Context

Context is a pre-formatted string built from recalled memories. It's designed to be injected directly into an LLM prompt as system context.

Format

[preferences:fact-1]
User is vegetarian

[preferences:fact-2]
Allergic to peanuts

Each memory is labeled with its [namespace:key] so the LLM understands where the information came from.

Usage

When you call recall_memory, you get back a context string along with the individual items:

ctx = client.recall_memory(
    namespace="preferences",
    prompt="What does the user eat?",
    num_chunks=10,
)

# Use the formatted context in your own LLM call
messages = [
    {"role": "system", "content": f"Use this context:\n{ctx.context}"},
    {"role": "user", "content": "What should I order for lunch?"},
]

Or Let Neocortex Handle It

If you don't want to manage the LLM call yourself, use recall_with_llm it fetches context and queries the LLM in one step:

response = client.recall_with_llm(
    prompt="What should I order for lunch?",
    namespace="preferences",
    api_key="your-openai-key",
)
print(response.text)