Mongodb Atlas - Vector Search - Build an AI agent for Blog Post Content - #2

This pull request refactors the project to transform it from an HR/employee-focused agent demo into the "TLD Blog Agent"—an AI assistant for searching and summarizing blog content from itsthatlady.dev. The changes include major updates to the README and codebase to reflect the new domain, a new database seeding process that fetches and embeds real blog posts, and improvements to the API and user experience, including a web-based chat interface.

Major project/domain refactor:

• All references to "employee" and "HR database" have been replaced with "blog post" and "blog_database" throughout the codebase, including database names, collection names, and tool descriptions (agent.ts, seed-database.ts) [1] [2] [3] [4].
• The agent's system prompt and tool are now specialized for answering questions about blog content, with improved response formatting and instructions for linking to relevant posts (agent.ts).

Database seeding and vector search:

• The seeding script now fetches real blog posts from itsthatlady.dev, extracts and cleans their content, generates embeddings, and stores them in MongoDB Atlas. The old synthetic employee data generation has been removed (seed-database.ts) [1] [2].
• Instructions have been added to the README for creating a vector search index in MongoDB Atlas after seeding (README.md).

API and user experience improvements:

• The Express server now serves a web-based chat interface from the public/ directory, with updated endpoints and improved logging for chat interactions (index.ts, README.md) [1] [2].
• The README has been rewritten to match the new blog agent functionality, with updated examples, project structure, and usage instructions (README.md) [1] [2].

Tooling and model updates:

• The agent now uses Anthropic's latest Claude model for chat and OpenAI only for embeddings, reflecting the new use case (agent.ts, seed-database.ts) [1] [2].

Summary of most important changes:

Domain and functionality refactor:

• Replaced all employee/HR-related logic with blog post search and summarization, including database names, collection names, and agent tool descriptions (agent.ts, seed-database.ts) [1] [2] [3] [4].
• Updated the agent's system prompt and response format to focus on blog content, including instructions for linking to relevant posts (agent.ts).

Database and embeddings:

• Seeding script now fetches real blog posts, extracts and cleans their content, generates embeddings, and stores them in MongoDB Atlas; removed synthetic employee data generation (seed-database.ts) [1] [2].
• Added README instructions for creating a MongoDB Atlas vector search index for blog post embeddings (README.md).

User interface and API:

• Added a web-based chat interface and updated Express server to serve it, with improved logging and updated API endpoint examples (index.ts, README.md) [1] [2].

Model and tool usage:

• Switched to using Anthropic Claude for chat and OpenAI for embeddings only, matching the new application focus (agent.ts, seed-database.ts) [1] [2].

Documentation:

• Rewrote the README to reflect the new TLD Blog Agent, including updated usage, project structure, and endpoint documentation (README.md) [1] [2].