-First\u003c/em\u003e\n\u003c/h3\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cstrong\u003e🚧 Work in Progress\u003c/strong\u003e — We're actively building the ultimate AI memory engine.\u003cbr/\u003e\n  Ideas, feedback, and contributions are welcome!\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB/actions\"\u003e\u003cimg src=\"https://img.shields.io/github/actions/workflow/status/cyberlife-coder/VelesDB/ci.yml?branch=main\u0026style=flat-square\" alt=\"Build\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB/blob/main/LICENSE\"\u003e\u003cimg src=\"https://img.shields.io/badge/license-ELv2-blue?style=flat-square\" alt=\"License\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB\"\u003e\u003cimg src=\"https://img.shields.io/github/stars/cyberlife-coder/VelesDB?style=flat-square\" alt=\"Stars\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://deepwiki.com/cyberlife-coder/VelesDB\"\u003e\u003cimg src=\"https://deepwiki.com/badge.svg\" alt=\"Ask DeepWiki\"\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"https://img.shields.io/badge/🏎️_Search-57_microsec-blue?style=for-the-badge\" alt=\"Search Latency\"/\u003e\n  \u003cimg src=\"https://img.shields.io/badge/📦_Binary-15MB-orange?style=for-the-badge\" alt=\"Binary Size\"/\u003e\n  \u003cimg src=\"https://img.shields.io/badge/🎯_Recall-100%25-success?style=for-the-badge\" alt=\"Recall\"/\u003e\n\u003c/p\u003e\n\n[![Star History Chart](https://api.star-history.com/svg?repos=cyberlife-coder/velesdb\u0026type=Date)](https://star-history.com/#cyberlife-coder/velesdb\u0026Date)\n\n---\n\n## 🎯 The Problem We Solve\n\n\u003e **\"My RAG agent needs both semantic search AND knowledge relationships. Existing tools force me to choose or glue multiple systems together.\"**\n\n### Three Pain Points That Cost You Time \u0026 Money\n\n| Pain Point | Business Impact | VelesDB Solution |\n|------------|-----------------|------------------|\n| **🐌 Latency kills UX** | Cloud vector DBs add 50-100ms/query. 10 retrievals = **1+ second delay** | **57µs local** — 1000x faster |\n| **🔗 Vectors alone aren't enough** | Semantic similarity misses relationships (\"Who authored this?\") | **Vector + Graph unified** in one query |\n| **🔒 Privacy \u0026 deployment friction** | Cloud dependencies, API keys, GDPR concerns | **15MB binary** — works offline, air-gapped |\n\n### 💰 The ROI of Switching to VelesDB\n\n| Metric | Before (Cloud Stack) | After (VelesDB) | Savings |\n|--------|---------------------|-----------------|---------|\n| **Infrastructure** | 3 databases + sync | 1 binary | **70% less code** |\n| **Cloud costs** | $500-5000/mo | $0 (local) | **100% savings** |\n| **Latency** | 100-300ms | \u003c 1ms | **100x faster** |\n| **Compliance** | Complex (data leaves premises) | Simple (local-first) | **HIPAA/GDPR ready** |\n| **Dev time** | 3 integrations to maintain | 1 API | **3x faster shipping** |\n\n---\n\n## 🏆 Why Developers Choose VelesDB\n\n\u003ctable align=\"center\"\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\" width=\"25%\"\u003e\n\u003ch3\u003e🧠 Vector + Graph + Columns\u003c/h3\u003e\n\u003cp\u003eUnified semantic search, relationships, AND structured data.\u003cbr/\u003e\u003cstrong\u003eNo glue code needed.\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"center\" width=\"25%\"\u003e\n\u003ch3\u003e⚡ 57µs Search\u003c/h3\u003e\n\u003cp\u003eNative HNSW + AVX-512 SIMD.\u003cbr/\u003e\u003cstrong\u003e1000x faster than cloud.\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"center\" width=\"25%\"\u003e\n\u003ch3\u003e📦 15MB Binary\u003c/h3\u003e\n\u003cp\u003eZero dependencies.\u003cbr/\u003e\u003cstrong\u003eWorks offline, air-gapped.\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"center\" width=\"25%\"\u003e\n\u003ch3\u003e🌍 Run Anywhere\u003c/h3\u003e\n\u003cp\u003eServer, Browser, Mobile, Desktop.\u003cbr/\u003e\u003cstrong\u003eSame Rust codebase.\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/table\u003e\n\n---\n\n## 🏢 Coming From Another Vector DB?\n\n| If you use... | VelesDB gives you... |\n|---------------|----------------------|\n| **Pinecone** | No API keys, no cloud costs, **100x faster locally**, + Graph + Columns |\n| **Qdrant** | Single binary (15MB vs 100MB+), native WASM/Mobile, **unified Vector+Graph** |\n| **Milvus** | Zero config vs complex cluster setup, **embedded mode** |\n| **pgvector** | Purpose-built for vectors, **700x faster search**, native graph support |\n| **ChromaDB** | Production-grade Rust vs Python prototype, **enterprise-ready** |\n| **Neo4j + Pinecone** | **One database instead of two**, unified query language |\n\n```sql\n-- The query that defines VelesDB: Vector + Graph in ONE statement\nMATCH (doc:Document)-[:AUTHORED_BY]-\u003e(author:Person)\nWHERE similarity(doc.embedding, $question) \u003e 0.8\n  AND doc.category = 'research'\nRETURN author.name, author.email, doc.title\nORDER BY similarity() DESC\nLIMIT 5;\n```\n\n**This query would require 2 databases and complex sync logic elsewhere. With VelesDB: one query, sub-millisecond response.**\n\n---\n\n## 🌍 Full Ecosystem / Écosystème Complet\n\nVelesDB is designed to run **where your agents live** — from cloud servers to mobile devices to browsers.