desplega-ai/Run Argus Session

Run an Argus agent session against a target URL and post the report as a PR comment.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
scenarioScenario identifier. Used for sticky-comment marker and to look up a scenario file under .github/argus-scenarios/<scenario>.md. Optional — defaults to "argus". When no scenario file, prompt_template, or prompt_template_file is provided, the built-in DEFAULT_PROMPT is used.noargus
prompt_templateInline prompt body. When set, overrides prompt_template_file and scenario file lookup. Supports ${VAR} substitution.no""
prompt_template_filePath (relative to the workspace) to a prompt template file. Used when prompt_template is empty.no""
base_urlTarget URL the agent will exercise. Substituted as ${BASE_URL} in the prompt.yes
argus_base_urlArgus backend base URL.nohttps://api.desplega.ai
ui_base_urlUI base URL used for session links in the PR comment.nohttps://app.desplega.ai
extra_varsJSON object of additional ${VAR} values merged over the built-ins.no{}
wait_modepoll (stream + poll outcome) or no (fire-and-forget).nopoll
timeout_sLocal wall-clock budget (seconds) for stream tail + outcome poll fallback when agent_end is not observed.no900
fail_onfailed (default) fails the job on non-pass outcome; never keeps the job green.nofailed
comment_on_prWhen true, upserts a sticky PR comment with the agent report.notrue
comment_keyOptional discriminator appended to the sticky-comment marker. Use when the same scenario runs against multiple base_urls on one PR.no""
github_tokenToken used to post the PR comment. Requires pull-requests: write.no${{ github.token }}
namedescription
session_idArgus session id.
instance_idArgus instance id.
poll_urlFull URL for polling the session outcome.
outcome_statusTerminal outcome status (completed | aborted | failed | error | timed_out), or empty for wait_mode=no.
comment_urlURL of the upserted PR comment, or empty when no comment was posted.