Skip to content

Features

CSS/XPATH Generator

Enable via use_llm_for_locator_proposals=false

An internal CSS/XPATH generator will generate a list of unique locator proposals for all relevant elements of a DOM Tree.
Relevant elements are decided based on the used Keyword.
For example, for a keyword Fill Text, the input and textarea elements are more relevant.
For a keyword Click, the button , a and other clickable elements are more relevant.

The generated locators are unique and based on attributes like class, id, placeholder , type and name. Also text values and next/previous siblings as well as parent elements are used to support the unique identification.

AI Supported Self-Healing

Enable via use_llm_for_locator_proposals=true

DOM Tree will be sent to a LLM, which will identify a list of locator proposals based on the failed locator. Locators will depend on the type of LLM.

LLMs can be selected via environment variables:

  • LLM_TEXT_MODEL (model used for picking final locator from proposal list)
  • LLM_LOCATOR_MODEL (model for generating locator proposals from DOM tree)

Collect additional information for locator proposals

Locator Database

Enable via

  • collect_locator_info: Boolean flag to enable or disable the collection of locator information. Default is false.
  • use_locator_db: Boolean flag to enable or disable the use of a locator database for selection of fixed locator. Default is false.
  • locator_db_file: Specifies the filename for the locator database. Default is "locator_db.json".

During a successful test run, additional details for each working locator can be collected and stored into a TinyDB database by setting collect_locator_info to true.

The locator database is just a .json file and the location and name can be set via locator_db_file.

During a test run in which locators shall be healed, the previously collected details in the TinyDB can be used to make better decision for the healed locator.
That behavior can be enabled by setting use_locator_db to True.

Shadow DOM

DOM Tree reduction

  • fix: Specifies the mode of operation, set to "realtime" for real-time healing. Default is "realtime".
  • collect_locator_info: Boolean flag to enable or disable the collection of locator information. Default is false.
  • use_locator_db: Boolean flag to enable or disable the use of a locator database. Default is false.
  • use_llm_for_locator_proposals: Boolean flag to enable or disable the use of a language model for generating locator proposals. If true, locator proposals will be identified from DOM Tree via LLM. If set to false, locator proposals are generated via CSS/XPATH generator. Default is false.
  • heal_assertions: Boolean flag to enable or disable the healing of assertions. Default is false. (not implemented yet)
  • locator_db_file: Specifies the filename for the locator database. Default is "locator_db.json".

Process for Self-Healing

graph TD
  A[end_library_keyword] --> B{Status?};
  B -->|FAIL| C{Broken Locator?};
  C --> |YES| D[Get Fixed Locator];
  D --> F{Get Locator proposals via}
  F --> |LLM| G[Use LLM for Locator Proposals];
  F --> |CSS/XPATH| H[Use CSS/XPATH Generator];
  G --> J[Add Details to Locator Proposals];
  H --> J[Add Details to Locator Proposals];
  B ----->|PASS| E[Continue with Execution];
  J --> LLM[Send Locator Proposals to LLM] --> K{Fixed Locator Found?};
  K -->|YES| E;
  K -->|NO| I;
  C -------> |NO| I[FAIL];