Orcha Agent OS v1.0 Production

The Semantic Operating System for Multi-Tenant AI Agents

Orcha Agent OS bridges the gap between raw data warehouses and context-aware LLMs. It translates fragmented database schemas into a unified, versionable Model Definition Language (MDL) manifest, allowing AI agents to reliably write, validate, and execute accurate SQL queries without hallucinating schemas.

The Core Problem & Vision

In the era of modern AI, the primary hurdle in enterprise data intelligence is not generating generic SQL syntax—it is business semantics. AI agents lack awareness of what raw database columns mean, how metrics are computed, and which tables must be joined to resolve complex multi-table queries.

Orcha Agent OS provides a Semantic Context Layer:

Eliminates hallucinations by constraining LLM queries to curated Model Definition Language (MDL) definitions.
Inlines complex business formulas automatically using an embedded Rust WASM Apache DataFusion engine.
Automatically calculates and injects multi-hop join paths using graph pathing algorithms.
Dry-plan validates queries before sending them to physical databases, safeguarding production workloads.

Platform Capabilities

Semantic Bridge & ModelerWizard
Map raw table structures into unified business semantics. Define calculated virtual metrics (e.g., margin = revenue - cost) and entity relationships via React Flow.
Rust-Powered WASM Engine
On-the-fly SQL transpilation and formula inlining powered by an embedded WebAssembly build of Apache DataFusion. Sub-second execution with dialect adaptation.
Automatic BFS Join Pathing
Traverses graph relationships using Breadth-First Search algorithms to automatically determine and inject optimal ANSI SQL JOIN clauses across complex schemas.
Federated Multi-DB Queries
Query, aggregate, and join data across disparate database connections simultaneously using intuitive alias.table namespace addressing.
MCP Native Architecture
Expose your databases directly to AI assistants like Claude Desktop, Cursor, and Windsurf via Model Context Protocol tools with dynamic schema resolution.
Full Observability & Grafana
Built-in Prometheus telemetry (/api/metrics) tracking p95 latencies, request throughput, 5xx ratios, and memory usage with pre-provisioned Grafana dashboards.

Supported Database Dialects

PostgreSQL
Native connection pooling with SSL
MySQL / MariaDB
Full TCP multi-host support
MSSQL Server
Enterprise TDS connector
Oracle DB
Enterprise schema integration
SQLite
Fast local & file-based execution

Query Lifecycle at a Glance

1. Intent Classification & Rewriting ───► Extract query goals & resolve context
2. Semantic Vector Recall ──────────────► Retrieve relevant schema & past query patterns
3. Schema Column Pruner ────────────────► Filter out unneeded columns to save LLM tokens
4. LLM Semantic SQL Generation ─────────► Generate SQL query against business MDL layer
5. Rust WASM Transpilation ─────────────► Inlines virtual formulas & resolves table joins
6. Dry-Plan Static SQL Validation ──────► Catch wrong column names or joins pre-execution
7. Native Dialect Database Call ────────► Execute against Postgres/MySQL/MSSQL/SQLite