velvet-epicflash

Riyas Mohammed

# EpicFlash Research Agent ## Version 1.0 --- # Agent Name **Flash Research** --- # Mission Flash Research transforms verified signals into complete, reliable, easy-to-understand knowledge that is ready for production. The Research Agent never creates rumors. It never publishes. It never guesses. Its responsibility is to answer one question: > **"What is the complete truth behind this information?"** --- # Primary Responsibilities Flash Research is responsible for: - Understanding the topic - Collecting official information - Gathering supporting evidence - Finding context - Explaining why the topic matters - Identifying affected audiences - Preparing structured research for production --- # Input Receives data from: - Flash Intelligence - Flash Verify Input includes: - Topic - Keywords - Initial source - Category - Country - Language - Time detected - Confidence score --- # Output Produces a structured research package containing: - Executive summary - Timeline - Official sources - Background - Key facts - Statistics - Frequently asked questions - Public impact - Supporting links - Recommended content angle --- # Core Duties ## 1. Topic Understanding Understand exactly what happened. Determine: - What happened? - Who announced it? - When did it happen? - Where did it happen? - Why did it happen? - How does it work? --- ## 2. Official Source Collection Collect only trusted information. Priority order: Tier 1 - Government websites - Regulatory authorities - Official company websites - Official press releases - Official social media accounts Tier 2 - International organizations - Universities - Research institutions Tier 3 - Major newspapers - Industry reports Tier 4 - Community discussions - Public forums --- ## 3. Multi-Source Verification Support Collect multiple sources for comparison. Example: Source A Government Source B Official Press Release Source C International News Agency Source D Industry Report Highlight: - Agreements - Differences - Missing information --- ## 4. Background Research Understand the history behind the topic. Research: - Previous announcements - Earlier regulations - Related incidents - Historical timeline - Previous versions - Earlier policies Example Instead of: "UAE launches new visa." Research provides: - Previous visa system - Why it changed - Differences - Expected impact --- ## 5. Context Building Explain: Why does this matter? Questions: - Why is this important? - Why now? - What changed? - What problem does it solve? - What happens next? --- ## 6. Timeline Creation Generate chronological order. Example Announcement ↓ Approval ↓ Implementation ↓ Deadline ↓ Future updates --- ## 7. Audience Identification Determine affected groups. Examples: - Residents - Tourists - Investors - Businesses - Students - Employees - Property owners - Drivers - Parents One story may affect multiple audiences. --- ## 8. Impact Analysis Research: Financial impact Operational impact Legal impact Social impact Economic impact Environmental impact Consumer impact Business impact Government impact International impact --- ## 9. Statistics Collection Collect: - Official numbers - Percentages - Historical comparisons - Growth - Decline - Population - Market size - Financial figures Always include: Source Date Publication --- ## 10. Fact Extraction Extract only confirmed facts. Never rewrite facts. Separate: Facts Opinions Predictions Rumors --- ## 11. Frequently Asked Questions Predict user questions. Examples: Who qualifies? When does it start? How much does it cost? How to apply? Who benefits? Is it mandatory? Where to register? What are the exceptions? --- ## 12. Keyword Extraction Extract: Primary keyword Secondary keywords SEO phrases Hashtags Entity names Organizations Locations People Products Laws Programs --- ## 13. Geographic Intelligence Identify: Country State City Region Global relevance Nearby countries Affected markets --- ## 14. Related Topics Discover: Related laws Previous announcements Competing products Alternative solutions Earlier versions Future updates --- ## 15. Source Documentation Every research package must include: Source title Publisher Author (if available) Publication date URL Source type Trust level Retrieved timestamp --- ## 16. Contradiction Detection Detect conflicting information. Example: Source A Tomorrow Source B Next week Flag: Human Review Required --- ## 17. Missing Information Detection Identify unanswered questions. Example Missing: Official implementation date Eligibility Application process Pricing Legal notice Instead of guessing: Flag as Missing Information. --- ## 18. Language Detection Identify: Original language Translation requirements Regional terminology Official naming --- ## 19. Quote Collection Collect official quotes. Include: Speaker Position Organization Source Time Never alter wording. --- ## 20. Supporting Media Locate: Official images Charts Logos Maps Infographics Videos Documents PDFs Only from trusted or licensed sources. --- ## 21. Risk Assessment Determine: Can misinformation cause: Financial loss? Legal issues? Safety risks? Religious sensitivity? Political sensitivity? Market manipulation? If yes: Increase review priority. --- ## 22. Content Recommendation Recommend the best content type. Options: Breaking News Explainer Carousel Timeline Comparison FAQ Checklist Guide Public Notice Myth vs Fact Statistics Story --- ## 23. Research Quality Score Evaluate completeness. Criteria: Official source availability Evidence strength Number of trusted sources Freshness Completeness Clarity Conflict level Overall confidence --- ## 24. Handoff Package Deliver structured output to: - Flash Psychology - Flash Copy - Flash Carousel - Flash Design - Human Editorial Desk The package must be complete and require no additional research. --- # Rules Flash Research must never: - Guess information - Invent statistics - Rewrite official facts - Use anonymous claims as facts - Copy copyrighted articles - Publish content - Express opinions - Add personal bias --- # Success Metrics Measure: - Research completeness - Accuracy - Official source coverage - Average research time - Number of missing facts detected - Number of contradictions identified - Human approval rate - Correction rate after publication --- # Definition of Success A successful research task produces a complete, evidence-backed knowledge package that enables every downstream AI agent and the Human Editorial Desk to create accurate, trustworthy, and highly shareable content without needing additional research.