\n\n| Domain      | Component                          | Description                              | Install                     |\n|-------------|------------------------------------|------------------------------------------|----------------------------|\n| **🦀 Core** | [velesdb-core](crates/velesdb-core) | Core engine (HNSW, SIMD, VelesQL)        | `cargo add velesdb-core`   |\n| **🌐 Server**| [velesdb-server](crates/velesdb-server) | REST API (11 endpoints, OpenAPI)         | `cargo install velesdb-server` |\n| **💻 CLI**  | [velesdb-cli](crates/velesdb-cli)   | Interactive REPL for VelesQL             | `cargo install velesdb-cli` |\n| **🐍 Python** | [velesdb-python](crates/velesdb-python) | PyO3 bindings + NumPy                    | `pip install velesdb`      |\n| **📜 TypeScript** | [typescript-sdk](sdks/typescript) | Node.js \u0026 Browser SDK                    | `npm i @wiscale/velesdb`   |\n| **🌍 WASM** | [velesdb-wasm](crates/velesdb-wasm) | Browser-side vector search               | `npm i @wiscale/velesdb-wasm` |\n| **📱 Mobile** | [velesdb-mobile](crates/velesdb-mobile) | iOS (Swift) \u0026 Android (Kotlin)           | [Build instructions](#-mobile-build) |\n| **🖥️ Desktop** | [tauri-plugin](crates/tauri-plugin-velesdb) | Tauri v2 AI-powered apps               | `cargo add tauri-plugin-velesdb` |\n| **🦜 LangChain** | [langchain-velesdb](integrations/langchain) | Official VectorStore                   | `pip install langchain-velesdb` |\n| **🦙 LlamaIndex** | [llamaindex-velesdb](integrations/llamaindex) | Document indexing                     | `pip install llama-index-vector-stores-velesdb` |\n| **🔄 Migration** | [velesdb-migrate](crates/velesdb-migrate) | From Qdrant, Pinecone, Supabase        | `cargo install velesdb-migrate` |\n\n---\n\n## 🎯 Use Cases\n\n| Use Case                      | VelesDB Feature                     |\n|-------------------------------|-------------------------------------|\n| **RAG Pipelines**             | Sub-ms retrieval                    |\n| **AI Agents**                 | Embedded memory, local context      |\n| **Desktop Apps (Tauri/Electron)** | Single binary, no server needed     |\n| **Mobile AI (iOS/Android)**   | Native SDKs with 32x memory compression |\n| **Browser-side Search**       | WASM module, zero backend           |\n| **Edge/IoT Devices**          | 15MB footprint, ARM NEON optimized  |\n| **On-Prem / Air-Gapped**      | No cloud dependency, full data sovereignty |\n\n---\n\n## 🚀 Quick Start\n\n### Option 1: Linux Package (.deb) ⭐ Recommended for Linux\n\nDownload from [GitHub Releases](https://github.com/cyberlife-coder/VelesDB/releases):\n\n```bash\n# Install\nsudo dpkg -i velesdb-1.1.0-amd64.deb\n\n# Binaries installed to /usr/bin\nvelesdb --version\nvelesdb-server --version\n```\n\n### Option 2: One-liner Script\n\n**Linux / macOS:**\n```bash\ncurl -fsSL https://raw.githubusercontent.com/cyberlife-coder/VelesDB/main/scripts/install.sh | bash\n```\n\n**Windows (PowerShell):**\n```powershell\nirm https://raw.githubusercontent.com/cyberlife-coder/VelesDB/main/scripts/install.ps1 | iex\n```\n\n### Option 3: Python (from source)\n\n```bash\n# Build from source (requires Rust)\ncd crates/velesdb-python\npip install maturin\nmaturin develop --release\n```\n\n```python\nimport velesdb\n\ndb = velesdb.Database(\"./my_vectors\")\ncollection = db.create_collection(\"docs\", dimension=768, metric=\"cosine\")\ncollection.upsert([{\"id\": 1, \"vector\": [...], \"payload\": {\"title\": \"Hello\"}}])\nresults = collection.search([...], top_k=10)\n```\n\n```bash\n# Install from PyPI\npip install velesdb\n```\n\n### Option 4: Rust (from source)\n\n```bash\n# Clone and build\ngit clone https://github.com/cyberlife-coder/VelesDB.git\ncd VelesDB\ncargo build --release\n\n# Binaries in target/release/\n./target/release/velesdb-server --help\n```\n\n```bash\n# Install from crates.io\ncargo install velesdb-cli\n```\n\n### Option 5: Docker (build locally)\n\n```bash\n# Build and run locally\ngit clone https://github.com/cyberlife-coder/VelesDB.git\ncd VelesDB\ndocker build -t velesdb .\ndocker run -d -p 8080:8080 -v velesdb_data:/data velesdb\n```\n\n```bash\n# Pull from GitHub Container Registry\ndocker pull ghcr.io/cyberlife-coder/velesdb:latest\n```\n\n### Option 6: Portable Archives\n\nDownload from [GitHub Releases](https://github.com/cyberlife-coder/VelesDB/releases):\n\n| Platform | File |\n|----------|------|\n| Windows | `velesdb-windows-x86_64.zip` |\n| Linux | `velesdb-linux-x86_64.tar.gz` |\n| macOS (ARM) | `velesdb-macos-arm64.tar.gz` |\n| macOS (Intel) | `velesdb-macos-x86_64.tar.gz` |\n\n### Start Using VelesDB\n\n```bash\n# Start the REST API server (data persisted in ./data)\nvelesdb-server --data-dir ./my_data\n\n# Or use the interactive CLI with VelesQL REPL\nvelesdb repl\n\n# Verify server is running\ncurl http://localhost:8080/health\n# {\"status\":\"healthy\",\"version\":\"1.1.0\"}\n```\n\n📖 **Full installation guide:** [docs/INSTALLATION.md](docs/INSTALLATION.md)\n\n\u003ca name=\"-mobile-build\"\u003e\u003c/a\u003e\n### 📱 Mobile Build (iOS/Android)\n\n```bash\n# iOS (macOS required)\nrustup target add aarch64-apple-ios aarch64-apple-ios-sim\ncargo build --release --target aarch64-apple-ios -p velesdb-mobile\n\n# Android (NDK required)\ncargo install cargo-ndk\ncargo ndk -t arm64-v8a -t armeabi-v7a build --release -p velesdb-mobile\n```\n\n📖 **Full mobile guide:** [crates/velesdb-mobile/README.md](crates/velesdb-mobile/README.md)\n\n---\n\n## 📖 Your First Vector Search\n\n```bash\n# 1. Create a collection\ncurl -X POST http://localhost:8080/collections \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\"name\": \"my_vectors\", \"dimension\": 4, \"metric\": \"cosine\"}'\n\n# 2. Insert vectors with metadata\ncurl -X POST http://localhost:8080/collections/my_vectors/points \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"points\": [\n      {\"id\": 1, \"vector\": [1.0, 0.0, 0.0, 0.0], \"payload\": {\"title\": \"AI Introduction\", \"category\": \"tech\"}},\n      {\"id\": 2, \"vector\": [0.0, 1.0, 0.0, 0.0], \"payload\": {\"title\": \"ML Basics\", \"category\": \"tech\"}},\n      {\"id\": 3, \"vector\": [0.0, 0.0, 1.0, 0.0], \"payload\": {\"title\": \"History of Computing\", \"category\": \"history\"}}\n    ]\n  }'\n\n# 3. Search for similar vectors\ncurl -X POST http://localhost:8080/collections/my_vectors/search \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\"vector\": [0.9, 0.1, 0.0, 0.0], \"top_k\": 2}'\n\n# 4. Or use VelesQL (SQL-like queries)\ncurl -X POST http://localhost:8080/query \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"query\": \"SELECT * FROM my_vectors WHERE vector NEAR $v AND category = '\\''tech'\\'' LIMIT 5\",\n    \"params\": {\"v\": [0.9, 0.1, 0.0, 0.0]}\n  }'\n```\n\n---\n\n## 🔌 API Reference\n\n### Collections\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/collections` | `GET` | List all collections |\n| `/collections` | `POST` | Create a collection |\n| `/collections/{name}` | `GET` | Get collection info |\n| `/collections/{name}` | `DELETE` | Delete a collection |\n\n### Points\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/collections/{name}/points` | `POST` | Upsert points |\n| `/collections/{name}/points/{id}` | `GET` | Get a point by ID |\n| `/collections/{name}/points/{id}` | `DELETE` | Delete a point |\n\n### Search (Vector)\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/collections/{name}/search` | `POST` | Vector similarity search |\n| `/collections/{name}/search/batch` | `POST` | Batch search (multiple queries) |\n| `/collections/{name}/search/multi` | `POST` | Multi-query search |\n\n### Graph\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/collections/{name}/graph/edges` | `GET` | Get edges for a node |\n| `/collections/{name}/graph/edges` | `POST` | Add edge between nodes |\n| `/collections/{name}/graph/traverse` | `POST` | BFS/DFS graph traversal |\n| `/collections/{name}/graph/nodes/{node_id}/degree` | `GET` | Get node degree (in/out) |\n\n### Indexes\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/collections/{name}/indexes` | `GET` | List indexes |\n| `/collections/{name}/indexes` | `POST` | Create index on property |\n| `/collections/{name}/indexes/{label}/{property}` | `DELETE` | Delete index |\n\n### VelesQL v2.0 (Unified Query)\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/query` | `POST` | Execute VelesQL (Vector + Graph + ColumnStore queries) |\n\n**VelesQL v2.0 Features:**\n- `GROUP BY` / `HAVING` with AND/OR operators\n- `ORDER BY` multi-column + `similarity()` function\n- `JOIN` with aliases across collections\n- `UNION` / `INTERSECT` / `EXCEPT` set operations\n- `USING FUSION(strategy='rrf')` hybrid search\n- `WITH (max_groups=100)` query-time config\n\n```sql\n-- Example: Analytics with aggregation\nSELECT category, COUNT(*), AVG(price) FROM products \nGROUP BY category HAVING COUNT(*) \u003e 5\n\n-- Example: Hybrid search with fusion\nSELECT * FROM docs USING FUSION(strategy='rrf', k=60) LIMIT 20\n\n-- Example: Set operations\nSELECT * FROM active UNION SELECT * FROM archived\n```\n\n\u003e **Note:** ColumnStore operations (INSERT, UPDATE, SELECT on structured data) are performed via the `/query` endpoint using VelesQL syntax.\n\n### Health\n\n| Endpoint | Method | Description |\n|----------|--------|-------------|\n| `/health` | `GET` | Health check |\n\n### Request/Response Examples\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cb\u003eCreate Collection\u003c/b\u003e\u003c/summary\u003e\n\n```bash\ncurl -X POST http://localhost:8080/collections \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"name\": \"my_vectors\",\n    \"dimension\": 768,\n    \"metric\": \"cosine\"  # Options: cosine, euclidean, dot, hamming, jaccard\n  }'\n```\n\n**Response:**\n```json\n{\"message\": \"Collection created\", \"name\": \"my_vectors\"}\n```\n\u003c/details\u003e\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cb\u003eUpsert Points\u003c/b\u003e\u003c/summary\u003e\n\n```bash\ncurl -X POST http://localhost:8080/collections/my_vectors/points \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"points\": [\n      {\n        \"id\": 1,\n        \"vector\": [0.1, 0.2, 0.3, ...],\n        \"payload\": {\"title\": \"Document 1\", \"tags\": [\"ai\", \"ml\"]}\n      }\n    ]\n  }'\n```\n\n**Response:**\n```json\n{\"message\": \"Points upserted\", \"count\": 1}\n```\n\u003c/details\u003e\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cb\u003eVector Search\u003c/b\u003e\u003c/summary\u003e\n\n```bash\ncurl -X POST http://localhost:8080/collections/my_vectors/search \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"vector\": [0.1, 0.2, 0.3, ...],\n    \"top_k\": 10\n  }'\n```\n\n**Response:**\n```json\n{\n  \"results\": [\n    {\"id\": 1, \"score\": 0.95, \"payload\": {\"title\": \"Document 1\"}},\n    {\"id\": 42, \"score\": 0.87, \"payload\": {\"title\": \"Document 42\"}}\n  ]\n}\n```\n\u003c/details\u003e\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cb\u003eBatch Search\u003c/b\u003e\u003c/summary\u003e\n\n```bash\ncurl -X POST http://localhost:8080/collections/my_vectors/search/batch \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"searches\": [\n      {\"vector\": [0.1, 0.2, ...], \"top_k\": 5},\n      {\"vector\": [0.3, 0.4, ...], \"top_k\": 5}\n    ]\n  }'\n```\n\n**Response:**\n```json\n{\n  \"results\": [\n    {\"results\": [{\"id\": 1, \"score\": 0.95, \"payload\": {...}}]},\n    {\"results\": [{\"id\": 2, \"score\": 0.89, \"payload\": {...}}]}\n  ],\n  \"timing_ms\": 1.23\n}\n```\n\u003c/details\u003e\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cb\u003eVelesQL Query\u003c/b\u003e\u003c/summary\u003e\n\n```bash\ncurl -X POST http://localhost:8080/query \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"query\": \"SELECT * FROM my_vectors WHERE vector NEAR $v LIMIT 10\",\n    \"params\": {\"v\": [0.1, 0.2, 0.3, ...]