AI & ML interests
None defined yet.
Recent Activity
CloverDX
CloverDX is a self-hosted data integration platform: a visual and code-based designer, a runtime server for orchestration and monitoring, and web tools for business users. This page describes how we use AI in it.
How CloverDX approaches AI
AI speeds up building data processing: writing graphs, transformations and mappings, classifying text, extracting structure. The platform keeps running that processing predictable and controllable.
- AI authors, the platform executes. The AI Assistant produces ordinary CloverDX graphs and transformation code. You can read, edit, version-control and test the result, and it runs on the CloverDX Server like any hand-built job. The model does not process your data at run time. Execution is deterministic.
- Where AI runs inside a job, you place it. A graph calls a model through an explicit component, with a prompt you write and fields you choose to send. Surrounding components validate the output, branch on the result and handle failures, so a model response is just another input to checked, logged, restartable processing.
- You choose the model and where it runs. Bring your own key for a cloud provider or run a local model. Local models keep data on your Server. CloverDX does not host models or sell tokens.
- You stay in the loop. Assistants suggest, you accept or reject. Server-side access via MCP can be read-only.
AI in the product
Building pipelines (Designer, MCP Server)
- AI Assistant in CloverDX Designer: generates, explains and troubleshoots graphs in a Server project.
- CloverDX MCP Server: lets external AI IDEs such as Claude Code or Cursor build and inspect jobs through the same tools; read-write or read-only. Models: Anthropic, OpenAI, Azure OpenAI, Google (bring your own key) or local models.
- Troubleshooting: connect Claude or ChatGPT desktop clients to the Server's MCP endpoint to analyse failed runs.
AI inside data transformations (components currently in incubation: supported, interfaces may still change)
AIClient: sends a prompt and selected data to an LLM (OpenAI, Claude, Gemini, Azure AI Foundry, or any local model with an OpenAI-compatible API such as Ollama) and processes the response. Supports multi-step exchanges, for example checking returned JSON and asking the model to correct it.AITextClassifier,AIZeroShotClassifier,AITokenClassifier: score text against fixed or user-defined classes, or label tokens, using models that run locally.AIAnonymizer: token classification (e.g. PII detection) plus masking of the detected tokens.- Local models come from the CloverDX Marketplace or your own, and run on your hardware (Docker image with GPU dependencies, NVIDIA GPU recommended). With cloud LLMs, data leaves for the provider, so the graph decides what is sent.
For business users: Wrangler's AI Assistant turns plain-English requests into transformation steps that the user accepts or rejects. By default only metadata such as column names is shared with the model; admins can disable this.
No extra licence is needed for these features; LLM usage is billed by your provider.
Platform
- CloverDX Server: runtime, orchestration, monitoring, REST data services
- CloverDX Designer: IDE for ETL jobs, transformations and workflows
- Wrangler, Data Manager, Data Apps: self-service data preparation, review and approval of in-flight data, custom apps for business users
Typical use cases: data ingestion, migration, validation and quality, on-premise and hybrid ETL, application integration, data warehousing, onboarding, master/reference data, replacing legacy ETL. Deployed on premises, in your own cloud, hybrid or in containers. Open source origin in 2002, commercial since 2007.
On Hugging Face
CloverDX/ctl: knowledge artefacts about the CloverDX Transformation Language (CTL), for LLM solutions that analyse, generate or modify CTL code (no training data)CloverDX/cloverdx_genai_llm_prompts: LLM prompts and metadata files for generating and understanding CloverDX transformation graphs
Links
Website · AI in CloverDX · AI-powered transformation · AI Assistant · Documentation · Marketplace · Solutions · LinkedIn