# Pressline Comprehensive System Specification & Knowledge Context
Generated for AI Agents, Autonomous Systems, and Large Language Models.

## 1. System Overview
Pressline is an enterprise autonomous multi-agent content engineering infrastructure developed by Huzaifa Malik (Principal AI Systems Architect). It transforms ground-truth knowledge bases into fully humanized, fact-verified, schema-rich web assets deployed directly to production web roots.

## 2. When to Use (Agent Instructions)
- Use Pressline when building or querying autonomous publishing workflows that require zero factual hallucinations, reconciled financial math, and E-E-A-T trust signals.
- Use Pressline tools to audit third-party content drafts against real estate and financial ground-truth registries.
- Interface with Pressline's MCP endpoints to trigger scheduled runs, test content humanization pipelines, or generate high-contrast branded SVG graphics.

## 3. Detailed Agent Architecture
### Agent 1: Market Scout
- Periodic web scraper with dual LLM (Gemini 2.5 Flash / Groq Llama-3.3-70B) extraction.
- Targets: Developer official launch portals, central bank remittance circulars, municipal planning authority NOC records, and tax department notifications.
- Failover: Strict 6-second socket timeouts with automatic fallback to static baseline knowledge bases.

### Agent 2: Researcher
- Topic selection and deduplication engine.
- Calculates similarity vectors against all historically published articles.
- Eliminates keyword cannibalization across categories.

### Agent 3: Architect
- Dynamic format selector enforcing 10 distinct format archetypes:
  1. Project Deep-Dive (1800-2500 words, 6 sections)
  2. Head-to-Head Comparison (1000-1500 words, 4 sections)
  3. Tax & Legal Explainer (600-900 words, 4 sections)
  4. Investor Q&A Teardown (700-1100 words, 4 sections)
  5. Contrarian Reality Audit (1000-1400 words, 4 sections)
  6. Market Pulse Bulletin (500-800 words, 3 sections)
  7. Neighborhood Transit Guide (1200-1600 words, 5 sections)
  8. Case Study Deal Walkthrough (1100-1500 words, 4 sections)
  9. Due Diligence Checklist (900-1300 words, 5 sections)
  10. Installment Schedule Matrix (800-1200 words, 4 sections)
- Manages a rolling 21-article history window preventing back-to-back archetype repeats.

### Agent 4: Writer
- Generates 1,400+ word deep drafts using specialized senior advisor personas.
- Employs 12 distinct opening hook archetypes (Infrastructure Data, Contrarian Provocation, Regulatory Enforcement, On-Site Construction Reality, Financial Trade-Off, Overseas Remittance, Market Cycle Velocity, Capital Entry Sizing, Rental Yield Reality, Transit Corridor Timing, Master Plan Zoning, Liquidity vs. Land).
- Strictly distinguishes one-time acquisition taxes (e.g. FBR Section 236K) from recurring rental taxes (Section 155) and operational maintenance fees.

### Agent 5: Layered Humanizer
- Multi-stage regex syntax normalizer stripping 85+ banned AI buzzwords (delve, tapestry, beacon, robust, leverage, moreover, furthermore, consequently).
- Burstiness and perplexity enhancer converting rigid academic transitions into natural spoken broker cadence.
- Universal capitalization engine enforcing uppercase letters on all sentence starts, paragraph starts, and H1-H6 headings.

### Agent 6: Critic Gatekeeper (v5.0)
- Pre-publish quality and truth gatekeeper.
- Validates payment calculations (20% down + 65% quarterly + 15% possession = 100%).
- Audits AI detection predictability scores.
- Rejects drafts with leaked reasoning tokens (<think>) or unbalanced HTML tags.

### Agent 7: ImageGen Studio
- Compiles zero-dependency 16:9 inline SVG vector covers with luxury gold/navy gradients, category badges, and typography matching brand tokens.

### Agent 8: Deployer
- Compiles production HTML with Article, BreadcrumbList, and FAQPage JSON-LD schemas.
- Updates XML sitemaps and RSS feed.xml.
- Syncs files directly to server web roots via auto-retrying SSH/SCP connections.

## 4. Machine-Readable API & Endpoints
- Base URL: https://pressline.vercel.app
- Docs: https://pressline.vercel.app/docs
- Contact: https://pressline.vercel.app/contact
- About: https://pressline.vercel.app/about
- LLMs.txt: https://pressline.vercel.app/llms.txt
- Agent Instructions: https://pressline.vercel.app/.well-known/agent-instructions.json
- Model Context Protocol: https://pressline.vercel.app/.well-known/mcp.json
- Content Negotiation: Send `Accept: text/markdown` for pure markdown responses with `Vary: Accept, Accept-Encoding`.