}\n  }'\n```\n\n**Response:**\n```json\n{\n  \"results\": [\n    {\"id\": 1, \"score\": 0.95, \"payload\": {\"title\": \"Document 1\"}}\n  ],\n  \"timing_ms\": 2.34,\n  \"rows_returned\": 1\n}\n```\n\u003c/details\u003e\n\n---\n\n## 🧪 Real-World Business Scenarios\n\n\u003e **Each scenario shows a business problem that traditionally requires 2-3 databases. VelesDB solves it with ONE query.**\n\n---\n\n### 💼 Business Scenario 1: E-commerce Product Discovery\n**Industry:** Retail / E-commerce  \n**Problem:** \"Show me products similar to this photo, from trusted suppliers, under $500\"\n\n```sql\n-- Traditional approach: Pinecone (image search) + Neo4j (supplier trust) + PostgreSQL (price)\n-- VelesDB: ONE query\n\nMATCH (product:Product)-[:SUPPLIED_BY]-\u003e(supplier:Supplier)\nWHERE \n  similarity(product.image_embedding, $uploaded_photo) \u003e 0.7  -- Vector: visual similarity\n  AND supplier.trust_score \u003e 4.5                               -- Graph: relationship data\n  AND (SELECT price FROM inventory WHERE sku = product.sku) \u003c 500  -- Column: real-time price\nORDER BY similarity() DESC\nLIMIT 12\n```\n\n**Business Impact:**\n| Metric | Before | After VelesDB |\n|--------|--------|---------------|\n| Query latency | 350ms (3 DBs) | **2ms** |\n| Infrastructure | $2,400/mo | **$0** (local) |\n| Dev complexity | 3 integrations | **1 API** |\n\n---\n\n### 💼 Business Scenario 2: Fraud Detection in Real-Time\n**Industry:** Banking / FinTech  \n**Problem:** \"Flag transactions that look suspicious based on pattern + network + history\"\n\n```sql\n-- Detect fraud: semantic pattern + transaction network + account history\nMATCH (tx:Transaction)-[:FROM]-\u003e(account:Account)-[:LINKED_TO*1..3]-\u003e(related:Account)\nWHERE \n  similarity(tx.behavior_embedding, $known_fraud_pattern) \u003e 0.6  -- Vector: behavioral similarity\n  AND related.risk_level = 'high'                                 -- Graph: network analysis\n  AND (SELECT SUM(amount) FROM transactions \n       WHERE account_id = account.id \n       AND timestamp \u003e NOW() - INTERVAL '24 hours') \u003e 10000       -- Column: velocity check\nRETURN tx.id, account.id, similarity() as fraud_score\n```\n\n**Business Impact:**\n| Metric | Before | After VelesDB |\n|--------|--------|---------------|\n| Detection time | 2-5 seconds | **\u003c 10ms** |\n| False positives | 15% | **8%** (better context) |\n| Compliance | Cloud concerns | **On-premise OK** |\n\n---\n\n### 💼 Business Scenario 3: Healthcare Diagnosis Assistant\n**Industry:** Healthcare / MedTech  \n**Problem:** \"Find similar patient cases with treatment outcomes, HIPAA-compliant\"\n\n```sql\n-- Medical RAG: symptoms + patient network + treatment history\nMATCH (patient:Patient)-[:HAS_CONDITION]-\u003e(condition:Condition)\n      -[:TREATED_WITH]-\u003e(treatment:Treatment)\nWHERE \n  similarity(condition.symptoms_embedding, $current_symptoms) \u003e 0.75  -- Vector: symptom matching\n  AND condition.icd10_code IN ('J18.9', 'J12.89')                     -- Column: specific diagnoses\n  AND (SELECT success_rate FROM treatment_outcomes \n       WHERE treatment_id = treatment.id) \u003e 0.8                       -- Column: outcome data\nRETURN treatment.name, AVG(success_rate) as effectiveness\n```\n\n**Business Impact:**\n| Metric | Before | After VelesDB |\n|--------|--------|---------------|\n| Data location | Cloud (HIPAA risk) | **100% on-premise** |\n| Query time | 500ms+ | **\u003c 5ms** |\n| Integration | 3 vendors | **1 binary** |\n\n---\n\n### 💼 Business Scenario 4: AI Agent Memory (RAG + Context)\n**Industry:** AI / SaaS  \n**Problem:** \"My AI agent needs conversation history + knowledge base + user preferences\"\n\n```sql\n-- Agent memory: semantic recall + conversation graph + user context\nMATCH (user:User)-[:HAD_CONVERSATION]-\u003e(conv:Conversation)\n      -[:CONTAINS]-\u003e(message:Message)\nWHERE \n  similarity(message.embedding, $current_query) \u003e 0.7     -- Vector: relevant past messages\n  AND conv.timestamp \u003e NOW() - INTERVAL '7 days'          -- Column: recent conversations\n  AND (SELECT preference_value FROM user_preferences \n       WHERE user_id = user.id AND key = 'topic') = message.topic  -- Column: user prefs\nORDER BY conv.timestamp DESC, similarity() DESC\nLIMIT 10\n```\n\n**Business Impact:**\n| Metric | Before | After VelesDB |\n|--------|--------|---------------|\n| Context retrieval | 100-200ms | **\u003c 1ms** |\n| Memory footprint | 500MB+ | **15MB binary** |\n| Works offline | ❌ | **✅** |\n\n---\n\n### Scenario 0: Technical Deep-Dive (Vector + Graph + ColumnStore)\n**Goal:** Demonstrate the power of cross-model queries - finding semantically similar documents through graph relationships with structured data filtering\n\n```sql\n-- 🔮 The VelesDB Advantage: One query across all three stores\nMATCH (doc:Document)-[:AUTHORED_BY]-\u003e(author:Author)\nWHERE \n  similarity(doc.embedding, $research_question) \u003e 0.8   -- Vector: semantic search\n  AND doc.category = 'peer-reviewed'                     -- Column: structured filter\n  AND (SELECT citation_count FROM author_metrics         -- Column: subquery\n       WHERE author_id = author.id) \u003e 50\nORDER BY similarity() DESC\nLIMIT 5\n```\n\n**What's happening:**\n1. **Graph traversal**: `MATCH` finds document→author relationships\n2. **Vector search**: `similarity()` ranks by semantic relevance to your question\n3. **Columnar filter**: `category = 'peer-reviewed'` filters structured metadata\n4. **Columnar subquery**: Joins with `author_metrics` table for citation counts\n\n**Expected Output:**\n```json\n{\n  \"results\": [\n    {\n      \"doc.title\": \"Neural Memory Consolidation in AI Agents\",\n      \"author.name\": \"Dr. Sarah Chen\",\n      \"similarity\": 0.94,\n      \"citation_count\": 127\n    }\n  ],\n  \"timing_ms\": 0.8\n}\n```\n\n**Why this matters:** This query would require 3 separate databases and complex synchronization logic in a traditional stack. With VelesDB: **one query, sub-millisecond response**.