karolmckgh/Multi-Model Code Review

Run code reviews across multiple AI models via GitHub Copilot SDK, then synthesize findings into a consensus table posted as a PR comment. Supports per-contributor PAT tokens for Copilot premium request attribution.

Trust Signals

Scorecard Score: not yet scored
Maintenance Recency: Stale
License: None

Inputs

name	description	required	default
models	JSON object mapping display names to Copilot model IDs. Example: '{"GPT-5.3 Codex": "gpt-5.3-codex", "Claude Sonnet 4.6": "claude-sonnet-4.6"}'	no	{"GPT-5.3 Codex": "gpt-5.3-codex", "Claude Sonnet 4.6": "claude-sonnet-4.6", "Gemini 3.1 Pro": "gemini-3.1-pro"}
synthesizer-model	Copilot model ID used to merge findings into a consensus table.	no	gpt-5.3-codex
max-diff-size	Maximum diff size in characters. Larger diffs are chunked by file.	no	30000
review-focus	Comma-separated review focus areas. Options: security, bugs, performance, error-handling, all	no	all
custom-prompt	Override the default review prompt entirely. If set, review-focus is ignored.	no	""
copilot-token	REQUIRED — GitHub PAT with 'copilot' scope from an account with an active Copilot subscription. Set as a repo/org secret (e.g. COPILOT_GITHUB_TOKEN) and pass it here. See README for step-by-step setup instructions.	no	""
copilot-token-map	JSON object mapping GitHub usernames to secret names holding their Copilot PAT. Example: '{"octocat": "COPILOT_PAT_OCTOCAT", "monalisa": "COPILOT_PAT_MONALISA"}' The action resolves the PR author's token automatically.	no	{}
max-retries	Maximum number of retries per model on transient failures.	no	2
post-comment	Whether to post the review as a PR comment.	no	true
session-timeout	Timeout in seconds for each model session (default 300 = 5 min).	no	300

Outputs

name	description
review-body	The full markdown review body (useful if post-comment is false).
findings-json	Raw JSON findings from all models.