{"id":43383402,"url":"https://github.com/23skdu/longbow","last_synced_at":"2026-02-02T09:23:09.988Z","repository":{"id":329244694,"uuid":"1115589668","full_name":"23skdu/longbow","owner":"23skdu","description":"Apache Arrow Flight clustered vector cache for high throughput Agent memory sharing ","archived":false,"fork":false,"pushed_at":"2026-01-29T18:09:56.000Z","size":332342,"stargazers_count":3,"open_issues_count":1,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-01-29T22:29:57.358Z","etag":null,"topics":["agentic-ai","agents","arrow","cache","clustering","database","golang","golang-application","hub","memory-cache","mesh","mesh-networks","mesh-processing","swarm-agents","swarm-algorithm","swarm-intelligence","swarms","vector-database","vectors"],"latest_commit_sha":null,"homepage":"https://github.com/23skdu/longbow","language":"Go","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/23skdu.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":"docs/security.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":"2025-12-13T06:29:53.000Z","updated_at":"2026-01-29T18:09:49.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/23skdu/longbow","commit_stats":null,"previous_names":["23skdu/longbow"],"tags_count":13,"template":false,"template_full_name":null,"purl":"pkg:github/23skdu/longbow","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/23skdu%2Flongbow","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/23skdu%2Flongbow/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/23skdu%2Flongbow/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/23skdu%2Flongbow/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/23skdu","download_url":"https://codeload.github.com/23skdu/longbow/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/23skdu%2Flongbow/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29009614,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-02T08:40:12.472Z","status":"ssl_error","status_checked_at":"2026-02-02T08:40:10.926Z","response_time":58,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: 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":["agentic-ai","agents","arrow","cache","clustering","database","golang","golang-application","hub","memory-cache","mesh","mesh-networks","mesh-processing","swarm-agents","swarm-algorithm","swarm-intelligence","swarms","vector-database","vectors"],"created_at":"2026-02-02T09:23:07.464Z","updated_at":"2026-02-02T09:23:09.981Z","avatar_url":"https://github.com/23skdu.png","language":"Go","funding_links":[],"categories":[],"sub_categories":[],"readme":"![CI](https://github.com/23skdu/longbow/actions/workflows/ci.yml/badge.svg)\n![Helm Validation](https://github.com/23skdu/longbow/actions/workflows/helm-validation.yml/badge.svg)\n![Markdown Lint](https://github.com/23skdu/longbow/actions/workflows/markdown-lint.yml/badge.svg)\n\n# Longbow\n\n\u003cimg width=\"1024\" height=\"559\" alt=\"image\" src=\"https://github.com/user-attachments/assets/16b4632a-f09b-42ab-9b05-9ab5a25566bf\" /\u003e\n\nLongbow is a distributed, high-performance vector cache built for modern AI/Agentic workloads. It leverages zero-copy data paths, SIMD optimizations, and advanced storage backends to deliver sub-millisecond latency.\n\n## Key Features\n\n- **High Performance**: Built on Apache Arrow for zero-copy data transfer.\n- **Distributed**: Consistent hashing and gossip-based membership (SWIM protocol).\n- **Optimized Storage**: Optional `io_uring` WAL backend for high-throughput ingestion.\n- **Hardware Aware**: NUMA-aware memory allocation and SIMD vector distance calculations.\n- **Smart Client**: Resilient Go SDK that handles request routing transparently.\n\n## Architecture\n\nLongbow uses a shared-nothing architecture where every node is loosely coupled.\n\nSee [Architecture Guide](docs/architecture.md) for a deep dive.