Configuration Files
next.config.mjs
This file configures Next.js build options and image handling.
- TypeScript: ignores build errors
- Images: disables built‐in optimization
/** @type {import('next').NextConfig} */
const nextConfig = {
typescript: { ignoreBuildErrors: true },
images: { unoptimized: true },
}
export default nextConfig
postcss.config.mjs
Defines PostCSS plugins for Tailwind CSS integration.
/** @type {import('postcss-load-config').Config} */
const config = {
plugins: {
'@tailwindcss/postcss': {},
},
}
export default config
package.json
Specifies project metadata, scripts, and dependencies.
| Property | Description |
|---|---|
| name | Package name (“my-project”) |
| version | Version number (0.1.0) |
| private | Prevents accidental publish |
Scripts
| Command | Runs |
|---|---|
| dev | next dev |
| build | next build |
| start | next start |
| lint | eslint . |
Dependencies & DevDependencies
- Frameworks & UI:
next,react,tailwindcss,@radix-ui,lucide-react - Form & Validation:
react-hook-form,zod - Analytics & Export:
@vercel/analytics,vaul - Dev Tools:
typescript,postcss,autoprefixer
React Components
theme-provider.tsx
Wraps next-themes provider to enable dark/light theme switching.
'use client'
import { ThemeProvider as NextThemesProvider, type ThemeProviderProps } from 'next-themes'
export function ThemeProvider({ children, ...props }: ThemeProviderProps) {
return <NextThemesProvider {...props}>{children}</NextThemesProvider>
}
pipeline-progress.tsx
Visualizes pipeline stages with icons and progress state.
- Props:
currentStage: PipelineStageisRunning: boolean- Behavior:
- Marks stages as complete, current, or pending
- Renders connectors and animated pulse on current
import { PipelineStage } from "@/lib/types"
import { Search, BookOpen, ShieldCheck, FlaskConical, Lightbulb, Users, Download } from "lucide-react"
const STAGES = [
{ key: PipelineStage.INPUT, label: "Input", icon: Search },
{ key: PipelineStage.RESEARCH, label: "Research", icon: BookOpen },
{ key: PipelineStage.VERIFICATION, label: "Verify", icon: ShieldCheck },
{ key: PipelineStage.SCIENTIFIC_METHOD, label: "Science", icon: FlaskConical },
{ key: PipelineStage.BRAINSTORMING, label: "Brainstorm", icon: Lightbulb },
{ key: PipelineStage.PMOPS, label: "P.M.O.P.S.", icon: Users },
{ key: PipelineStage.EXPORT, label: "Export", icon: Download },
]
// ... getStageIndex, PipelineProgress component ...
stats-summary.tsx
Displays key metrics from a research session as cards.
- Props:
session: ResearchSession | null - Stats:
- Articles Analyzed
- Verified Truths
- Flagged Poisoned
- Avg. Credibility
- Ideas Generated
- Votes Cast
import { Card, CardContent } from "@/components/ui/card"
import { BookOpen, CheckCircle2, ShieldAlert, FlaskConical, Lightbulb, Vote } from "lucide-react"
export function StatsSummary({ session }: { session: ResearchSession | null }) {
if (!session) return null
// compute values…
return (
<div className="grid …">
{/* map over stats array */}
</div>
)
}
research-dashboard.tsx
Orchestrates the entire UI: input, progress, error, agents, stats, and result panels in tabs.
- Manages
userRequest,session,exportText, pipeline state, agent statuses, and errors - Calls
/api/pipelineon “Run Pipeline” - Offers Export and Reset buttons
- Renders panels: Verification, Scientific, Brainstorm, P.M.O.P.S., Verified Truths, Code Audit, Export Preview
import { useState, useCallback, useRef, useEffect } from "react"
import { Card, Button, Textarea, Tabs, Separator } from "@/components/ui"
import { PipelineProgress } from "@/components/pipeline-progress"
import { AgentCards } from "@/components/agent-cards"
import { VerificationPanel } from "@/components/verification-panel"
import { ScientificPanel } from "@/components/scientific-panel"
import { BrainstormPanel } from "@/components/brainstorm-panel"
import { PMOPSPanel } from "@/components/pmops-panel"
import { VerifiedTruthsPanel } from "@/components/verified-truths-panel"
import { CodeAuditPanel } from "@/components/code-audit-panel"
import { StatsSummary } from "@/components/stats-summary"
// … full component …
agent-cards.tsx
Shows each agent’s name, status badge, description, current task, and skills.
- Props:
agents: Record<AgentRole, AgentDefinition> - Uses
AGENT_ICONSlookup
import { Card, CardHeader, CardTitle, CardContent } from "@/components/ui/card"
import { Badge } from "@/components/ui/badge"
import { AgentRole, AgentDefinition } from "@/lib/types"
// getStatusColor, getStatusLabel helpers
export function AgentCards({ agents }: { agents: Record<AgentRole, AgentDefinition> }) { /* … */ }
verification-panel.tsx
Lists verification results per article.
