Demo Generator

Template-based generation of personalized demo landing pages and outreach assets for Japanese small-business prospects.

Overview

Demo Generator turns approved lead or worklog data into lightweight, shareable HTML demo pages. In the broader workflow, lead-finder identifies promising businesses, this repository generates a tailored demo page or preview asset for each target, and the outreach layer uses those assets during contact or follow-up. The current implementation is practical and operational: it reads TSV or Excel-based lead lists, applies existing HTML templates, assigns images, and writes static output files for review.

Why This Project Exists

For local-business outreach, a generic pitch is weak. A fast, concrete demo page gives the prospect something specific to react to. This repository reduces the manual work of building those first-pass assets one by one and keeps the process repeatable across batches.

Core Features

• Generate static HTML demo pages from operational Excel worklogs or a small TSV sample list
• Maintain multiple reusable template variants (A through F)
• Auto-assign page images from existing pools using simple atmosphere heuristics
• Support deterministic batch generation with --start-id, --template, --output-template, and --excel
• Write generation logs to CSV for batch review
• Produce optional SNS/demo preview sets from the same templates and images

Workflow

1. Start with approved leads from lead-finder or an internal worklog.
2. Prepare input data in input/ as either an Excel file for the operational path or input/list.tsv for the simple sample path.
3. Run the generator against a selected template.
4. Review the generated HTML and output/generation_log.csv.
5. Use the page or asset in the downstream outreach workflow.

Pipeline context:

lead-finder -> demo-generator -> outreach automation

Screenshot

Generated demo page: static HTML output created from an approved lead record and reusable template.

Recommended Usage

1. Main operational path: auto_generate.py

Use this when generating pages from Excel-based worklogs or normalized handoff CSVs from lead-finder.

pip install openpyxl
python auto_generate.py

Behavior:

• scans input/ for the first Excel file whose name includes 営業
• reads the active sheet, or reads --csv when provided
• normalizes brand name, URL, and ID fields
• writes HTML under output/<TEMPLATE>/
• writes output/generation_log.csv

Useful examples:

python auto_generate.py --excel "input/営業ログ(東京).xlsx" --template A --output-template A
python auto_generate.py --csv "../lead-finder/web_app/output/handoff_normalized_leads_fukuoka_city_20260512_171821.csv" --template A --output-template A
python auto_generate.py --start-id 8000 --template B --output-template B

Normalized handoff CSVs should provide at least id or lead_id, a name field such as brand_name, business_name, or company_name, and a URL field such as reference_url, website, or url.

Operational note:

• auto_generate.py expects page images to already exist under output/<template>/images/

2. Simple sample path: generate.py

Use this for a quick, repo-local preview based on input/list.tsv.

python generate.py

This writes files like output/demo01A_index.html and references images from input/images/.

3. Optional preview asset path: scripts/create_sns_site.py

Use this to generate preview/demo variants into sns/.

python scripts/create_sns_site.py

Tech Stack

• Python
• Standard library: csv, os, argparse, datetime, random
• openpyxl for Excel input
• Static HTML/CSS templates in templates/
• Local image assets in input/images/ and output/<template>/images/

Project Structure

demo-generator/
|-- auto_generate.py        # Main operational generator
|-- generate.py             # Simple TSV-based generator
|-- templates/              # Source HTML templates (A-F)
|-- input/                  # Sample input, source images, operational worklogs
|-- output/                 # Generated HTML and logs
|-- scripts/                # Helper scripts and preview generation
|-- sns/                    # Generated preview/demo pages
|-- README.md
|-- AGENTS.md
`-- .gitignore

Notes:

• auto_generate.py is the recommended public starting point.
• generate.py is the simpler sample path.
• scripts/generate_by_id_range.py is a narrow helper, not the main interface.
• extract_and_run.py and some files in scripts/ are currently empty placeholders.

Setup

Requirements in the current checkout are minimal:

• Python 3
• openpyxl for the Excel-based workflow

Install:

pip install openpyxl

The repository does not currently include a pinned requirements.txt or pyproject.toml.

Usage

Excel-based batch generation

python auto_generate.py
python auto_generate.py --excel "input/営業ログ(東京).xlsx"
python auto_generate.py --csv "../lead-finder/web_app/output/handoff_normalized_leads_fukuoka_city_20260512_171821.csv"
python auto_generate.py --start-id 8000 --template A --output-template A

TSV-based sample generation

python generate.py

Preview or SNS sample generation

python scripts/create_sns_site.py

Example Output

Main operational output:

output/
|-- A/
|   |-- 08000A.html
|   |-- 08001A.html
|   `-- images/
`-- generation_log.csv

Simple sample output:

output/demo01A_index.html
output/demo06B_index.html

Sample TSV row:

id	brand_name	reference_url	template	image	therapist_image
demo01	Demo Site	https://example.com	A	image01.jpg	image26.jpg

Limitations

• Personalization is currently template-driven; it does not generate custom copy with an LLM
• The Excel path assumes operational worklogs with recognizable columns for brand name, URL, and optional ID
• The repository currently mixes source files with generated output and local working data
• There is no automated test suite or pinned dependency manifest in the current checkout
• Some tracked folders in the current repo, especially output/, sns/, and private worklogs under input/, are better treated as local or generated data than permanent public source

Related Repositories

• lead-finder: discovers and scores candidate businesses
• demo-generator: generates personalized demo pages and assets
• playwright-automation: uses approved outputs during outreach execution

Roadmap

• Keep the public README centered on the operational path and the sample path
• Separate source assets from generated outputs more cleanly
• Add a small redacted sample dataset for public use
• Add a pinned dependency file and a lightweight verification path
• Reduce tracked bulk output so the repository is easier to review