SadBot

A melancholy robot that discusses existentialism. Built with Go (backend) and plain HTML/CSS/JS (frontend). Optionally powered by a local LLM via any OpenAI-compatible inference server.

Quick Start

# Build and run with canned responses
make run

# Open http://localhost:8080

LLM Mode

SadBot can use a local LLM for AI-powered existential conversations. When the LLM is unavailable, it falls back to built-in canned responses.

Supported Servers

Any server exposing /v1/chat/completions (OpenAI-compatible):

Server	Default URL	Make Target
Ollama	http://localhost:11434	make run-ollama
llama.cpp	http://localhost:8080	make run-llama
LM Studio	http://localhost:1234	make run-lmstudio
vLLM	http://localhost:8000	make run-vllm

Usage

# Ollama with llama3
ollama run llama3  # in another terminal
make run-ollama

# Ollama with a different model
make run-ollama MODEL=mistral

# llama.cpp (runs SadBot on port 3000 to avoid conflict)
./llama-server -m model.gguf  # in another terminal
make run-llama

# Custom server
make run-llm LLM_URL=http://localhost:5000 MODEL=my-model

# Full manual control
./sadbot \
  -llm-url=http://localhost:11434 \
  -llm-model=llama3 \
  -llm-temperature=0.9 \
  -llm-max-tokens=256 \
  -llm-timeout=30s \
  -port=3000

CLI Flags

Flag	Default	Description
-port	8080	Server port
-llm-url	(none)	Base URL of the inference server
-llm-model	(none)	Model name (required for Ollama)
-llm-api-key	(none)	API key (most local servers don't need this)
-llm-temperature	0.8	Temperature (0.0-2.0)
-llm-max-tokens	512	Max tokens per response
-llm-timeout	60s	Timeout for LLM requests

API

Endpoint	Method	Description
/api/chat	POST	Send a message, get a response
/api/status	GET	Server status and LLM info
/api/topics	GET	Available existential topics

POST /api/chat

// Request
{ "session_id": "abc123", "message": "Is life absurd?" }

// Response
{
  "text": "Camus would say yes...",
  "mood": "melancholy",
  "turn_count": 3,
  "topics_visited": ["absurdism"],
  "robot_face": ":/",
  "llm_powered": true
}

Development

make build        # Build binary
make test         # Run tests
make test-race    # Run tests with race detector
make vet          # Run go vet
make clean        # Remove binary

Project Structure

cmd/server/main.go           Server entry point, CLI flags, graceful shutdown
internal/robot/sadbot.go      Dialogue engine, mood system, session management
internal/handlers/handlers.go HTTP handlers, rate limiting, input validation
internal/llm/client.go        OpenAI-compatible LLM client with retries
internal/llm/prompt.go        SadBot personality system prompt
static/index.html             Frontend (animated robot, chat UI, starfield)

Tests

60 tests total, all passing with -race:

• internal/robot: 14 tests (dialogue, mood, sessions, concurrency)
• internal/handlers: 11 tests (API, rate limiting, validation)
• internal/llm: 35 tests (client, retries, health checks, concurrency)

Zero Dependencies

The entire project uses only the Go standard library. No external modules.