\n\n---\n\n### Scenario 0b: Multi-Vector Fusion Search (NEAR_FUSED)\n**Goal:** Search using multiple query vectors simultaneously with intelligent result fusion\n\n```sql\n-- 🔮 Multi-modal search: combine text + image embeddings\nSELECT * FROM products \nWHERE vector NEAR_FUSED [$text_embedding, $image_embedding] \n  USING FUSION 'rrf' (k=60)\n  AND category = 'electronics'\nORDER BY similarity() DESC\nLIMIT 10\n```\n\n**Fusion Strategies Available:**\n\n| Strategy | Syntax | Best For |\n|----------|--------|----------|\n| **RRF** | `USING FUSION 'rrf' (k=60)` | Robust rank-based fusion (recommended) |\n| **Average** | `USING FUSION 'average'` | General purpose, balanced results |\n| **Maximum** | `USING FUSION 'maximum'` | Emphasize documents scoring high in ANY query |\n| **Weighted** | `USING FUSION 'weighted' (avg_weight=0.5, max_weight=0.3, hit_weight=0.2)` | Custom control over fusion factors |\n\n**Real-World Use Cases:**\n\n```sql\n-- E-commerce: \"show me products like this photo that match 'wireless headphones'\"\nSELECT * FROM products \nWHERE vector NEAR_FUSED [$image_vector, $text_vector] \n  USING FUSION 'weighted' (avg_weight=0.6, max_weight=0.3, hit_weight=0.1)\nLIMIT 8\n\n-- RAG: Multi-perspective document retrieval\nSELECT * FROM documents \nWHERE vector NEAR_FUSED [$question_embedding, $context_embedding, $user_profile_embedding]\n  USING FUSION 'rrf'\nLIMIT 5\n\n-- Semantic + Lexical hybrid (BM25 + Vector)\nSELECT * FROM articles\nWHERE content MATCH 'artificial intelligence'\n  AND vector NEAR $semantic_embedding\nORDER BY similarity() DESC\nLIMIT 10\n```\n\n**Expected Output:**\n```json\n{\n  \"results\": [\n    {\"id\": 42, \"score\": 0.91, \"fusion_details\": {\"rrf_rank\": 1, \"sources\": 2}},\n    {\"id\": 17, \"score\": 0.87, \"fusion_details\": {\"rrf_rank\": 2, \"sources\": 2}}\n  ],\n  \"timing_ms\": 1.2\n}\n```\n\n---\n\n### Scenario 0c: Distance Metrics for Every Use Case\n**Goal:** Choose the right metric for your data type and domain\n\nVelesDB supports **5 distance metrics** - each optimized for specific use cases:\n\n| Metric | Best For | Example Domain |\n|--------|----------|----------------|\n| **Cosine** | Text embeddings, normalized vectors | NLP, semantic search |\n| **Euclidean** | Spatial data, absolute distances | Geolocation, clustering |\n| **DotProduct** | Pre-normalized embeddings, retrieval | RAG, recommendations |\n| **Hamming** | Binary vectors, fingerprints | Image hashing, DNA |\n| **Jaccard** | Set similarity, sparse features | Tags, categories |\n\n**1. Cosine Similarity (NLP / Semantic Search)**\n```bash\n# Create collection with cosine metric\ncurl -X POST http://localhost:8080/collections \\\n  -d '{\"name\": \"documents\", \"dimension\": 768, \"metric\": \"cosine\"}'\n```\n```sql\n-- Find semantically similar documents (angle-based, ignores magnitude)\nSELECT * FROM documents \nWHERE vector NEAR $query_embedding\nORDER BY similarity() DESC\nLIMIT 10\n```\n**Use case:** ChatGPT-style RAG, document similarity, semantic Q\u0026A\n\n---\n\n**2. Euclidean Distance (Spatial / Clustering)**\n```bash\ncurl -X POST http://localhost:8080/collections \\\n  -d '{\"name\": \"locations\", \"dimension\": 3, \"metric\": \"euclidean\"}'\n```\n```sql\n-- Find nearest physical locations (absolute distance matters)\nSELECT * FROM locations \nWHERE vector NEAR $gps_coordinates\n  AND category = 'restaurant'\nORDER BY similarity() ASC  -- Lower = closer\nLIMIT 5\n```\n**Use case:** Geospatial search, k-means clustering, anomaly detection\n\n---\n\n**3. Dot Product (RAG / Recommendations)**\n```bash\ncurl -X POST http://localhost:8080/collections \\\n  -d '{\"name\": \"products\", \"dimension\": 512, \"metric\": \"dot\"}'\n```\n```sql\n-- Maximize relevance score (pre-normalized embeddings)\nSELECT * FROM products \nWHERE vector NEAR $user_preference_vector\n  AND in_stock = true\nORDER BY similarity() DESC\nLIMIT 8\n```\n**Use case:** Recommendation engines, MaxIP retrieval, MIPS problems\n\n---\n\n**4. Hamming Distance (Binary Vectors / Fingerprints)**\n```bash\ncurl -X POST http://localhost:8080/collections \\\n  -d '{\"name\": \"image_hashes\", \"dimension\": 256, \"metric\": \"hamming\"}'\n```\n```sql\n-- Find near-duplicate images (bit-level comparison, 6ns latency!)\nSELECT * FROM image_hashes \nWHERE vector NEAR $perceptual_hash\n  AND source = 'user_uploads'\nORDER BY similarity() ASC  -- Fewer bit differences = more similar\nLIMIT 10\n```\n**Use case:** Image deduplication, perceptual hashing, DNA sequence matching, malware signatures\n\n---\n\n**5. Jaccard Similarity (Sets / Sparse Features)**\n```bash\ncurl -X POST http://localhost:8080/collections \\\n  -d '{\"name\": \"user_tags\", \"dimension\": 100, \"metric\": \"jaccard\"}'\n```\n```sql\n-- Find users with similar interests (set overlap)\nSELECT * FROM user_tags \nWHERE vector NEAR $current_user_tags\nORDER BY similarity() DESC\nLIMIT 20\n```\n**Use case:** Tag-based recommendations, category matching, collaborative filtering\n\n---\n\n**Performance by Metric (768D vectors):**\n\n| Metric | Latency | Throughput | SIMD Optimized |\n|--------|---------|------------|----------------|\n| **Cosine** | 78 ns | 13M ops/sec | ✅ AVX-512 |\n| **Euclidean** | 44 ns | 23M ops/sec | ✅ AVX-512 |\n| **DotProduct** | 66 ns | 15M ops/sec | ✅ AVX-512 |\n| **Hamming** | **6 ns** | **164M ops/sec** | ✅ POPCNT |\n| **Jaccard** | 89 ns | 11M ops/sec | ✅ AVX2 |\n\n\u003e **Tip:** Hamming is 10x faster than float metrics - ideal for binary embeddings on edge devices!