LinkForge 0.1.0

First release of LinkForge, a LinkedIn content intelligence platform.

Overview

LinkForge ingests a profile's posts, runs natural-language analysis over the content and its comments, trains a per-profile engagement model, and produces data-driven recommendations for future posts. It exposes a FastAPI backend and a Streamlit dashboard, backed by PostgreSQL with the pgvector extension.

Features

• Profile and post ingestion via a Playwright-based scraper with cookie-based authentication.
• Sentiment analysis (VADER) augmented with lexical dimensions for technical discussion: pragmatic/balanced, tribalism, and technical-depth scores.
• Theme detection, theme-confidence scoring, and comment-driven polarization scoring.
• 384-dimension sentence embeddings stored as pgvector columns for similarity queries.
• A scikit-learn engagement predictor trained per request, producing an engagement estimate and a success probability relative to the profile's own distribution.
• Next-post recommendations: suggested topic, tone, structure, and hook.
• CSV and JSON export of analysis results from the dashboard.

Stack

Python 3.11+, FastAPI, Streamlit, PostgreSQL 16 + pgvector, SQLAlchemy 2.0 (async) and Alembic, Playwright, sentence-transformers, scikit-learn, vaderSentiment.

See the README for setup, configuration, and API reference.