Lamaq/iPoly-LOC

fix: extract explicit company_name from prediction level, strictly fallback to LinkedIn posts and macro news without Crunchbase hallucination

feat: increase LinkedIn recent posts context density in prompts for better drafting

fix: logic to use entire scraped data, handle news fallbacks to macro domain context, no Crunchbase hallucination

fix: reasoning and drafting prompts to prioritize live fetched news over mock data

fix: robust company_name + industry extraction waterfall in node_deep_research

fix: robust news fetching with rate-limit detection, caching, and dynamic industry

feat: dynamic LLM-generated execution flowchart based on POC data

fix: robust company name extraction for accurate news ingestion

feat: user-centric oauth and calendar endpoints

Fix: ml_output shape mismatch in gatekeeper and all downstream nodes

Integrate real scraped POC data for personalization

Fix: numpy serialization error in ML predict endpoint

V1.4: ML model integration with SHAP explainability and Domain navigation

V1.3: Upgrade Email and Calendar to Official Google API Client (OAuth2)

Fix: email_workflow imports and python-multipart dependency

V1.2: Email Workflow Engine + LEAD_JPMORGAN + Gmail/Calendar services

Add real demo leads: TerraByte + Visa Friendly with LinkedIn profiles and ML scores

Fix MongoDB SSL: certifi CA bundle + engine init at startup

Deploy V1.1 backend to HF Spaces - FastAPI + LangGraph + 15 endpoints

initial commit