ReverieHacks (Aug 2026)
NodeForge.ai
4-Node Agentic Code-Review & Security Pipeline
Agentic code-review and security pipeline built for ReverieHacks (Aug 2026), orchestrating 4 inspectable AI nodes (Human Input, Query Expansion, Execution & Verification, Output Sanitization) with side-by-side baseline scoring, PR review, SARIF/JSON export, and live Mermaid flowcharts.
System ID
06
NodeForge.ai
AI Nodes
M-014 Stages
Inspectable Pipeline
Report Export
M-02SARIF / JSON
PR Review Ready
Visualization
M-03Mermaid
Live Flowcharts
Model Routing
M-04OpenRouter
+ Featherless / Vercel
Section 01
Problem Statement
Opaque single-prompt AI code reviewers provide no visibility into intermediate reasoning, baseline comparisons, or machine-readable security findings.
Section 02
System Constraints
- >Expose every reasoning stage through 4 inspectable pipeline nodes
- >Deliver side-by-side baseline scoring alongside PR review diagnostics
- >Emit standard SARIF and JSON outputs plus live Mermaid architecture graphs
Section 03
System Architecture & Dataflow
1. Human Input
2. Query Expansion
3. Execution & Verification
4. Output Sanitization (SARIF)
devhuang1@nodeforge:~$ nodeforge inspect
[NodeForge.ai // ReverieHacks (Aug 2026) • MIT]
4-Node Agentic Code-Review & Security Pipeline:
[01] Human Input -> Scope PR diff, policy targets & review constraints
[02] Query Expansion -> Expand checks & codebase context (OpenRouter / Featherless)
[03] Execution & Verification -> Side-by-side baseline scoring & PR review checks
[04] Output Sanitization -> Deterministic findings + SARIF / JSON export + Mermaid
PASS: All 4 inspectable nodes active (export ready: SARIF / JSON)
Section 04
Key Engineering Decisions
- 01Structured the review workflow into 4 inspectable nodes: Human Input, Query Expansion, Execution & Verification, and Output Sanitization.
- 02Added side-by-side baseline scoring, PR review workflows, and live Mermaid flowcharts in Next.js.
- 03Supported SARIF/JSON security export backed by OpenRouter and Featherless model routing.