Aller au contenu principal

Code Generation

Category: AI & ML
Node id: code-generation

AI-powered code generation with multi-language and framework support

What it does​

  • Multi-language code generation (auto-detect or specified)
  • Configurable code style (clean, functional, OOP, minimal)
  • Optional unit test generation
  • Documentation and inline comment generation
  • Framework-specific code generation
  • Structured JSON output with code, explanation, and dependencies
  • Design pattern and constraint-aware generation

Use cases​

  • Generating boilerplate code from requirements
  • Creating FHIR integration code with test coverage
  • Scaffolding microservice endpoints with documentation
  • Prototyping data transformation pipelines

Runtime​

  • Performance tier: variable

Input ports​

NameTypeRequiredDescription
requirementtextYesCode requirement description
contextjsonNoAdditional context (sub-keys: context, constraints, examples, existing_code, api_docs, design_pattern)

Output ports​

NameTypeRequiredDescription
resultjsonNoMain result containing code, language, explanation, tests, dependencies, and design_pattern
metadatajsonNoExecution metadata including tokens, line_count, and char_count
successbooleanNoWhether code generation succeeded

Settings​

Basic​

SettingNameTypeRequiredDefaultAllowed valuesDescription
ProviderproviderenumNoopenaiopenai, anthropicLLM provider
ModelmodelstringNo""Model name (auto-selected if empty)
LanguagelanguagestringNoautoTarget programming language (auto for auto-detect)
StylestyleenumNocleanclean, functional, oop, minimalCode style
FrameworkframeworkstringNo""Target framework or library
Include Testsinclude_testsbooleanNofalseGenerate unit tests
Include Docsinclude_docsbooleanNotrueInclude documentation
Include Commentsinclude_commentsbooleanNotrueAdd inline comments

Auth​

SettingNameTypeRequiredDefaultAllowed valuesDescription
API Keyapi_keysecretNoAPI key for the LLM provider
Base URLbase_urlstringNo""Custom API base URL

Advanced​

SettingNameTypeRequiredDefaultAllowed valuesDescription
TemperaturetemperaturenumberNo0.2min 0; max 1Sampling temperature
Max Tokensmax_tokensnumberNo4000min 1Maximum tokens for response

Errors​

CodeDescription
REQUIREMENT_REQUIREDNo requirement input was provided
LLM_REQUEST_FAILEDLLM API call failed during code generation
UNSUPPORTED_PROVIDERProvider must be openai or anthropic

Runtime notes​

  • result.tests is empty unless include_tests=true in config — always check before wiring to a test runner node
  • result.code may contain the raw LLM response if JSON parsing fails — the fallback note appears in result.explanation
  • The context input supports sub-keys (context, constraints, examples, existing_code, api_docs, design_pattern) — pass as a flat JSON object
  • Temperature defaults to 0.2 for deterministic output; increase for more creative/varied code generation

See Build any workflow for how to put nodes together and Ports, types and data flow for how port types connect.

Quelque chose n'est pas clair ? Envoyez vos commentaires sur la documentation par courriel.