Refinement

Refinement is the iterative process of improving generated code after the initial generation pass:

Specification
Your .aexol file
Parse to AST
Syntax tree
AI Generation
Model + KB context
Output Code
TS, Python, Rust, Go, JS

Refinement Types

Via Remote MCP

Remote MCP
curl -X POST "https://api.aexol.ai/mcp" \
  -H "Authorization: Bearer sk-aexol-team-..." \
  -d '{
    "jsonrpc": "2.0", "id": 1, "method": "tools/call",
    "params": {
      "name": "remote_refine_aexol_content",
      "arguments": {
  "content": "type User { id: string }",
  "instruction": "Add email, name fields and UserRole enum"
}
    }
  }'
Remote MCP
curl -X POST "https://api.aexol.ai/mcp" \
  -H "Authorization: Bearer sk-aexol-team-..." \
  -d '{
    "jsonrpc": "2.0", "id": 1, "method": "tools/call",
    "params": {
      "name": "remote_refine_inference_artifact",
      "arguments": {
  "taskId": "task_...",
  "instruction": "Add input validation, error handling, strict types"
}
    }
  }'
Remote MCP
curl -X POST "https://api.aexol.ai/mcp" \
  -H "Authorization: Bearer sk-aexol-team-..." \
  -d '{
    "jsonrpc": "2.0", "id": 1, "method": "tools/call",
    "params": {
      "name": "remote_refine_from_task",
      "arguments": {
  "taskId": "from_task_...",
  "instruction": "Add workflow definitions for all entities"
}
    }
  }'

MCP Tools

ToolWhat it refines
remote_refine_aexol_contentAexol spec (add fields, types, workflows)
remote_refine_inference_artifactGenerated code artifact
remote_refine_from_taskAI Import task result

Knowledge Base Integration

When your project has Knowledge Base documents, refinement automatically incorporates your team's conventions. E.g., if KB says "all API responses use { data, error, meta } envelope", refinement will restructure code accordingly.

Best Practices

  • Be specific — "Add try/catch to all async functions and zod validation" not "Make it better"
  • Iterate incrementally — smaller, focused passes produce better results
  • Use KB — upload conventions before refining

Next Steps