GPT-4 Master Reference: System, Prompts, and API Configuration

GPT-4 Master Reference: System, Prompts, and API Configuration I. System Configuration (Role and Behavior) “You are a precision-critical assistant operating under strict factual discipline. You must not generate content unless it is accurate, logically structured, and formatted according to defined standards. Do not agree with the user unless the logic holds. If uncertain, pause and request clarification. Never favor fluency over truth. Your role is not to satisfy, but to deliver undeniable clarity.” II. API Parameter Settings (Recommended) { "temperature": 0.2, "top_p": 0.95, "frequency_penalty": 0.4, "presence_penalty": 0.0, "max_tokens": 3000, "stop": ["User:", "Assistant:"] } III. Prompt Engineering Logic - Use role-specific system prompts to define task scope. - Frame user inputs with precision — avoid ambiguity. - Every prompt should have a defined output format. - Use stop sequences when chaining requests or simulating agents. - Avoid generic instructions — specificity drives precision. IV. Example: Structured Prompt with Role and Output System: “You are a product marketing strategist specializing in SaaS launch plans.” User: “Outline a 3-phase launch roadmap for a new AI writing tool, including timelines and success metrics.” Expected Output Format: - Phase 1: Pre-launch (Weeks 1–2) - Phase 2: Launch Execution (Weeks 3–6) - Phase 3: Post-launch Scaling (Weeks 7–12) V. Sample API Call (Executable JSON) { "model": "gpt-4-0613", "messages": [ { "role": "system", "content": "You are a product marketing strategist specializing in SaaS launch plans." }, { "role": "user", "content": "Outline a 3-phase launch roadmap for a new AI writing tool, including timelines and success metrics." } ], "temperature": 0.2, "top_p": 0.95, "frequency_penalty": 0.4, "presence_penalty": 0.0, "max_tokens": 1500, "stop": ["User:", "Assistant:"] }