Context Injection

ContextInjection fills {placeholder} templates in your prompt from the active session context before each model call. Write {user_name} in a system or user message, put user_name in the flow context, and the model sees the resolved value. It is model-level only (model_middleware=).

Usage

import railtracks as rt
from railtracks.prebuilt.middleware import ContextInjection


# ContextInjection is model-level only. It fills {placeholders} in the prompt
# from the active session context before each model call.
CtxAgent = rt.agent_node(
    name="context-injection-demo",
    llm=rt.llm.OpenAILLM("gpt-4o"),
    system_message="You are helping {user_name}. Keep answers short.",
    model_middleware=[ContextInjection()],
)

flow = rt.Flow(
    "ContextInjectionFlow",
    entry_point=CtxAgent,
    context={"user_name": "Alex"},
)
# flow.invoke("Who are you helping?")  ->  the model sees "You are helping Alex."

Ordering

List position matters: place ContextInjection before (outside) any middleware that must see the injected prompt. For example, an input guard listed after it will see the filled-in template rather than the raw {placeholder}.