- Props:
verifications: Verification[]articles: Article[]- Renders badge via
getStatusBadge - Shows credibility progress and individual checks
- Displays flagged issues if any
import { Progress, ScrollArea, Badge, Card, CardHeader, CardContent, CardTitle } from "@/components/ui"
import { VerificationStatus } from "@/lib/types"
import { CheckCircle2, XCircle, AlertTriangle, Shield } from "lucide-react"
// getStatusBadge helper
export function VerificationPanel({ verifications, articles }: { verifications: Verification[], articles: Article[] }) { /* … */ }
scientific-panel.tsx
Presents scientific validation for each article.
- Props:
validations: ScientificValidation[]articles: Article[]- Uses
ValidationRowsubcomponent for each check - Shows overall verdict badge and extracted insights
import { Badge, Progress, ScrollArea, Card, CardHeader, CardContent, CardTitle } from "@/components/ui"
import { Brain, Repeat2, Gauge, TrendingUp, Eye, ShieldAlert } from "lucide-react"
// getVerdictBadge, ValidationRow, ScientificPanel …
brainstorm-panel.tsx
Displays generated ideas and novel procedures.
- Props:
results: BrainstormResult[] - Renders idea cards with feasibility & innovation progress bars
- Lists Novel Procedures, Method Extensions, Cross-Domain Applications
import { Lightbulb, FlaskConical, ArrowRight, Globe } from "lucide-react"
import { Card, Badge, Progress, ScrollArea } from "@/components/ui"
// BrainstormPanel component …
pmops-panel.tsx
Shows P.M.O.P.S. discussion:
- Props:
discussions: PMOPSDiscussion[] - Highlights Winning Proposal, All Proposals, Voting Results, and Chat Log
- Uses icons:
Trophy,Users,Vote,MessageSquare
import { Card, Badge, ScrollArea, Separator } from "@/components/ui"
import { Vote, MessageSquare, Trophy, Users } from "lucide-react"
// PMOPSPanel component …
verified-truths-panel.tsx
Lists final verified truths.
- Props:
truths: VerifiedTruth[] - Shows title, content, credibility progress, methods, sources, and insights
import { CheckCircle2, BookOpen, FlaskConical } from "lucide-react"
import { Card, Badge, Progress, ScrollArea } from "@/components/ui"
// VerifiedTruthsPanel …
code-audit-panel.tsx
Presents code quality metrics and audited issues.
- Props:
session: ResearchSession - Computes metrics via
computeMetrics - Renders:
- Code Quality Health with score and key tiles
- Code Audit Report listing
AUDITED_ISSUES - Pipeline Integrity Checks (hardcoded passed checks)
import { ShieldCheck, AlertTriangle, Bug, CheckCircle2, Activity, FileSearch } from "lucide-react"
import { AUDITED_ISSUES, generateQualityReport } from "@/lib/code-monitor"
import { Card, Badge, Progress, ScrollArea } from "@/components/ui"
// CodeAuditPanel component …
Frontend Utilities & Library
utils.ts
Utility cn() to merge Tailwind CSS classes safely.
import { clsx, type ClassValue } from 'clsx'
import { twMerge } from 'tailwind-merge'
export function cn(...inputs: ClassValue[]) {
return twMerge(clsx(inputs))
}
code-monitor.ts
Tracks pipeline events and generates a quality report.
- Interfaces:
MonitorEvent,QualityReport,PipelineMetrics,CodeIssue - Constants:
AUDITED_ISSUES(list of resolved code issues) - Functions:
createEventLog()recordEvent(log, type, message, stage?, data?)generateQualityReport(events, metrics)
export interface MonitorEvent { /* … */ }
export const AUDITED_ISSUES: CodeIssue[] = [ /* … */ ]
export function createEventLog(): MonitorEvent[] { return [] }
export function recordEvent( log: MonitorEvent[], type: MonitorEvent["type"], message: string, stage?: string, data?: Record<string, unknown> ): MonitorEvent[] { /* … */ }
export function generateQualityReport(events: MonitorEvent[], metrics: PipelineMetrics): QualityReport { /* … */ }
export.ts
Builds a structured export document and renders it as plain text.
- Function:
generateExportDocument(session): ExportDocument - Section Builders:
buildExecutiveSummarybuildTruthSectionbuildMethodologySectionbuildVerificationSectionbuildScientificSectionbuildPMOPSSectionbuildRejectedDataSectionbuildAnalyticsSection- Renderer:
renderExportToText(doc): string
export function generateExportDocument(session: ResearchSession): ExportDocument { /* … */ }
export function renderExportToText(doc: ExportDocument): string { /* … */ }
types.ts (TS)
Defines core types for the frontend library.
- Enums:
PipelineStage,VerificationStatus,CredibilityLevel,AgentRole - Interfaces:
AgentDefinition,ResearchSession,Topic,Article,CitationVerification,VerificationCheckScientificValidation,ValidationResultBrainstormResult,BrainstormIdeaPMOPSDiscussion,PMOPSProposal,ChatMessage,VoteResultVerifiedTruthExportDocument,ExportSectionPipelineRequest,PipelineResponse
export enum PipelineStage { INPUT = "input", /* … */ }
export interface ResearchSession { /* … */ }
export interface Verification { /* … */ }
// … other types …
agents.ts
Defines AI agent metadata and system prompts.
