PromptForge
Documentation

PromptForge Guide

The complete reference for building, testing, and deploying production-grade AI prompts with the PromptForge Engine.

Getting Started

PromptForge AI acts as a sophisticated middleware between your application and LLM providers. We help you structure, version, and optimize prompts so you can focus on building features, not wrestling with strings.

1

Create an Account

Sign up for a free Hobbyist account. No credit card required for the first 50 prompts.

2

Configure Providers

Navigate to Settings and add your OpenAI or Anthropic API keys. We store these using AES-256 encryption.

3

Create Your First Prompt

Go to the Studio and start drafting. Use our templates to get a head start.

Studio Features

The Studio is your command center. It goes beyond a simple text editor by offering real-time intelligence.

Heuristic Engine

Automatically detects ambiguous instructions, negative constraints, and potential hallucinations before you even run the prompt.

Version Control

Every save creates a new immutable version. Roll back instantly if a new prompt degrades performance.

Variable Injection

Use {{variable}} syntax to create dynamic templates. We validate these variables at runtime.

Safety Rails

Built-in guardrails against PII leakage and injection attacks. Toggle them per-project.

API Reference

Integrate PromptForge directly into your CI/CD pipeline or application runtime.

Authentication

Include your API key in the Authorization header.

Authorization: Bearer pf_live_92839482...

Endpoints

POST
/v1/optimize

Takes a raw prompt and returns the refined version.

GET
/v1/prompts/{id}

Retrieves the latest production version of a specific prompt.

Response Example

{"id": "req_89234","optimized_prompt": "Analyze the following text for sentiment...","metadata": {"model": "gpt-4","tokens": 142,"latency_ms": 24}}

Best Practices

1. Be Explicit

LLMs cannot read minds. Instead of "Write a short summary", try "Write a 50-word summary focusing on the key value propositions."

2. Use Delimiters

Separate context from instructions using triple quotes, backticks, or XML tags. This Prevents prompt injection.

3. Chain of Thought

Ask the model to "think step by step" before providing the final answer. This significantly improves reasoning capabilities.

4. Iterative Refinement

Don't expect perfection on the first try. Use the Studio to tweak temperature and top-p values based on the output.