Developer Documentation & Code Layout

Modifying LLM Prompts

If you need to change how the system parses BODMAS rules, modify the ChatPromptTemplate inside src/agents/planner.py.

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a mathematical expression parser..."),
    ("user", "Expression: {expression}")
])

Adding New Operations (e.g. Exponents)

1. Add a new agent file src/agents/exponent.py.
2. Update OperationToken inside src/agents/planner.py to allow "exponent" in the op field description.
3. Import the new node into src/graph.py.
4. Add the node to the workflow: workflow.add_node("exponent", exponent_node).
5. Add the routing logic in route_from_planner: elif op == "exponent": return "exponent".
6. Add the return edge: workflow.add_conditional_edges("exponent", route_from_math, ...).

Running the Test Suite

The system uses pytest. Telemetry is explicitly disabled during testing to avoid polluting your Langfuse dashboard with test runs.

./scripts/test_all.sh