AI Project Analyzer — Feature Documentation (EN)

Overview

AI Project Analyzer helps sellers create listings faster by analyzing a project link (website or repository) and auto‑populating the create‑listing form on /create-listing. Users can review and edit all fields before saving a draft or submitting for review.

Current Implementation:

• Mistral AI (blue theme) - Direct API integration with Mistral Large/Small
• n8n AI Agent (green theme) - Workflow automation via n8n webhook (deprecated in favor of Mistral)

The system uses Mistral AI as the primary analyzer with robust error handling and JSON parsing recovery mechanisms.

Key Capabilities

• Mistral AI: Direct AI analysis powered by Mistral Large/Small with automatic model fallback
• Robust Error Handling: JSON parsing recovery with 5-strategy truncation and fallback mechanisms
• Rate Limit Management: User-friendly 429 error messages with retry guidance
• Works with public websites and code repositories (GitHub/GitLab)
• Secure invocation via Supabase Edge Function with JWT verification
• Attractive, responsive UI with AI-futuristic loading animations and clear states
• Safe merge logic: preserves any user‑typed data, only fills missing fields
• Financial Defaults: Generates realistic market averages when specific data unavailable

Data Fields Auto‑Populated

The analyzer aims to fill the following groups (where information is publicly discoverable):

• Basic Information
  • title/title_en/title_ro
  • category (normalized slug from DB listing_categories)
  • website_url
• Overview
  • business_model (markdown)
  • description (summary/overview)
  • established_date (YYYY‑MM‑DD)
  • number_of_customers (integer)
• Financial Data
  • monthly_revenue (USD)
  • monthly_profit (USD)
  • price (from AI sale_price)
  • allow_negotiation (boolean)
  • price_reasoning (markdown)
• Traffic & Audience
  • monthly_traffic
  • target_audience (markdown)
• Business Details (markdown; structured accordions)
  • Introduction: business_description, business_model, key_strengths, team_founder
  • Business Details: included_assets, operations_time_required, technology_stack, financial_performance, expenses, growth_opportunities
  • Conclusion & Motivation: reason_for_selling, benefits_for_buyer, risks_challenges, ideal_buyer_profile
• Resources & Technologies
  • technologies (multi‑select; array of strings)
  • included_assets (array of strings)

If data is not found, numeric fields default to 0 and strings to an empty string.

Architecture

Frontend Components

1. Mistral AI Analyzer (N8NAIProjectAnalyzer)

• Theme: Blue/Purple gradient with AI-futuristic animations
• Icon: CPU/Bot icons with orbiting particles
• Authentication: Requires JWT (from useAuth context)
• Endpoint: https://nujcbrbtejvhuajfwgeo.supabase.co/functions/v1/analyze-project
• Features:
  • Accepts a project URL input
  • Calls the Supabase Edge Function with the current user JWT
  • Displays advanced loading animations with dynamic text
  • Emits AI result back to parent via onDataGenerated callback
  • Handles 429 (rate limit) errors with user‑friendly messages
  • Comprehensive error handling and logging
  • Maps AI response to FormData structure correctly

2. Legacy n8n AI Analyzer (Deprecated)

• Status: Deprecated in favor of Mistral AI
• Reason: CORS issues and complexity of external webhook management
• Replacement: All functionality moved to Mistral AI implementation

2. Form Integration (CreateListingContent)

• Renders Mistral AI analyzer for new listings only (not during edit)
• Merges AI data into formData using safe, field‑by‑field mapping
• Normalizes category to a valid slug from listing_categories
• Maps structured markdown sections into dedicated description_* form fields (accordions)
• Preserves user‑typed data (AI only fills empty fields)
• Comprehensive logging for debugging AI data flow

Backend

Mistral AI Backend (Supabase Edge Function)

