graph_with_llm

Overview ¶

Example: graph_with_llm — StateGraph with real LLM calls inside nodes.

What you'll learn:

• How to wire a real LLM provider (OpenAI, Anthropic, Gemini, Ollama) into graph nodes
• Building a multi-step workflow where each node calls the LLM for reasoning
• Conditional routing based on LLM classification output
• Using tools inside graph nodes for grounded agent behavior
• Checkpointing and resume with SQLite persistence

Prerequisites:

• Go 1.22+
• At least one LLM API key (or Ollama running locally)

Environment variables (set ONE of these):

OPENAI_API_KEY=sk-...               → uses GPT-4o
ANTHROPIC_API_KEY=sk-ant-...        → uses Claude Sonnet 4
GEMINI_API_KEY=AIza...              → uses Gemini 2.0 Flash
(none)                              → uses Ollama at localhost:11434

Run:

export OPENAI_API_KEY=sk-your-key
go run ./examples/graph_with_llm/

YAML equivalent (same behavior, no Go code):

See examples/yaml-configs/graph-agent.yaml
CHRONOS_CONFIG=examples/yaml-configs/graph-agent.yaml go run ./cli/main.go run "Explain goroutines"