PersonaAI – Personal Branding Intelligence Agent

A multi-agent AI system that helps professionals build a consistent LinkedIn personal brand through an end-to-end feedback loop.

Built with Python, FastAPI, LangChain, LangGraph, Groq (Llama 3.3 70B), SQLAlchemy, and Streamlit.

---

What It Does

PersonaAI takes the guesswork out of personal branding on LinkedIn. It creates a structured brand profile, generates posts aligned to your brand strategy, predicts engagement before you publish, tracks real performance metrics, and gives AI-powered feedback that gets smarter with more data.

The Workflow

Define Brand → Generate Post → Predict Engagement → Publish → Log Metrics → Get Feedback → Improve

---

Architecture

PersonaAI/
├── main.py                  # FastAPI entry point
├── agents/
│   ├── brand_agent.py       # Brand profile creation (LangChain + ReAct)
│   ├── content_agent.py     # LinkedIn post generation
│   ├── feedback_agent.py    # Engagement analysis & recommendations
│   └── orchestrator.py      # LangGraph StateGraph orchestration
├── services/
│   └── llm_service.py       # Groq LLM wrapper (Llama 3.3 70B)
├── database/
│   ├── models.py            # SQLAlchemy ORM models
│   └── db.py                # Engine, session, dependency
├── schemas/
│   └── schemas.py           # Pydantic request/response models
├── utils/
│   ├── auth.py              # API key authentication
│   ├── rate_limiter.py      # Request rate limiting
│   └── observability.py     # Prometheus metrics + structured logging
├── static/                  # Frontend UI (6-tab dashboard)
├── streamlit_dashboard.py   # Monitoring dashboard
├── Dockerfile
├── docker-compose.yml
├── requirements.txt
└── .env.template

---

Features

• 4 AI Agents + 1 Orchestrator — Brand, Content, Feedback, and Predictor agents, all coordinated through a LangGraph StateGraph
• Engagement Prediction — scores your post out of 100 with a rating breakdown before you publish
• Feedback Loop — analyzes past engagement data and recommends strategy improvements
• Dashboard — performance overview, post history, averages, and best-performing topics
• Prompt Versioning — version-controlled prompt templates with rollback support
• Audit Logging — every LLM call logged with trace IDs
• API Security — Bearer token authentication + per-endpoint rate limiting
• Prometheus Metrics — request counts, latency histograms, error rates at /metrics
• Streamlit Monitoring — visual observability dashboard for engagement trends
• Docker Deployment — containerized with docker-compose
• Human-in-the-Loop — no auto-posting; you review everything before it goes live

---

Quick Start

Prerequisites

• Python 3.10+
• A free Groq API key → console.groq.com/keys

Setup

cd PersonaAI

python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # macOS / Linux

pip install -r requirements.txt

copy .env.template .env
# Open .env and add your GROQ_API_KEY and set an API_KEY

Run

uvicorn main:app --reload

• App UI → http://127.0.0.1:8000
• Swagger Docs → http://127.0.0.1:8000/docs
• Prometheus Metrics → http://127.0.0.1:8000/metrics

Streamlit Dashboard

streamlit run streamlit_dashboard.py

Docker

docker-compose up --build

---

API Endpoints

All endpoints require Authorization: Bearer <your-api-key> header.

Method	Endpoint	Description
POST	/brand	Create a brand profile
POST	/generate	Generate a LinkedIn post
POST	/predict	Predict engagement score before publishing
POST	/engagement	Log engagement metrics (likes, comments, shares)
POST	/feedback	Get AI-powered strategy feedback
POST	/orchestrate	Run the full pipeline (brand → content → feedback)
GET	/metrics	Prometheus metrics
GET	/health	Health check
GET	/audit-logs	View LLM call audit trail
GET	/prompts	View prompt version history

---

Environment Variables

Variable	Description	Required
GROQ_API_KEY	Groq API key for Llama 3.3 70B	Yes
API_KEY	Bearer token for endpoint auth	Yes
DATABASE_URL	SQLite connection string	No (defaults to sqlite:///./personaai.db)

---

Tech Stack

Backend — Python, FastAPI, SQLAlchemy, SQLite, Pydantic

AI/Agents — LangChain (ReAct agents with tool-use), LangGraph (StateGraph orchestration), Groq (Llama 3.3 70B)

Observability — Prometheus, Streamlit

Security — Bearer token auth, SlowAPI rate limiting, audit logging

Deployment — Docker, docker-compose, Gunicorn