Second Renaissance — AI Guide

A web-based chatbot that explains the ideas of the Second Renaissance movement, powered by Claude (claude-sonnet-4-6). Drop documents into a folder and the bot uses them as its knowledge base — no code changes required.

---

Quick start

1. Install dependencies

pip install -r requirements.txt

Requires Python 3.10+.

2. Set your API key

cp .env.example .env
# Open .env and replace the placeholder with your real key:
#   ANTHROPIC_API_KEY=sk-ant-...

3. Add documents

Place .txt, .md, or .pdf files in the documents/ folder. See documents/placeholder.txt for full guidance.

4. Run the server

python server.py

Open http://localhost:8000 in your browser.

---

Project structure

2RAI/
├── server.py          # FastAPI backend — document loading, chat endpoint
├── config.py          # Model, max_tokens, system prompt preamble
├── requirements.txt
├── .env.example       # Copy to .env and add your API key
├── documents/         # ← Drop your content here
│   └── placeholder.txt
├── static/
│   └── index.html     # Single-page chat UI
└── README.md

---

Configuration

Edit config.py to change any of these settings without touching the server code:

Setting	Default	Description
MODEL	claude-sonnet-4-6	Anthropic model to use
MAX_TOKENS	1024	Maximum tokens in each response
DOCUMENTS_FOLDER	documents	Path to the documents folder
SYSTEM_PROMPT_PREAMBLE	(see file)	Bot persona and instructions

---

Document management

The documents/ folder is the only place you need to touch to manage content:

• Add a file → corpus reloads automatically within ~1 second
• Remove a file → corpus reloads automatically
• Edit a file → corpus reloads automatically

Supported formats: .txt, .md, .pdf

If a PDF cannot be parsed (scanned images, corrupted files), a warning is logged and the file is skipped — the server keeps running with the remaining documents.

Prompt caching

The full document corpus is sent as a cached system-prompt block. The first request after a corpus change pays full input-token cost; subsequent requests within 5 minutes reuse the Anthropic prompt cache (~90 % cheaper for the document portion). Cache usage is visible in the Anthropic console.

---

How it works

1. Startup — server.py reads every supported file in documents/, extracts text (using pdfplumber for PDFs), and assembles a single corpus string.
2. File watching — watchdog monitors the folder; any change triggers a debounced reload (1 s delay to avoid partial-write races).
3. Chat endpoint — POST /chat accepts { message, history } and calls the Claude API with streaming enabled. The corpus is injected into the system prompt with cache_control: ephemeral.
4. Frontend — A plain HTML/JS page reads the SSE stream token-by-token and renders a live, markdown-aware response.

---

Troubleshooting

Symptom	Likely cause	Fix
ANTHROPIC_API_KEY … not set on startup	Missing env var	Add key to .env
PDF skipped with warning	Scanned/image-only PDF	Use a text-based PDF or convert to .txt
Responses cut off	MAX_TOKENS too low	Increase in config.py
Old content still appearing	Cache TTL (5 min)	Wait or send a new message after the TTL expires