• functions/analyze-project
  • Verifies JWT (config verify_jwt = true)
  • Validates URL input
  • Builds optimized JSON‑only prompt for Mistral (max 3 sentences per field)
  • Calls Mistral Chat Completions API with temperature 0.7
  • Model Selection: mistral-small-latest (with fallback to mistral-large-latest)
  • Token Optimization: Reduced max_tokens to 1000 for efficiency
  • Robust JSON Parsing: 5-strategy truncation and repair logic
  • Error Recovery: Fallback to minimal valid JSON if parsing fails
  • Rate Limit Handling: User-friendly 429 error messages
  • Financial Defaults: Generates market averages when data unavailable
  • Timeout Management: 60-second timeout with AbortController

Environment & Configuration

Mistral AI

• Secret: MISTRAL_API_KEY (Supabase project secret)
• Config: supabase/config.toml
  • [functions.analyze-project] verify_jwt = true
• Models: mistral-small-latest (primary), mistral-large-latest (fallback)
• Rate Limits: Handles 429 errors gracefully with user guidance

n8n Integration (Optional)

• Webhook URL: https://n8n.10alert.com/webhook/95e9d69b-25c0-4642-bd84-9a52ad68f912
• Purpose: Admin notifications and listing events
• Configuration: Set as Supabase secrets N8N_WEBHOOK_URL, N8N_WEBHOOK_CHAT, N8N_WEBHOOK_LISTING

How It Works (Step‑by‑Step)

Mistral AI Flow

1. User opens /create-listing (authenticated)
2. Enters a project URL in the AI analyzer card
3. Clicks "Analyze Project" button
4. Frontend displays AI-futuristic loading animation with dynamic text
5. Frontend sends POST to functions/v1/analyze-project with Authorization: Bearer <JWT>
6. Edge Function validates JWT and calls Mistral API
7. Robust Processing: JSON parsing with 5-strategy recovery if needed
8. Financial Defaults: Generates market averages for missing financial data
9. Edge Function returns structured JSON response
10. Frontend merges AI fields into form state (preserves user input)
11. User reviews and edits populated fields as needed
12. User saves draft or submits for review (existing flow unchanged)

Error Handling Flow

1. 429 Rate Limit: User sees "Service is currently overloaded. Please try again in 10 minutes."
2. JSON Parse Error: System attempts 5-strategy truncation and repair
3. Complete Failure: Falls back to minimal valid JSON with "Unable to determine" values
4. Timeout: 60-second timeout with user-friendly error message

API Contract

Request

{
  "url": "https://example.com"
}

Response (example; fields may be empty if data is unavailable)

{
  "title": "Project Name",
  "category": "saas",
  "website_url": "https://example.com",
  "business_model": "Markdown body...",
  "description": "Overview paragraph...",
  "established_date": "2023-01-01",
  "number_of_customers": 1250,
  "monthly_revenue": 15000,
  "monthly_profit": 8500,
  "sale_price": 125000,
  "allow_negotiation": true,
  "price_justification": "Markdown body...",
  "monthly_traffic": 45000,
  "target_audience": "Markdown body...",
  "competitors": "Markdown list...",
  "growth_opportunities": "Markdown body...",
  "reason_for_selling": "Markdown body...",
  "tech_stack": ["React", "Node.js", "PostgreSQL", "AWS"],
  "included_resources": ["Codebase & Files", "Domain", "Analytics"],
  "introduction": {
    "business_description": "Markdown...",
    "business_model": "Markdown...",
    "key_strengths": "Markdown...",
    "team_founder": "Markdown..."
  },
  "business_details": {
    "included_assets": "Markdown...",
    "operations_time_required": "Markdown...",
    "technology_stack": "Markdown...",
    "financial_performance": "Markdown...",
    "expenses": "Markdown...",
    "growth_opportunities": "Markdown..."
  },
  "conclusion_motivation": {
    "reason_for_selling": "Markdown...",
    "benefits_for_buyer": "Markdown...",
    "risks_challenges": "Markdown...",
    "ideal_buyer_profile": "Markdown..."
  }
}

Field Mapping Highlights

• Category normalization

  • AI returns a label (e.g., "SaaS"); frontend normalizes to slug and matches against DB listing_categories.slug/name.
  • Fallback: empty string when no confident match is found (prevents invalid slugs).

