Code Review Pipeline

PR Review Pipeline

Orchestrates the full pull request review: fetch, analyse, check style, scan security, summarise, decide

Code Analysis

Analyses code for patterns, complexity, and potential bugs across common languages

Defang Content

Neutralises potentially dangerous content in code snippets, URLs, and commands before display or processing

Format Conversion

Converts content between output formats: markdown, plain text, email HTML, JSON, CSV, presentation outline

Security Scanning

Identifies security vulnerabilities, insecure patterns, and dependency risks in code

Structured Data Extraction

Extracts structured fields and data points from unstructured text

Style Checking

Validates code against style guidelines, naming conventions, and formatting standards

Approval Decision

Makes an approve or request-changes decision with structured reasoning

Extract Structured Data

Extracts structured fields and key-value pairs from unstructured text

Inline Comments

Generates precise inline code comments for specific issues found during review

Review Summary

Generates a structured, readable summary of all review findings for a pull request

Security Report

Produces a detailed security findings report with risk ratings and remediation guidance

Style Violations

Lists style guide violations with specific fix suggestions and auto-fix indicators

GitHub MCP

GitHub Model Context Protocol service for pull request access and review posting

LLM Service

Language model service for analysis, synthesis, and document generation

Ollama Local

Local LLM inference via Ollama for private, offline code review

Review Standards

Coding standards and review criteria used as the baseline for all analysis passes