Releases: paulomtts/py-ai-toolkit
Releases · paulomtts/py-ai-toolkit
Feat: convenience method for generate task subtrees
Feat: embeddings
Support for embedding generation.
Fix: split validations issues
FIxed an issue with split validations where validation nodes would output strings due to mismatching overloads.
Enh: create_validation_nodes overloads
Added overloads to create_validation_nodes()
Feat: split validation nodes
Validation can now be split with one node per test.
Fix: eval issues
- Eval issues now follow the _eval naming pattern for prompting purposes
- Only failed tests are fed back into the task's eval for reprocessing
Fix: .task() issue
Fixed an issue where BaseWorkflow.task() would return a boolean instead of the node's output.
Fix: exposed LLMConfig import
Exposed missing LLMConfig import.
Fix: create_validation_nodes output evals
create_validation_nodes now evaluates outputs from source node at runtime.
BaseValidation & others
- Added
BaseValidationclass & builder method (inBaseWorkflow). - Enhanced type injection to inherit all
Fieldproperties from the parent class. - Alternative models made available for validation nodes.