\n\n---\n\n### Scenario 1: Medical Research Assistant\n**Goal:** Find recent oncology studies with specific gene mentions, ordered by relevance\n\n```sql\nSELECT study_id, title, publication_date \nFROM medical_studies \nWHERE \n  vector NEAR $cancer_research_embedding \n  AND content LIKE '%BRCA1%' \n  AND publication_date \u003e '2025-01-01'\nORDER BY similarity() DESC \nLIMIT 5\n```\n\n**Parameters:**\n- `$cancer_research_embedding`: [0.23, 0.87, -0.12, ...] (embedding for \"advanced cancer immunotherapy\")\n\n**Expected Output:**\n```json\n{\n  \"results\": [\n    {\n      \"study_id\": \"onco-2025-042\", \n      \"title\": \"BRCA1 Mutations in Immunotherapy Response\",\n      \"publication_date\": \"2025-03-15\",\n      \"score\": 0.92\n    },\n    {\n      \"study_id\": \"onco-2025-017\",\n      \"title\": \"Gene Editing Approaches for Metastatic Cancer\",\n      \"publication_date\": \"2025-02-28\",\n      \"score\": 0.87\n    }\n  ]\n}\n```\n\n---\n\n### Scenario 2: E-commerce Recommendation Engine\n**Goal:** Recommend products similar to a user's purchase history, within their price range\n\n```sql\nSELECT product_id, name, price \nFROM products \nWHERE \n  vector NEAR $user_preferences \n  AND price BETWEEN 20.00 AND 100.00 \n  AND category = 'electronics'\nORDER BY similarity() DESC, price ASC \nLIMIT 8\n```\n\n**Parameters:**\n- `$user_preferences`: [0.78, -0.23, 0.45, ...] (embedding based on user's purchase history)\n\n**Expected Output:**\n```json\n{\n  \"results\": [\n    {\n      \"product_id\": \"prod-67890\",\n      \"name\": \"Wireless Noise-Cancelling Headphones\",\n      \"price\": 89.99,\n      \"score\": 0.95\n    },\n    {\n      \"product_id\": \"prod-54321\",\n      \"name\": \"Bluetooth Portable Speaker\",\n      \"price\": 59.99,\n      \"score\": 0.91\n    }\n  ]\n}\n```\n\n---\n\n### Scenario 3: Cybersecurity Threat Detection\n**Goal:** Find similar malware patterns observed in the last 7 days\n\n```sql\nSELECT malware_hash, threat_level, first_seen \nFROM threat_intel \nWHERE \n  vector NEAR $current_threat_embedding \n  AND first_seen \u003e NOW() - INTERVAL '7 days'\n  AND threat_level \u003e 0.8\nORDER BY similarity() DESC, first_seen DESC\nLIMIT 10\n```\n\n**Parameters:**\n- `$current_threat_embedding`: [0.12, -0.87, 0.34, ...] (embedding of current malware signature)\n\n**Troubleshooting Tip:** If no results appear, verify:\n1. Threat intelligence feed is updating daily\n2. Vector dimensions match collection configuration\n3. Timestamp format matches ISO 8601 (YYYY-MM-DD HH:MM:SS)\n\n---\n\n## ⚡ Performance\n\n\n### 🔥 Core Vector Operations (768D - BERT/OpenAI dimensions)\n\n| Operation | Latency | Throughput | vs. Naive |\n|-----------|---------|------------|----------|\n| **Dot Product (1536D)** | **66 ns** | **15M ops/sec** | 🚀 **8x faster** |\n| **Euclidean (768D)** | **44 ns** | **23M ops/sec** | 🚀 **6x faster** |\n| **Cosine (768D)** | **78 ns** | **13M ops/sec** | 🚀 **4x faster** |\n| **Hamming (Binary)**| **6 ns** | **164M ops/sec** | 🚀 **10x faster** |\n\n### 📊 System Performance (10K Vectors, Local)\n\n| Benchmark | Result | Details |\n|-----------|--------|---------|\n| **HNSW Search** | **57 µs** | p50 latency (Cosine) |\n| **VelesQL Parsing**| **554 ns** | Simple SELECT |\n| **VelesQL Cache Hit**| **48 ns** | HashMap pre-allocation |\n| **Recall@10** | **100%** | Perfect mode (brute-force SIMD) |\n| **BM25 Search** | **33 µs** | Adaptive PostingList (10K docs) |\n\n### 🎯 Search Quality (Recall)\n\n| Mode | Recall@10 | Latency (128D) | Use Case |\n|------|-----------|----------------|----------|\n| Fast | 92.2% | ~26µs | Real-time, high throughput |\n| Balanced | 98.8% | ~39µs | Production recommended |\n| Accurate | 100% | ~67µs | High precision |\n| **Perfect** | **100%** | ~220µs | Brute-force SIMD |\n\n### 🛠️ Optimizations Under the Hood\n\n- **SIMD**: AVX-512/AVX2 auto-detection with 32-wide FMA\n- **Prefetch**: CPU cache warming for HNSW traversal (+12% throughput)\n- **Contiguous Layout**: 64-byte aligned memory for cache efficiency\n- **Batch WAL**: Single disk write per batch import\n- **Zero-Copy**: Memory-mapped files for fast startup\n\n\u003e 📊 Full benchmarks: [docs/BENCHMARKS.md](docs/BENCHMARKS.md)\n\n### 📦 Vector Quantization (Memory Reduction)\n\nReduce memory usage by **4-32x** with minimal recall loss:\n\n| Method | Compression | Recall Loss | Use Case |\n|--------|-------------|-------------|----------|\n| **SQ8** (8-bit) | **4x** | \u003c 2% | General purpose, Edge |\n| **Binary** (1-bit) | **32x** | ~10-15% | Fingerprints, IoT |\n\n```rust\nuse velesdb_core::quantization::{QuantizedVector, dot_product_quantized_simd};\n\n// Compress 768D vector: 3072 bytes → 776 bytes (4x reduction)\nlet quantized = QuantizedVector::from_f32(\u0026embedding);\n\n// SIMD-optimized search (only ~30% slower than f32)\nlet similarity = dot_product_quantized_simd(\u0026query, \u0026quantized);\n```\n\n\u003e 📖 Full guide: [docs/QUANTIZATION.md](docs/QUANTIZATION.md)\n\n---\n\n## 🚀 Transformative Benefits: How VelesDB Changes Development\n\n### ⚡ Eliminates Database Sprawl\nVelesDB replaces 3+ specialized databases (vector DB + graph DB + document store) with a **single unified engine**.