Public agent for anyone

This post explains how the website assistant works.

Stack

• Streamlit for the responsive, embedded UI
• OpenRouter as the LLM backend (supporting 200+ models)
• Markdown files as the content source (scoped to /content directory only)

Overview

The website chat interface is a single-screen, embeddable chat widget designed to be integrated into static websites via an <iframe>. It provides an iMessage-style conversation UI with blue user bubbles and dark gray assistant bubbles, all styled with a modern dark theme.

What It Does

This is a secure, content-aware assistant that:

1. Indexes all markdown files in the repository's content/ directory
2. Searches for relevant articles based on user questions using keyword matching
3. Fetches full article content when needed to provide grounded, contextual answers
4. Maintains conversation history to enable follow-up questions within the same session
5. Never accesses source code, environment secrets, or files outside the /content boundary

Security & Scope

The chat interface enforces a strict content boundary:

• ✅ Accessible: All files matching content/*/.md
• ❌ Not accessible: Source code, .env files, secrets, configuration files, or any path outside content/
• ✅ Startup validation: If CONTENT_DIR is misconfigured, the app fails fast with a clear error

This design ensures users can safely expose the chat widget publicly without revealing sensitive project information.

Interaction Model

The assistant can:

1. Search relevant articles by keyword from user queries
2. Retrieve full article content by file slug
3. Answer questions grounded in the retrieved markdown files
4. Remember context across multiple messages in a session

Features & Customization

• Iframe-ready: Embed via https://YOUR-STREAMLIT-APP.streamlit.app/?embed=true
• Configurable LLM: Switch models via OPENROUTER_MODEL env var
• Content limit controls: Set MAX_ARTICLE_CONTEXT_CHARS to control context window
• Safety modes: Optional content filtering with configurable blocked terms
• Message limits: User input and context are size-limited to prevent abuse

Learn More

For detailed documentation, implementation details, and deployment instructions, visit the GitHub repository:

→ streamlit-openrouter-chat-embed

The repo includes setup guides, environment configuration examples, embedding instructions, and contribution guidelines.