• Structured Description (accordions)

  • introduction.* → description_introduction_*
  • business_details.* → description_content_*
  • conclusion_motivation.* → description_conclusion_*

• Technologies/Included Assets

  • tech_stack → technologies[]
  • included_resources → included_assets[]

• Financials

  • sale_price → price
  • price_justification → price_reasoning(_en/_ro)

Security Considerations

• JWT verification enforced by Supabase (function‑level and in code)
• CORS configured for Supabase Functions endpoint
• No secrets in frontend; AI key stored as Supabase secret
• Defensive parsing: AI output must be valid JSON; otherwise a 500 error is returned

UX Notes

• Analyzer card is visible only for new listings (not during edit)
• Clearly communicates loading/progress and any errors
• Never overwrites existing user input; only fills blanks

Troubleshooting

Common Issues

• 404 Not Found

  • Ensure the frontend calls the fully‑qualified Supabase Functions URL
  • Check that Edge Function is deployed: supabase functions deploy analyze-project

• 401 Invalid JWT

  • User must be logged in; frontend sends Authorization: Bearer <session.access_token>
  • Check if JWT has expired (logout and login again)

• "Failed to parse AI analysis"

  • System automatically attempts 5-strategy JSON repair
  • Check Edge Function logs for detailed error information
  • Verify Mistral API response contains valid JSON

• 429 Service tier capacity exceeded

  • User sees friendly message: "Service is currently overloaded. Please try again in 10 minutes."
  • System automatically switches to mistral-small-latest model

• Category remains empty

  • AI label may not match DB category slug
  • Check listing_categories table for available slugs
  • Add missing category or adjust normalization logic

• Price not populated

  • Verify sale_price mapping to price field
  • Check console logs for mapping debug information

Debug Information

• Frontend Logs: Check browser console for AI data flow
• Edge Function Logs: Check Supabase dashboard for server-side errors
• Mapping Logs: Console shows sale_price from AI, mapped price, etc.

Testing Checklist

Basic Functionality

• [ ] Public website URL analysis (e.g., https://windowsreport.com)
• [ ] GitHub repository URL analysis
• [ ] Authentication required (redirects to login if not authenticated)

Data Population

• [ ] Basic Information: title, category (slug), website_url
• [ ] Financial Data: monthly_revenue, monthly_profit, price (sale_price)
• [ ] Structured Descriptions: introduction, business_details, conclusion_motivation
• [ ] Technologies: tech_stack array, included_assets array
• [ ] Business Details: established_date, number_of_customers, monthly_traffic

Error Handling

• [ ] 429 rate limit shows user-friendly message
• [ ] Invalid URLs show appropriate error
• [ ] JSON parsing errors are handled gracefully
• [ ] Timeout after 60 seconds

User Experience

• [ ] Loading animation displays correctly
• [ ] User-typed fields are preserved on re-analysis
• [ ] Form validation works with AI-populated data
• [ ] Only new listings show AI analyzer (not during edit)

Change Log (Feature)

v1.0.0 (Initial Release)

• Added AI Project Analyzer component with Mistral integration
• Implemented analyze-project Edge Function
• Category slug normalization and structured mapping

v1.1.0 (Robust Error Handling)

• Added 5-strategy JSON parsing recovery
• Implemented rate limit handling with user-friendly messages
• Added financial defaults for missing data
• Enhanced logging for debugging

v1.2.0 (UI/UX Improvements)

• AI-futuristic loading animations with dynamic text
• Improved error states and user feedback
• Comprehensive field mapping with debug logging
• Deprecated n8n webhook in favor of Mistral AI

v1.3.0 (Production Ready)

• Fixed price mapping (sale_price → price)
• Optimized token usage and model selection
• Added timeout management with AbortController
• Enhanced error recovery mechanisms