Skip to content
Radhflow

Roadmap

Radhflow is under active development. Here is what is done, what is being built, and what is coming.

Shipped

Section titled “Shipped”
  • Core pipeline engine — flow.yaml parser, graph executor, topological ordering
  • CLI — rf init, rf run, rf validate, rf inspect
  • Four data types — Value, Record, Table, Stream
  • NDJSON interchange with .schema.json companions
  • Schema validation at construction time
  • DuckDB-backed data operations — filter, map, sort, limit, dedup, join, group, sql
  • File connectors — CSV, JSON, NDJSON read/write
  • Git-backed workspace with auto-commit
  • Visual canvas — React Flow editor with node palette, config panels, data preview
  • YAML and canvas stay in sync — edit either
  • Pipeline execution from the UI with status animation

In Progress

Section titled “In Progress”
  • Cloud deployment — EU infrastructure (Hetzner / Scaleway), cron scheduling, managed credentials
  • MCP server — create, validate, and run pipelines from AI agents (Claude Code, Cursor, Copilot)
  • Memory module — key-value store and semantic graph for cross-run state
  • Spreadsheet connectors — Google Sheets and Excel with OAuth, upsert, column mapping
  • Browser nodes — headless browser extraction via agent-browser with interactive step builder

Planned

Section titled “Planned”
  • Conductor and Code Agent — natural language prompt to working pipeline
  • Nix-managed CLI nodes with bubblewrap sandboxing
  • HTTP / REST API connector with auth, pagination, rate limiting
  • Router nodes for conditional branching
  • Secret management in local credential vault
  • Firecracker microVM isolation for agent-generated code
  • Community node registry — publish, discover, install third-party nodes
  • Multi-tenant cloud with team workspaces and role-based access
  • Pipeline composition — nest pipelines as nodes in other pipelines
  • GPU node support for ML inference workloads

Not Planned

Section titled “Not Planned”
  • Real-time streaming execution — Radhflow runs batch pipelines. If you need sub-second event processing, use Kafka or Flink.
  • GUI-only mode — The visual canvas is a view into flow.yaml, not a replacement for it. YAML is always the source of truth.
  • US cloud hosting — Infrastructure runs on EU providers. This is a design decision, not a gap.
  • Self-hosted enterprise — The hosted service uses managed infrastructure. Radhflow runs locally or on Radhflow’s cloud. Customer-managed deployment is not on the roadmap.