Landing Page Optimizer Agent

Open-source web UI for the Landing Page Optimizer Agent — powered by UPLIZD. The hosted flow analyzes landing performance and shapes A/B tests using gpt-4o-mini, a composable instruction block ({agent} + {rules} for Growth Hacker persona and execution rules), plus Google Analytics (Composio, OAuth2), Firecrawl (CRAWL, timeout 10), and Web Search (PERFORM_SEARCH).

---

How it works

This repo is only the playground UI + Express proxy. Integrations run inside your UPLIZD flow—configure OpenAI, Composio, and Firecrawl in-studio, not in .env.

Browser  →  React UI  →  POST /api/run (proxy)  →  UPLIZD API  →  Landing Page Optimizer flow

---

Workflow logic (matches the Langflow canvas)

Piece	Role
Chat Input	User shares URLs, GA context, experiment goals, competitor names
Prompt template	Fuses {agent} + {rules} from Text Input nodes → Agent Instructions
Language Model	gpt-4o-mini
Google Analytics (Composio)	Traffic, behavior, conversion signals (OAuth2)
Firecrawl	Crawl landing URLs; template uses CRAWL action, timeout 10
Web Search	PERFORM_SEARCH for benchmarks / competitor copy
Chat Output	Recommendations, test hypotheses, next steps

Prerequisites (in the flow): OpenAI API key, Composio API key, Firecrawl API key; complete GA auth through Composio.

---

Features

• Chat playground aligned with Chat Input / Chat Output
• Secure proxy — UPLIZD_API_KEY never sent to the browser
• Quick prompts for GA + crawl + search workflows
• Dark mode — follows system preference
• Session ID — “New session” rotates session_id

---

Extension ideas

Ideas for fork-level or flow-level extensions without changing this repo’s proxy/UI contract (POST /api/run, chat in → chat out):

• Experiment backlog — In the UPLIZD flow, add rules so the agent outputs a ranked table: hypothesis, primary metric, minimum sample, guardrails (e.g. bounce vs. conversion).
• Segment-aware optimization — Extend {rules} so GA context is requested by traffic source or landing cohort before recommending copy or layout tests.
• Competitor tear-down mode — Quick prompts that force Firecrawl + Web Search to produce a side-by-side (hero, CTA, proof, pricing cues) before suggesting tests on your URL.
• Locale / brand voice — Extra Text Input or rules block for tone, forbidden claims, and language so outputs stay compliant while testing headlines.

---

Stack

Layer	Tech
Frontend	React 18 + Vite
Proxy	Express (Node.js)
Workflow	UPLIZD Marketplace

---

Prerequisites

• Node.js ≥ 18
• UPLIZD account + API key
• Flow installed from Marketplace when published (or import the template)

---

Quick start

1 — Install the flow

→ Landing Page Optimizer Agent on UPLIZD Marketplace

1. Install the flow and open it in the editor.
2. Copy Flow ID from the URL.
3. Settings → API Keys — copy your UPLIZD key.
4. In the flow: set OpenAI on the model node; Composio on Google Analytics + Web Search toolsets; Firecrawl API on the crawl node; verify CRAWL + timeout 10 and PERFORM_SEARCH.

2 — Configure this repo

git clone https://github.com/YOUR_ORG/landing-page-optimizer-agent.git
cd landing-page-optimizer-agent
cp .env.example .env

UPLIZD_API_KEY=your_api_key_here
UPLIZD_FLOW_ID=your_flow_id_here

3 — Run locally

npm run install:all
npm run dev

Open http://localhost:5173

---

Environment variables

Variable	Required	Description
UPLIZD_API_KEY	✅	Your UPLIZD API key
UPLIZD_FLOW_ID	✅	Flow ID after install
UPLIZD_BASE_URL	—	Default https://studio.uplizd.ai
PORT	—	Proxy (default 3001)
CORS_ORIGIN	—	Default http://localhost:5173
FLOW_INPUT_TYPE / FLOW_OUTPUT_TYPE	—	Default chat

---

Project structure

landing-page-optimizer-agent/
├── .env.example
├── .gitattributes
├── package.json
├── server/index.js
├── web/
└── docs/
    ├── workflow.png
    ├── setup.md
    └── contributing.md

---

Contributing

See docs/contributing.md.

---

License

MIT — see LICENSE.