fix: make recommendation engine robust and prioritize recommendations in dashboard

feat: implement IA recommendation engine based on company keywords

feat: implement local database management system with admin stats and clear functions

feat: enhance tender detail scraping with guarantees, items and direct links to portal technical sheet

feat: implement Llama 3.2 Vision support in chatbot for image analysis

fix: improve chatbot self-awareness and expand model engine mapping

Fix false positive on questions and make Q&A generation conditional

Enhance Market Questions: Generate and display full list of 22 questions

fix: backend crash due to missing SearchRecord import

feat: persistent search history and professional PDF export with branded headers

feat: enhance History view with raw agent thoughts and remove Reports tab

feat: upgrade proposal draft generation with professional structure

fix: rename reserved 'metadata' attribute to 'tab_metadata' in SQL models

feat: integrate buyer risk intelligence and scraped tender details into agent analysis

🚀 Production: 3-day deep search fix + Lifecycle UI + Data persistence

💎 Deep Analysis: Enabled extraction and display of Evaluation Criteria and Attachments from Mercado Público

🔧 Fix: Resolved UNIQUE constraint error in tender synchronization

⚙️ Backend: Auto-sync with Mercado Público when local DB is empty

🚀 Fix: Company profile persistence and input logic

🚀 Feature: Enhanced elite agent chat intelligence

🚀 Fix: Add synthetic attachments to generated tenders

🚀 Fix: NameError List to list

🚀 Feature: Synthetic tender generation with coherent documents

🚀 Fix: Realistic buyer fallback for 'Unknown' results

🚀 Fix: Ensure app_port 7860 and simple health check

🚀 Fix: Move SQLite to /tmp on Linux for HF compatibility

🚀 Fix: Correct README metadata color

🚀 Fix: Port 7860 and absolute DB path for HF Spaces

🚀 Production ready for HF Spaces (Security Cleaned)

fix: remove demo/fake data seeding to ensure only real API data is shown

fix: force Chile timezone for API calls and fix status mapping to resolve 500 errors

fix: remove unused imports in tenders router to resolve backend crash

fix: ensure API calls always include required date parameter and improve status feedback

fix: restore missing API functions in mercado_publico.py to resolve backend startup error

feat: enhance advanced search with Date, Org, and Type filters using MP API

debug: add LLM logging and model fallbacks

fix: chat response visibility and backend fallback

feat: implement Groq integration, robust multi-agent synthesis and UX performance optimization

Fix async LLM calls and agents orchestration to resolve chat connectivity error

Implement Agent Chat feature with expert selection and real-time context

Add Requirement Response (Q&A Style) feature for document analysis

Improve Agent Analysis UX with parallel execution, model support, and status log

Fix: Enable Featherless fallback for agents when Gemini unavailable

Improve Mercado Público search integration and tender search UX

fix: implement rate-limiting and retry logic for Mercado Público APIs

fix: refine OC API parameters and handle 500 errors gracefully

feat: implement Purchase Order (OC) monitoring and real-time market monitor UI

feat: enhance AI agents with detailed Mercado Publico data and update frontend UI

🐛 Fix: Persistent Company Profile via Database to resolve 404 in production

🚀 Production Ready: Multi-Model Scraper with Synthetic Fallback Intelligence and UI Polish