CAR BLOG AGENT

Source in this repository

Overview

Car Blog Agent is an automated Python script that drives an AI agent to stick to a strict narrative and produce ongoing writing content. The narrative, audience targets, and market framing are hardcoded in the script: categories (Electric Vehicles, Used Cars, New Cars, SUVs & Trucks), audience config (technician, consumer, industry, general) with fixed prompt keys and keywords, and article templates. The script feeds this structure to an AI (Ollama or OpenAI) so generated articles stay on-message and match the intended format.

Architecture

Key components

• CarMarketDataCollector: Supplies hardcoded market trends (EV, used, new, SUV/truck) and data points baked into the script
• BlogArticleGenerator: Sends narrative and templates to the AI (Ollama or OpenAI) so output stays on-message
• BlogPublisher: Saves articles and builds the blog index
• CarBlogAgent: Orchestrates the script and keeps the AI on the strict narrative

Features

• Hardcoded narrative: Categories, trends, data points, and audience config are fixed in the script
• Strict narrative: Prompts and templates constrain the AI so content matches the intended framing
• Ongoing content: Script can be run repeatedly to produce a steady stream of articles
• AI integration: Uses Ollama (local) or OpenAI API; narrative is enforced via hardcoded prompts and structure
• Automated publishing: Writes Markdown articles and generates the blog index
• Keyword and audience: Per-audience keywords and prompt keys are defined in the script

Market Categories

Audience Targeting

Key source file

car_blog_agent.py

Single script containing the hardcoded narrative and all logic:

• CarTrend: In-script structure for trends; actual trend text and data points are hardcoded in collector methods
• BlogArticle: Structure for the AI output (title, content, summary, keywords, etc.)
• AUDIENCE_CONFIG: Hardcoded audience types, prompt keys, and keyword lists
• CarMarketDataCollector: Returns fixed trends (EV, used, new, SUV/truck) with hardcoded metrics and impact text
• BlogArticleGenerator: Passes narrative and templates to the AI so articles follow the strict format
• BlogPublisher: Writes articles and index to disk
• CarBlogAgent: Runs the pipeline and keeps the AI aligned to the script’s narrative

AI integration

The script feeds hardcoded narrative, templates, and audience config to an AI so it produces articles that stay on-message. The AI does not choose the narrative; the script does.

OpenAI API

Optional. Configure with API key; the script still enforces the same strict narrative and structure.

Ollama

Local model (e.g. deepseek-r1:1.5b) for content generation. Narrative and format are still defined in the script. Configure model name, temperature, and max_tokens in the script or via your environment.

Dependencies

• requests - HTTP requests for API integration
• openai - OpenAI API client (optional)
• csv - CSV data processing
• datetime - Date and time handling
• random - Random selection for trend sampling
• typing - Type hints for code clarity
• dataclasses - Data structure definitions

Usage

From the project directory, run the script (e.g. python car_blog_agent.py). You can configure an optional OpenAI API key, use_ollama, and ollama_config (model name, temperature, max_tokens) in the script. Specify number of articles and audience mix as needed; generated articles are saved to blog_articles/.

Output Format

Articles are saved as Markdown files with the following structure:

• Title and metadata (publish date, category, word count, keywords)
• Executive summary
• Current market data with metrics
• Market impact analysis
• Key insights and recommendations
• Conclusion with data source and confidence level

Project Structure

car-blog-agent/
├── car_blog_agent.py    # Main application
└── blog_articles/       # Generated articles (created at runtime)
    ├── index.md         # Blog index
    └── *.md             # Individual articles