Rayforce — Columnar Analytics and Graph Traversal in One Pipeline

Open source · Pure C · Zero dependencies Analytics and graphs.One fast engine. Rayforce fuses columnar analytics, graph traversal, and recursive queries into one embeddable execution pipeline—built for teams that measure latency in microseconds, not meetings. ; 100,000 deterministic market rows (select {from: trades where: (> qty 100) by: symbol trades: (count qty) notional: (sum (* price qty))}) Loading 100,000 rows Load100K rows Optimizerewrite DAG Filter60K match Group4 symbols symboltradesnotional NVDA15,0001,181,242,500 GOOG15,0001,191,143,000 AAPL15,0001,176,404,200 MSFT15,0001,185,774,600 Join the community One execution model Stop moving databetween engines. Relational operators and graph traversals belong in the same plan. Rayforce sees the whole workload, rewrites it together, and keeps the hot path close to the metal. Explore the execution pipeline ↗ 01 ComposeTables + graphs + rules Lazy DAG 02 OptimizeRewrite the whole plan Multi-pass 03 FuseRemove intermediates Bytecode 04 ExecuteStream cache-sized morsels Parallel 16Klines of focused C 0external dependencies IPCclient/server transport built in MITlicensed and embeddable Rayfall language Ask complex questions.Keep the syntax small. Use Rayfall interactively, embed the C API, or connect through a growing set of client interfaces. Every surface reaches the same optimizer and execution core. One engine, three surfaces ; Aggregate high-value flow, then traverse counterparties (set flow (select {from: trades where: (> Notional 1000000) by: Counterparty Volume: (sum Notional)})) (.graph.var-expand network flow 1 3) EXAMPLE RESULT5 rows · optimized CounterpartyDepthVolume LYNX_02184.20M ALPHA_17261.08M NODE_08244.71M FORT_04327.19M EDGE_12312.42M Built as one system Small footprint.Serious machinery. Everything you need to move from raw columns to connected answers—without assembling a second platform around it. 01Columnar analyticsVectorized filters, joins, groups, windows, pivots, and time-series primitives.↗ 02Native graph engineCSR traversal, shortest paths, centrality, communities, WCO joins, and HNSW.↗ 03Recursive DatalogExpress reachability and recursive relationships as rules compiled into the same DAG.↗ 04Storage that stays openCSV, splayed and partitioned tables, memory maps, IPC, and block offloading.↗ Rayforce in the wild Built for real work,not synthetic demos. Rayforce is already part of production systems and open-source projects spanning trading, market connectivity, investment analytics, and risk. Usage confirmed by the Rayforce project. Consumer names link to their public sites. Rayforce Cloud Coming soon The engine you can embed.The platform you won’t have to operate. We’re bringing Rayforce’s unified analytics and graph pipeline to a managed cloud experience—so teams can move from local prototype to production workload without rebuilding the data path. RAYFORCE / CLOUDPREVIEW Deploymenteu-westReady QUERY SERVICEPreviewENGINEManaged Open at the core Read every line.Own every workload. One codebase. One public header. No external runtime. Clone Rayforce, compile it, embed it, and keep control of your data path. —GitHub stars —forks Linux · macOSsupported today