gulersevil/action-gh-model-prompt

Render a prompt file, call GitHub Models, and return JSON outputs.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
modelModel id (e.g., openai/gpt-4o-mini)noopenai/gpt-4o-mini
prompt_pathPath to the prompt fileyes
placeholders_jsonJSON object whose keys map to {{KEY}} in the promptyes{}
systemSystem message; default enforces strict JSONnoOutput STRICT JSON only.
response_formatjson_object | textnojson_object
api_urlOverride inference endpointnohttps://models.github.ai/inference/chat/completions
timeout_msHTTP timeout in msno120000
strip_hash_commentsRemove lines starting with '# ' before sendingnotrue
request_debugEcho rendered prompt/payload to logs (careful in public repos)nofalse
fail_on_invalid_jsonFail if response isn't valid JSON (when response_format=json_object)notrue
jq_pickOptional JSONPath-like (jq) selector, e.g. .verdict.release_riskno""
pick_aggregationHow to aggregate picked values across batches: none | first | join | worst_severitynonone
tokenGitHub token; defaults to env.GITHUB_TOKENno
batch_sizeIf >0, split FILE_LIST into batches and call model per batchno0
namedescription
raw_responseRaw HTTP response from the models API
message_contentchoices[0].message.content
textPlain text content (single) or joined text (batched)
jsonParsed JSON (pretty-printed) when available
pickedValue of jq_pick if provided
picked_markdownMarkdown list/sections derived from picked value(s)
reportMarkdown report derived from JSON content