\n\n## Getting Started\n\n### Prerequisites\n\n- Go 1.25+\n- Linux (recommended for best performance) or macOS\n\n### Installation\n\n```bash\ngit clone https://github.com/23skdu/longbow.git\ncd longbow\ngo build -o bin/longbow ./cmd/longbow\n```\n\n### Running a Local Cluster\n\n```bash\n./scripts/start_local_cluster.sh\n```\n\n### Running Benchmarks\n\nLongbow includes a distributed benchmark tool:\n\n```bash\ngo build -o bin/bench-tool ./cmd/bench-tool\n./bin/bench-tool --mode=ingest --concurrency=4 --duration=10s\n```\n\n## Configuration\n\nLongbow is configured via environment variables. See [Configuration](docs/configuration.md) for details.\n\nNotable flags:\n\n- `STORAGE_USE_IOURING=true` (Enable new Linux storage engine)\n- `LONGBOW_GOSSIP_ENABLED=true` (Enable distributed discovery)\n\n- **Protocol**: Apache Arrow Flight (over gRPC/HTTP2).\n- **Search**: High-performance HNSW vector search with hybrid (Dense + Sparse) support and polymorphic vector types (F32, F16, Int8, Complex64/128).\n- **Distance Metrics**: Pluggable metrics (Euclidean, Cosine, Dot Product) with SIMD optimizations for all supported types.\n- **Filtering**: Metadata-aware predicate filtering for searches and scans.\n- **Lifecycle**: Support for vector deletion via tombstones.\n- **Durable**: WAL with Apache Parquet format snapshots.\n- **Storage**: In-memory ephemeral storage for zero-copy high-speed access.\n- **Observability**: Structured JSON logging and 100+ Prometheus metrics.\n\n## Architecture \u0026 Ports\n\nTo ensure high performance under load, Longbow splits traffic into two dedicated gRPC servers:\n\n- **Data Server (Port 3000)**: Handles heavy I/O operations (`DoGet`, `DoPut`, `DoExchange`).\n- **Meta Server (Port 3001)**: Handles lightweight metadata operations (`ListFlights`, `GetFlightInfo`, `DoAction`).\n\n**Why?**\nSeparating these concerns prevents long-running data transfer operations from blocking metadata requests. This ensures\nthat clients can always discover streams and check status even when the system is under heavy write/read load.\n\n## Observability \u0026 Metrics\n\nLongbow exposes Prometheus metrics on a dedicated port to ensure observability without impacting the main Flight\nservice.\n\n- **Scrape Port**: 9090\n- **Scrape Path**: /metrics\n\n### Custom Metrics\n\n### Key Metrics\n\n| Metric Name | Type | Description |\n| :--- | :--- | :--- |\n| `longbow_flight_ops_total` | Counter | Total number of Flight operations (DoGet, DoPut, etc.) |\n| `longbow_flight_duration_seconds` | Histogram | Latency distribution of Flight operations |\n| `longbow_flight_rows_processed_total` | Counter | Total rows processed in scans and searches |\n| `longbow_hnsw_search_duration_seconds` | Histogram | Latency of k-NN search operations |\n| `longbow_hnsw_node_count` | Gauge | Current number of vectors in the index |\n| `longbow_tombstones_total` | Gauge | Number of active deleted vector tombstones |\n| `longbow_index_queue_depth` | Gauge | Depth of the asynchronous indexing queue |\n| `longbow_memory_fragmentation_ratio` | Gauge | Ratio of system memory reserved vs used |\n| `longbow_wal_bytes_written_total` | Counter | Total bytes written to the WAL |\n| `longbow_snapshot_duration_seconds` | Histogram | Duration of the Parquet snapshot process |\n| `longbow_evictions_total` | Counter | Total number of evicted records (LRU) |\n| `longbow_ipc_decode_errors_total` | Counter | Count of IPC decoding errors or panics |\n\nFor a detailed explanation of all 100+ metrics, see [Metrics Documentation](docs/metrics.md).\n\nStandard Go runtime metrics are also exposed.\n\n## Usage\n\n### Running locally\n\n```bash\ngo run cmd/longbow/main.go\n```\n\n### Docker\n\n```bash\ndocker build -t longbow .\ndocker run -p 3000:3000 -p 3001:3001 -p 9090:9090 longbow\n```\n\n## Documentation\n\n- [Distance Metrics](docs/distance_metrics.md)\n- [Persistence \u0026 Snapshots](docs/persistence.md)\n- [Vector Search Architecture](docs/vectorsearch.md)\n- [Troubleshooting Guide](docs/troubleshooting.md)\n- [Metrics](docs/metrics.md)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2F23skdu%2Flongbow","html_url":"https://awesome.ecosyste.ms/projects/github.com%2F23skdu%2Flongbow","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2F23skdu%2Flongbow/lists"}