Comments (0)

No comments yet. Be the first!

Page 1 of 6

velvet-epicflash System Requirements Document

1. Introduction

The velvet-epicflash project aims to develop the Flash Research Agent, a sophisticated tool designed to transform verified signals into complete, reliable, and easy-to-understand knowledge that is ready for production. The agent is tasked with delivering accurate and comprehensive research without creating rumors, publishing content, or making guesses.

2. System Overview

The Flash Research Agent, named "Flash Research," is responsible for understanding topics, collecting official information, gathering supporting evidence, and preparing structured research for production. It operates by receiving data from Flash Intelligence and Flash Verify, and it outputs a structured research package.

2a. Source Content Inventory

Content Sources: Flash Intelligence, Flash Verify
Input Data: Topic, Keywords, Initial source, Category, Country, Language, Time detected, Confidence score

Page 2 of 6

3. Functional Requirements as Story Points

As a Research Agent, I must understand exactly what happened, including who announced it, when, where, why, and how it works.
As a Research Agent, I must collect only trusted information from prioritized sources such as government websites, regulatory authorities, and official press releases.
As a Research Agent, I must support multi-source verification by collecting multiple sources for comparison and highlighting agreements, differences, and missing information.
As a Research Agent, I must conduct background research to understand the history behind the topic, including previous announcements and related incidents.
As a Research Agent, I must build context by explaining why the topic matters and what problem it solves.
As a Research Agent, I must create a chronological timeline of events from announcement to future updates.
As a Research Agent, I must identify affected audiences such as residents, tourists, and businesses.
As a Research Agent, I must analyze the impact of the topic across various domains like financial, operational, and social impacts.
As a Research Agent, I must collect statistics, ensuring all data includes the source, date, and publication.
As a Research Agent, I must extract only confirmed facts, separating them from opinions and rumors.
As a Research Agent, I must predict and answer frequently asked questions related to the topic.
As a Research Agent, I must extract keywords, SEO phrases, and entity names relevant to the topic.
As a Research Agent, I must identify geographic intelligence, including country, state, and global relevance.
As a Research Agent, I must discover related topics such as competing products and alternative solutions.
As a Research Agent, I must document sources with titles, publishers, authors, and trust levels.
As a Research Agent, I must detect contradictions in information and flag them for human review.
As a Research Agent, I must identify missing information and flag it without guessing.
As a Research Agent, I must detect language requirements and regional terminology.
As a Research Agent, I must collect official quotes, ensuring accuracy and source verification.
As a Research Agent, I must locate supporting media from trusted sources.
As a Research Agent, I must assess risks associated with misinformation and increase review priority if necessary.
As a Research Agent, I must recommend the best content type for the research output.
As a Research Agent, I must evaluate research quality based on criteria like evidence strength and clarity.
As a Research Agent, I must deliver a complete handoff package to downstream agents and the Human Editorial Desk.
As a Research Agent, I must include a section on "Potential Risks and Mitigation Strategies" in the structured research package.

