Skip to content

Robot Framework Heal

Failure triage, self-healing and root-cause analysis for Robot Framework UI tests — Browser/Playwright, SeleniumLibrary and Appium.

Every failed keyword is classified into a failure class, healed when possible, and always turned into a clean, enriched error record with evidence, the healing attempts, and a suggested permanent fix.

Get started in 5 minutes What you get

Why heal

  • Heals more than locators — timing, viewport, overlays, missing form fields, assertion drift, and locator drift, each with its own strategy.
  • Verified, never guessed — every proposed fix is checked against the live page before the keyword reruns.
  • Runs on any model — any OpenAI-compatible endpoint (vLLM, Ollama, LiteLLM, MiniMax, OpenRouter) or a pydantic-ai provider string. Capability is probed, not assumed; small models work via a prompted-JSON floor.
  • Never edits your suite by surprise — fixes are read-only copies + diffs by default; applying them to source is opt-in and blast-radius aware.
  • Explains every failure — even unhealable ones get a root-cause record.

Quickstart

*** Settings ***
Library    Browser    timeout=3s
Library    Heal

# .env (auto-loaded)
HEAL_MODEL=openai/gpt-4.1-nano
HEAL_BASE_URL=https://openrouter.ai/api/v1
HEAL_API_KEY=sk-...

heal doctor --role locator   # verify the endpoint
robot -d results suites/     # heal during the run

See the getting-started tutorial for the full walkthrough.

How it heals

flowchart LR
    F[keyword fails] --> D[deterministic detectors]
    D -->|match| H[heal by class]
    D -->|silent| T[triage agent] --> H
    H -->|verified + rerun| P[PASS + fix proposal]
    H -->|can't heal| R[root-cause record]
    P --> E[report]
    R --> E

Documentation map

  • Tutorials — learn by doing
  • How-to guides — provider setups, Selenium/Appium, CI, fixing files, MCP
  • Reference — every HEAL_* setting, the CLI, failure classes, drivers
  • Explanation — the design and the benchmarks behind it