\n\n```mermaid\ngraph LR\nA[App] --\u003e V[VelesDB]\n```\n\n**Impact:**\n- ✅ **70% less infrastructure code**\n- ✅ **No synchronization headaches**\n- ✅ **Single query language for all operations**\n\n### 💡 Enables New Application Categories\nWith air-gapped deployment and 15MB binary size:\n```mermaid\npie title Deployment Locations\n    \"On-Prem Servers\" : 35\n    \"Edge Devices\" : 25\n    \"Mobile Apps\" : 20\n    \"Browser WASM\" : 20\n```\n\n**Impact:**\n- ✅ **Build HIPAA-compliant healthcare apps**\n- ✅ **Create military-grade analytics** for air-gapped environments\n- ✅ **Enable privacy-first consumer apps** with zero data sharing\n\n### 🚀 Redefines Performance Expectations\n\n| Pipeline Step | Cloud Vector DB | VelesDB |\n|---------------|-----------------|---------|\n| Network round-trip | 50-100ms | **0ms** (local) |\n| Vector search | 10-50ms | **0.057ms** |\n| Graph traversal | 20-100ms | **0.1ms** |\n| **Total latency** | **100-250ms** | **\u003c 1ms** |\n\n\u003e 💡 **100x faster** enables use cases impossible with cloud: real-time autocomplete, instant RAG, sub-frame game AI\n\n**Impact:**\n- ✅ **Build real-time AI agents** that respond faster than human perception\n- ✅ **Enable complex RAG chains** with 10+ sequential retrievals\n- ✅ **Create instant search experiences** with no loading spinners\n\n### 💼 Unified API Simplifies Development\nOne consistent API across all platforms:\n```rust\n// Same API everywhere\nlet results = db.search(query_vector, filters, graph_traversal);\n```\n\n**Impact:**\n- ✅ **Learn once, deploy everywhere**\n- ✅ **Shared codebase** between web, mobile, and desktop\n- ✅ **Eliminate platform-specific database code**\n\n---\n\n## ✨ Core Features That Transform Development\n\n| Feature | Technical Capability | Real-World Impact |\n|---------|----------------------|-------------------|\n| **🧠 Vector + Graph Fusion** | Unified query language for semantic + relationship queries | **Build smarter AI agents** with contextual understanding |\n| **⚡ 57µs Search** | Native HNSW + AVX-512 SIMD | **Create real-time experiences** previously impossible |\n| **📦 15MB Binary** | Zero dependencies, single executable | **Deploy anywhere** - from servers to edge devices |\n| **🔒 Air-Gapped Deployment** | Full functionality without internet | **Meet strict compliance** in healthcare/finance |\n| **🌍 Everywhere Runtime** | Consistent API across server/mobile/browser | **Massive code reuse** across platforms |\n| **🧠 SQ8 Quantization** | 4x memory reduction | **Run complex AI** on resource-constrained devices |\n| **📝 VelesQL** | SQL-like unified query language | **Simplify complex queries** - no DSL learning curve |\n\n---\n\n## 🏆 Real-World Impact Stories\n\n### 🏥 Healthcare Diagnostics Assistant\n**Before VelesDB:**\n- 300ms latency per query\n- Patient data in cloud\n- Separate systems for medical knowledge and patient data\n\n**With VelesDB:**\n- **0.6ms diagnosis suggestions**\n- **On-device patient data**\n- **Unified medical knowledge graph**\n\n```mermaid\npie title Performance Improvement\n    \"Diagnosis Speed\" : 85\n    \"Accuracy\" : 10\n    \"Privacy\" : 5\n```\n\n### 🏭 Manufacturing Quality Control\n**Before VelesDB:**\n- Cloud dependency caused production delays\n- Separate systems for defect images and part metadata\n\n**With VelesDB:**\n```sql\nMATCH (part)-[HAS_DEFECT]-\u003e(defect)\nWHERE defect.vector NEAR $image_vec\nAND part.material = 'titanium'\n```\n- **50% fewer defective shipments**\n- **Offline factory floor operation**\n- **Unified defect database**\n\n---\n\n## 🤝 Contributing\n\nWe welcome contributions! Here's how to get started:\n\n### Development Setup\n\n```bash\n# Clone the repo\ngit clone https://github.com/cyberlife-coder/VelesDB.git\ncd VelesDB\n\n# Run tests\ncargo test --all-features\n\n# Run with checks (before committing)\ncargo fmt --all\ncargo clippy --all-targets --all-features -- -D warnings\n```\n\n### Project Structure\n\n```\nVelesDB/\n├── crates/\n│   ├── velesdb-core/     # Core engine library\n│   │   ├── src/\n│   │   │   ├── collection/   # Collection management\n│   │   │   ├── index/        # HNSW index\n│   │   │   ├── storage/      # Persistence layer\n│   │   │   ├── velesql/      # Query language parser\n│   │   │   └── simd/         # SIMD optimizations\n│   │   └── tests/\n│   ├── velesdb-server/   # REST API server\n│   ├── velesdb-mobile/   # iOS/Android bindings (UniFFI)\n│   ├── velesdb-wasm/     # WebAssembly module\n│   └── velesdb-python/   # Python bindings (PyO3)\n├── benches/              # Benchmarks\n└── docs/                 # Documentation\n```\n\n### Good First Issues\n\nLooking for a place to start? Check out issues labeled [`good first issue`](https://github.com/cyberlife-coder/VelesDB/labels/good%20first%20issue).\n\n---\n\n## 📊 Roadmap\n\n```mermaid\ngantt\n    title VelesDB Development Timeline\n    dateFormat YYYY-MM\n    section v1.2 ✅\n    Knowledge Graph Storage     :done, 2025-10, 2025-12\n    VelesQL MATCH Clause        :done, 2025-11, 2025-12\n    Vector-Graph Fusion         :done, 2025-12, 2026-01\n    ColumnStore CRUD            :done, 2026-01, 2026-01\n    VelesQL JOIN Cross-Store    :done, 2026-01, 2026-01\n    section v1.3 🔄\n    Aggregations VelesQL        :active, 2026-02, 2026-03\n    Agent Memory Patterns       :2026-02, 2026-04\n    Documentation               :2026-02, 2026-03\n    section v1.4 📋\n    E2E Test Suite              :2026-04, 2026-05\n    Durability \u0026 Recovery       :2026-04, 2026-06\n```\n\n### Progress Overview\n\n| Version | Status | EPICs Done | Progress |\n|---------|--------|------------|----------|\n| **v1.2.0** | ✅ Released | 15/15 | ![100%](https://progress-bar.xyz/100?title=Complete) |\n| **v1.3.0** | 🔄 In Progress | 0/6 | ![4%](https://progress-bar.xyz/4?title=Building) |\n| **v1.4.0** | 📋 Planned | 0/5 | ![0%](https://progress-bar.xyz/0?title=Planned) |\n\n---\n\n### v1.2.0 ✅ Released (January 2026)\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cb\u003e15 EPICs Completed - Click to expand\u003c/b\u003e\u003c/summary\u003e\n\n| EPIC | Feature | Impact |\n|------|---------|--------|\n| EPIC-001 | ✅ Code Quality Refactoring | Clean architecture |\n| EPIC-002 | ✅ GPU Acceleration (wgpu) | 10x throughput |\n| EPIC-003 | ✅ PyO3 Migration | Python 