robotframework-allurevisitor

Generate Allure input files from Robot Framework results. It is an offline alternative to the allure-robotframework listener: instead of hooking into a live test run, it post-processes an existing Robot Framework output.xml via Robot Framework's ResultVisitor API and writes the same *-result.json / *-container.json / *-attachment.* files using the official allure-python-commons engine.

This is handy when you already have an output.xml (from CI, a rerun, or a pabot run) and want an Allure report without re-running the tests.

Installation

pip install robotframework-allurevisitor

This installs the rf-allure console script (with allurevisitor as an alias). Rendering the Allure report additionally requires the Allure CLI.

Requirements

• Python >=3.13
• uv for working on the project from source

From source

uv sync

Usage

Command line (rf-allure, with allurevisitor as an alias):

uv run rf-allure path/to/output.xml -o allure-results --clean

Then render the report with the Allure CLI:

allure serve allure-results

Programmatic:

from robotframework_allurevisitor import generate

generate("output.xml", "allure-results", clean=True)

CLI options

What gets converted

• Suites → containers, with Suite Setup/Teardown recorded as the container's befores/afters.
• Tests → results, with parentSuite/suite/subSuite labels from the test full name, framework/language/host (and optional thread) labels, tags mapped to labels/links (allure.* tags, severity), and a stable historyId = md5(full_name).
• Keywords and control structures (FOR/IF/WHILE/TRY/…) → nested steps, with keyword assignment in the step name and arguments as step parameters.
• Status: PASS→passed, SKIP/NOT RUN→skipped, FAIL→failed, and library/import/syntax/timeout errors → broken.
• Timestamps are taken from the result data (epoch ms), never wall-clock, so conversion is reproducible.
• Log messages → HTML attachments folded into the owning step; embedded screenshots → binary attachments.

Reruns, merges, and pabot

Passing multiple sources combines them; because historyId is stable, repeated runs of the same test are grouped by Allure as retries with the latest attempt (by stop time) as the headline:

uv run rf-allure original.xml rerun.xml -o allure-results --clean

For a single, deduplicated last-wins result, pre-merge with Robot's own tooling and convert the merged file:

rebot --merge original.xml rerun.xml --output merged.xml
uv run rf-allure merged.xml -o allure-results --clean

For pabot, either convert the single merged output.xml, or convert each worker's file with a distinct --thread — but never both into the same results directory (identical historyId would inflate counts).

Allure's trend/history view additionally relies on copying the previous report's history/ directory into allure-results before allure generate; that is an Allure-tooling step, enabled by the stable historyId this tool produces.

Development

uv run pytest          # tests
uv run ruff check .    # lint
uv run ruff format .   # format

End-to-end test

tests/test_e2e_report.py runs the whole pipeline — Robot Framework → the packaged rf-allure CLI → a real Allure HTML report — and asserts the report's widgets/summary.json statistics and data/test-cases/ match the run. It is tagged with the e2e marker:

uv run pytest -m e2e          # only the end-to-end test
uv run pytest -m "not e2e"    # skip it (hermetic fast lane)

Generating the report needs an Allure CLI. The test resolves one automatically: it uses allure if on PATH, otherwise falls back to npx allure-commandline (which needs Node and Java). If neither is available it skips rather than failing. To install a generator:

# Option A: the standalone Allure CLI (https://allurereport.org/docs/install/)
# Option B: via npm (Java required)
npm install -g allure-commandline

The npx allure-commandline fallback downloads the package on first use.

See AGENTS.md for architecture and contribution notes.