Context Engineering Template for AI Systems

1. Define System Role and Instructions

• Specify the AI model’s overall role, behavior, and constraints.
  Example: “You are an expert sustainability analyst providing concise, data-backed insights.”

2. Collect User Input and Interaction History

• Include the latest user query plus relevant recent conversation context for continuity.
  Example: “User asked about carbon footprint reduction; previous discussion included transport emissions.”

3. Retrieve Relevant External Knowledge

• Identify and fetch pertinent documents, datasets, or knowledge base excerpts.
  Example: “Pull latest supply chain emission reports and recent research summaries.”

4. Incorporate User Profile and Preferences

• Add personalized data such as user role, expertise level, and interests.
  Example: “User is a policy advisor focusing on European regulations.”

5. Prepare Tool and API Context

• Provide information on callable tools, APIs, or modules the AI can use.
  Example: “AI can request live emissions data or access compliance databases through API calls.”

6. Context Formatting and Prioritization

• Organize, summarize, and truncate inputs to fit the model’s token limit while preserving priority info.
  Example: “Prioritize recent regulatory updates and critical user questions; omit redundant background.”

7. Specify Expected Output Formats or Constraints

• Define response style, structure, and length limits.
  Example: “Output as bullet points with citations, max 300 words.”

Example of a Filled Template

1. Role: Sustainability analyst specializing in supply chain emissions.
2. User Input: User asks, “How can I reduce transportation emissions in my logistics network?” Conversation history includes discussion of current fleet fuel use.
3. External Knowledge: Latest supply chain carbon reports 2023, lifecycle assessment data.
4. User Profile: Logistics manager with intermediate AI familiarity.
5. Tools: Access to real-time vehicle tracking API and carbon calculator API.
6. Formatting: Summarize key strategies; list pros and cons in bullets; limit to 5 points.
7. Expected Output: Bullet points with citations and actionable recommendations.