Gl0wFlow – 일반 영어 스크립팅 언어 및 AI용 Rust 런타임

Welcome To Featured AI Script Language Plain-English AI automation powered by GlowScript. GlowFlow is the runtime and product surface for GlowScript — a scripting language built for one job: writing AI automations that stay readable. Current version: GlowFlow v1.1.2 / GlowScript v1.1.2 GlowFlow lets you build agents, webhook handlers, workflow automations, API integrations, MCP tools, and business process scripts in a language that reads like ordered instructions — not framework boilerplate. Instead of scattering logic across Python glue code, shell scripts, YAML workflows, and prompt fragments, GlowScript puts everything in one readable surface. when webhook "/signup" read json body extract from body.message: name email company classify body.message as: urgent normal spam decide next_step from body.message: send_welcome_email ask_for_details escalate_to_human reply "Signup received for {body.email}" Python is flexible but makes simple automations feel lower-level than they need to be. Shell scripts are fragile. YAML workflows fall apart when logic gets complex. GlowFlow sits in the gap: - easier to read than Python automation code - safer and more structured than shell scripts - more expressive for AI workflows than YAML-heavy tools - natively MCP-compatible without extra wiring Requirements: Rust toolchain installed (rustup.rs ) # Clone and build git clone https://github.com/nikolakb/Gl0wFlow.git cd Gl0wFlow cargo build --release # Run your first script cargo run -- run examples/hello.glow # Run the investor demo cargo run -- run examples/investor_demo.glow # Start a webhook server cargo run -- serve examples/webhook_signup.glow 3000 Or install via script: ./scripts/install-from-repo.sh https://github.com/nikolakb/Gl0wFlow.git | Command | What it does | |---|---| glow run file.glow | Execute a script | glow build file.glow | Compile to native binary via Rust | glow check file.glow | Semantic analysis and diagnostics | glow format file.glow | Format and normalize a script | glow inspect file.glow | Print the AST | glow serve file.glow [port] | Run a webhook server (HTTP or HTTPS) | glow repl | Interactive REPL | glow new project | Scaffold a new project | Variables and output set name to "Ana" say "Hello {name}" Control flow if score >= 80 say "Great" else say "Try again" repeat 3 times say "Working" for each lead in leads say lead.email Tasks and functions task sync_sales get "https://api.shop.com/sales" save result to "sales.json" run task sync_sales AI as native syntax extract from message: name email company classify message as: urgent normal spam decide next_step from context: send_invoice ask_for_details escalate_to_human summarize report into short_summary rewrite reply in friendly tone Files, JSON, HTTP set data to load json "leads.json" set response to get "https://api.example.com/status" save json result to "output.json" Error handling try set data to read file "config.json" catch error set data to {status: "fallback"} set response to get "http://api.example.com" recover {status: "offline"} MCP tools use mcp filesystem tool create_invoice needs name, total return {name: name, total: total} export mcp tool create_invoice Context compression compress report into brief auto target w_j 1.618 gain 3.0 mode agent-safe preserve: summary action items keep: first sentence last sentence Agents build agent support_bot on webhook "/support" use mcp filesystem memory session system "You are a helpful support assistant" user body.message reply result - Full compiler pipeline: lexer → parser → AST → semantic analyzer → interpreter → Rust transpiler - Native binary compilation via glow build whenrustc is present - HTTP and HTTPS webhook serving - File, JSON, CSV operations - AI execution via openai ,ollama ,command , andmock providers - MCP filesystem tools and user-defined tool exports - First-class try /catch /throw /recover - Context-control built-ins: calculate tokens ,compress ,collapse - Formatter, AST inspector, REPL, project scaffolding - 160+ tests across the full language surface (consolidated to focused regression suite in v1.1.x) - A full production orchestration platform - A background daemon scheduler - A broad MCP adapter ecosystem beyond filesystem - Complete production hardening across every edge case GlowScript source → Lexer → Parser → AST → Semantic Analyzer → Interpreter (glow run) or → Rust Transpiler → rustc → native binary (glow build) The parser is intentionally small and readable. Diagnostics carry line and column information with suggested fixes. | Operation | Target | |---|---| | CLI startup, small script | < 100ms | | Parse + semantic check | < 50ms | | Interpreter, simple logic | < 100ms | | Webhook response, no AI | 50–300ms | | Full local automation | < 1s | AI provider calls, HTTP APIs, and external tools will dominate real-world latency. GlowFlow aims to add near-zero overhead on top of the work being orchestrated. Apache-2.0