Build the chat messages list

Background

Every LLM call in this course uses the same shape: a list of role-tagged messages. ResearchAgent.run() builds a system message (the agent's instructions) followed by a user message (the query), then passes them to llm.chat(messages).

[
    {"role": "system", "content": "..."},
    {"role": "user",   "content": "..."},
]

Problem statement

Implement build_messages(system_prompt, user_query) returning the two-message list.

Input

• system_prompt — the system instruction string.
• user_query — the user's question string.

Output

Returns a list of two dicts: a "system" message then a "user" message, each with "role" and "content" keys.

Examples

Example 1

Input:  system_prompt = "You are a research assistant.", user_query = "What is RL?"
Output: [{"role": "system", "content": "You are a research assistant."},
         {"role": "user",   "content": "What is RL?"}]

Constraints

• System message first, user message second.
• Each dict has exactly "role" and "content".

Notes

• Keeping this format fixed is what lets the same agent talk to a mock, OpenAI, or Anthropic without changes.