- Constants:
AGENT_DEFINITIONS: basic info perAgentRoleAGENT_SYSTEM_PROMPTS: system‐level instructions- Function:
createAgentTeam()returns fresh idle copies
export const AGENT_DEFINITIONS: Record<AgentRole, AgentDefinition> = { /* … */ }
export const AGENT_SYSTEM_PROMPTS: Record<AgentRole, string> = { /* … */ }
export function createAgentTeam(): Record<AgentRole, AgentDefinition> { /* … */ }
poison-shield.ts
Implements Data Poison Shield for verification.
- Prompt:
POISON_SHIELD_SYSTEM_PROMPT - Interface:
PoisonIndicator - Patterns:
POISON_INDICATORSarray of regex checks - Functions:
scanForPoisonIndicators(text)→ matchescalculatePoisonRiskScore(indicators)→ risk 0-100
export const POISON_SHIELD_SYSTEM_PROMPT = `…`
export interface PoisonIndicator { pattern: RegExp; type: string; /* … */ }
export function scanForPoisonIndicators(text: string): { indicator: PoisonIndicator; match: string; position: number }[] { /* … */ }
export function calculatePoisonRiskScore(indicators: { indicator: PoisonIndicator }[]): number { /* … */ }
pipeline.ts (TS)
Pure JS implementation of the research pipeline.
- Helpers:
generateId,now - Metadata:
SOURCE_METAfor source labeling/snippets - Stage 1:
parseUserRequest→Topic[] - Stage 2:
conductResearch→Article[] - Stage 3:
verifyArticle→Verification - Stage 4:
applyScientificMethod→ScientificValidation - Stage 5:
brainstormFromVerifiedData→BrainstormResult - Stage 6:
runPMOPS→PMOPSDiscussion - Stage 7:
compileVerifiedTruths→VerifiedTruth[] - Orchestrator:
runFullPipeline(userRequest)→ completes all stages
export function runFullPipeline(userRequest: string): ResearchSession { /* … */ }
flowchart TD
A[User Request] --> B[parseUserRequest]
B --> C[conductResearch]
C --> D[verifyArticle]
D --> E[applyScientificMethod]
E --> F[brainstormFromVerifiedData]
F --> G[runPMOPS]
G --> H[compileVerifiedTruths]
H --> I[Final Session]
Python Library Modules
init.py
Package entrypoint for ai_research_verifier.
- Sets
__version__,__author__ - Exposes
run_pipelineandResearchSession
__version__ = "1.0.0"
__author__ = "AI Research Verification Team"
from ai_research_verifier.pipeline import run_pipeline
from ai_research_verifier.types import ResearchSession
types.py
Core type definitions via dataclasses and enums.
- Enums:
PipelineStage,VerificationStatus,CredibilityLevel,AgentRole - Utilities:
generate_id,now_iso - Dataclasses:
Citation,Topic,ArticleVerificationCheck,VerificationValidationResult,ScientificValidationBrainstormIdea,BrainstormResultChatMessage,PMOPSProposal,VoteResult,PMOPSDiscussionVerifiedTruth,ResearchSession
@dataclass
class Article:
id: str = field(default_factory=lambda: generate_id("article"))
title: str = ""
# … other fields …
agents.py
Defines agent roles, display names, skills, system prompts, and PMOPS approaches.
- Constants:
AGENT_DISPLAY_NAMESAGENT_SKILLSAGENT_SYSTEM_PROMPTSAGENT_APPROACHES
AGENT_SYSTEM_PROMPTS = {
AgentRole.RESEARCH_ANALYST: "You are the Research Analyst Agent…",
# …
}
poison_shield.py
Data Poison Shield logic for Python agents.
- Prompt:
POISON_SHIELD_SYSTEM_PROMPT - Dataclass:
PoisonIndicator,PoisonScanResult - Constants:
POISON_INDICATORS(regex patterns) - Functions:
scan_for_poison(text)calculate_poison_risk(results)
def scan_for_poison(text: str) -> list[PoisonScanResult]:
# pure function, no side-effects
export.py
Generates and saves .txt exports from a ResearchSession.
- Function:
export_to_text(session) - File Writer:
save_export(session, filepath)
def export_to_text(session: ResearchSession) -> str:
# builds lines list, joins with "\n"
pipeline.py
Orchestrates the multi-agent pipeline in Python.
- Data:
SOURCE_META,CHECK_RULES - Stages:
parse_user_requestconduct_researchverify_articleapply_scientific_methodbrainstormrun_pmopscompile_verified_truths- Orchestrator:
run_pipeline(user_request)
def run_pipeline(user_request: str) -> ResearchSession:
# sequential, no loops or recursion
flowchart TD
I[Input] --> R[Research]
R --> V[Verification]
V --> S[Scientific Method]
S --> B[Brainstorm]
B --> P[P.M.O.P.S.]
P --> E[Compile Truths]
E --> O[Output Session]
Each file implements a distinct part of the full AI Research Verification System, from configuration and UI on the frontend to multi-agent orchestration and export logic in both JavaScript/TypeScript and Python.