3.12+ support |\n| EPIC-004 | ✅ Knowledge Graph Storage | GraphNode, GraphEdge, BFS |\n| EPIC-005 | ✅ VelesQL MATCH Clause | Cypher-inspired queries |\n| EPIC-006 | ✅ Agent Toolkit SDK | Python, WASM, Mobile |\n| EPIC-007 | ✅ Python Bindings Refactoring | Clean API |\n| EPIC-008 | ✅ Vector-Graph Fusion | `similarity()` in MATCH |\n| EPIC-009 | ✅ Graph Property Index | 10x faster MATCH |\n| EPIC-019 | ✅ Scalability 10M entries | Enterprise datasets |\n| EPIC-020 | ✅ ColumnStore CRUD | Real-time updates |\n| EPIC-021 | ✅ VelesQL JOIN Cross-Store | Graph ↔ Table queries |\n| EPIC-028 | ✅ ORDER BY Multi-Columns | Complex sorting |\n| EPIC-029 | ✅ Python SDK Core Delegation | DRY bindings |\n| EPIC-031 | ✅ Multimodel Query Engine | Unified execution |\n\n\u003c/details\u003e\n\n**Highlights:**\n- 🧠 **Knowledge Graph** - Full MATCH clause with BFS traversal\n- 🔮 **Vector-Graph Fusion** - `WHERE similarity() \u003e 0.8` in graph queries\n- 📊 **ColumnStore** - Real-time CRUD with JOIN support\n- 📦 **Published** - crates.io, PyPI, npm\n\n---\n\n### v1.3.0 🔄 In Progress (Q1 2026)\n\n| EPIC | Feature | Priority | Progress |\n|------|---------|----------|----------|\n| EPIC-016 | **SDK Ecosystem Sync** | 🔥 Critical | 🔄 21% (3/14 US) |\n| EPIC-017 | **Aggregations VelesQL** | 🔥 Critical | 📋 0% |\n| EPIC-010 | **Agent Memory Patterns SDK** | 🚀 High | 📋 0% |\n| EPIC-018 | **Documentation \u0026 Examples** | 🚀 High | 📋 0% |\n| EPIC-012 | **TypeScript SDK** | 📋 Medium | 📋 0% |\n| EPIC-013 | **LangChain/LlamaIndex** | 📋 Medium | 📋 0% |\n\n**Coming Soon:**\n```sql\n-- Aggregations (EPIC-017)\nSELECT category, COUNT(*), AVG(price) \nFROM products \nGROUP BY category\nHAVING COUNT(*) \u003e 10\n\n-- Agent Memory Patterns (EPIC-010)  \nINSERT INTO agent_memory (episode, embedding, context)\nVALUES ('task_123', $vec, '{\"goal\": \"find documents\"}')\n```\n\n---\n\n### v1.4.0 📋 Planned (Q2 2026)\n\n| EPIC | Feature | Focus |\n|------|---------|-------|\n| EPIC-011 | **E2E Test Suite** | Quality assurance |\n| EPIC-024 | **Durability \u0026 Crash Recovery** | Database-grade reliability |\n| EPIC-022 | **Unsafe Auditability** | Security audit |\n| EPIC-023 | **Loom Concurrency Proofs** | Thread safety |\n| EPIC-025 | **Miri/Fuzzing** | Memory safety |\n\n---\n\n### Future Vision\n\n```mermaid\npie title VelesDB Feature Distribution\n    \"Vector Search\" : 30\n    \"Knowledge Graph\" : 25\n    \"ColumnStore\" : 20\n    \"Query Engine\" : 15\n    \"SDKs \u0026 Integrations\" : 10\n```\n\n| Horizon | Features |\n|---------|----------|\n| **2026 H2** | Sparse vectors, Product Quantization (PQ) |\n| **2027** | Distributed mode (Premium), Cluster HA |\n| **Beyond** | Agent Hooks \u0026 Triggers, Multi-tenancy |\n\n---\n\n## 📜 License\n\nVelesDB is licensed under the [Elastic License 2.0 (ELv2)](LICENSE).\n\nELv2 is a source-available license that allows free use, modification, and distribution, with restrictions only on providing the software as a managed service.\n\n---\n\n## ⭐ Support VelesDB\n\nIf VelesDB helps your project, here's how you can support us:\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB\"\u003e\n    \u003cimg src=\"https://img.shields.io/badge/⭐_Star_on_GitHub-181717?style=for-the-badge\u0026logo=github\" alt=\"Star on GitHub\"/\u003e\n  \u003c/a\u003e\n  \u003ca href=\"https://twitter.com/intent/tweet?text=🚀%20Check%20out%20VelesDB%20-%20The%20Local%20Knowledge%20Engine%20for%20AI%20Agents!%20Vector%20%2B%20Graph%20%2B%20ColumnStore%20in%20one%2015MB%20binary.\u0026url=https://github.com/cyberlife-coder/VelesDB\u0026hashtags=VectorDatabase,AI,Rust,OpenSource\"\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Share_on_Twitter-1DA1F2?style=for-the-badge\u0026logo=twitter\u0026logoColor=white\" alt=\"Share on Twitter\"/\u003e\n  \u003c/a\u003e\n\u003c/p\u003e\n\n### ☕ Buy Me A Coffee\n\nBuilding VelesDB takes countless hours. If you find it useful, consider supporting development:\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://buymeacoffee.com/wiscale\" target=\"_blank\"\u003e\n    \u003cimg src=\"https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png\" alt=\"Buy Me A Coffee\" style=\"height: 60px; width: 217px;\"\u003e\n  \u003c/a\u003e\n\u003c/p\u003e\n\n### 🏷️ Show You Use VelesDB\n\nAdd the badge to your project:\n\n[![Powered by VelesDB](https://img.shields.io/badge/Powered_by-VelesDB-blue?style=flat-square)](https://github.com/cyberlife-coder/VelesDB)\n\n```markdown\n[![Powered by VelesDB](https://img.shields.io/badge/Powered_by-VelesDB-blue?style=flat-square)](https://github.com/cyberlife-coder/VelesDB)\n```\n\n---\n\n\u003cp align=\"center\"\u003e\n  \u003cstrong\u003eBuilt with ❤️ and 🦀 Rust\u003c/strong\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cstrong\u003eOriginal Author:\u003c/strong\u003e \u003ca href=\"https://github.com/cyberlife-coder\"\u003eJulien Lange\u003c/a\u003e — \u003ca href=\"https://wiscale.io\"\u003e\u003cstrong\u003eWiScale\u003c/strong\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB\"\u003e⭐ GitHub\u003c/a\u003e •\n  \u003ca href=\"https://deepwiki.com/cyberlife-coder/VelesDB/\"\u003e📖 Documentation\u003c/a\u003e •\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB/issues\"\u003e🐛 Issues\u003c/a\u003e •\n  \u003ca href=\"https://github.com/cyberlife-coder/VelesDB/releases\"\u003e📦 Releases\u003c/a\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003csub\u003eDon't forget to ⭐ star the repo if you find VelesDB useful!\u003c/sub\u003e\n\u003c/p\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fcyberlife-coder%2FVelesDB","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fcyberlife-coder%2FVelesDB","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fcyberlife-coder%2FVelesDB/lists"}