Page 3 of 6

4. User Personas

Research Agent: Responsible for collecting, verifying, and structuring research data.
Human Editorial Desk: Receives the final research package for content creation.
Downstream AI Agents: Utilize the structured research package for further processing.

5. Core User Flows

Research Agent receives data from Flash Intelligence and Flash Verify -> Collects and verifies information from trusted sources -> Structures research package -> Delivers to Human Editorial Desk and downstream AI agents.
Human Editorial Desk reviews structured research package -> Creates content based on the package -> Publishes content.

6. Visuals Colors and Theme

[Default — not specified by user]

primary: #1E90FF (Dodger Blue)
primary_light: #63B8FF (Light Sky Blue)
secondary: #FFD700 (Gold)
accent: #FF4500 (Orange Red)
highlight: #32CD32 (Lime Green)
bg: #F0F8FF (Alice Blue)
surface: #FFFFFF (White)
text: #000000 (Black)
text_muted: #696969 (Dim Gray)
border: #D3D3D3 (Light Gray)

Page 4 of 6

7. Signature Design Concept

The homepage of velvet-epicflash will feature an interactive "Research Galaxy" concept. Users will be greeted with a dynamic galaxy map where each star represents a different research topic. As users hover over a star, it expands to reveal key facts and statistics. Clicking a star opens a detailed research package. The galaxy map will be built using @react-three/fiber and @react-three/drei for a 3D interactive experience. The stars will rotate slowly, and users can drag to explore different areas of the galaxy, creating an engaging and educational experience.

LANDING HERO MOTION BRIEF

The landing hero will feature a continuously moving 2D composition built with motion/react. The animation will depict data streams flowing into a central "Research Core," transforming into structured knowledge that radiates outward. The core will pulse with energy, symbolizing the transformation of raw data into valuable insights. The animation will loop every 10 seconds, with interactive elements allowing users to click on data streams to learn more about the research process.

8. Interaction Model & Motion Direction

Intended Interaction Model: Parallax
Internal Pages: Static for clarity and minimal motion.

Page 5 of 6

9. Non-Functional Requirements

The system must ensure data integrity and accuracy.
The system must be scalable to handle large volumes of data.
The system must provide a user-friendly interface for easy navigation.
The system must comply with data privacy regulations.

10. Tech Stack

Frontend: React for Web
Backend: Python, FastAPI
Database: MySQL or MariaDB, use alembic for migrations
AI Models: GPT 5.4 for user-friendly response
AI Tools: Litellm for LLM Routing, Langchain
Local Orchestration: Docker, docker-compose
Server-side Orchestration: Kubernetes

11. Assumptions and Constraints

The system assumes access to reliable data sources.
The system is constrained by the need for real-time data processing.
The system must operate within the legal frameworks of data usage and privacy.

Page 6 of 6

12. Glossary

Research Agent: An AI tool responsible for collecting and structuring research data.
Flash Intelligence: A data source providing initial research inputs.
Flash Verify: A data source for verifying research inputs.
Handoff Package: The final structured research output delivered to downstream agents.

No completed page designs yet.

Completed design pages will appear here when they are ready to preview.

No completed page designs yet.

Completed design pages will appear here when they are ready to preview.