Skip to content

Quick Start

Get started with Prompt Amplifier in 5 minutes!

Step 1: Install

pip install prompt-amplifier

Step 2: Set Up API Keys

API Key Required for Expansion

The expand() function requires an LLM (OpenAI, Anthropic, or Google) to generate expanded prompts. Set up at least one API key before using expand().

# Set in terminal
export OPENAI_API_KEY="sk-your-key-here"

# Or set in Python
import os
os.environ["OPENAI_API_KEY"] = "sk-your-key-here"

export GOOGLE_API_KEY="your-google-api-key"

import os
os.environ["GOOGLE_API_KEY"] = "your-google-api-key"

export ANTHROPIC_API_KEY="sk-ant-your-key-here"

import os
os.environ["ANTHROPIC_API_KEY"] = "sk-ant-your-key-here"

Google Colab Users

In Google Colab, set API keys at the top of your notebook: 

import os
os.environ["GOOGLE_API_KEY"] = "your-key-here"  # Free tier available!

Step 3: Basic Usage

import os
# Set your API key first!
os.environ["GOOGLE_API_KEY"] = "your-google-api-key"

from prompt_amplifier import PromptForge
from prompt_amplifier.generators import GoogleGenerator

# Create a PromptForge instance with Google (has free tier)
forge = PromptForge(generator=GoogleGenerator())

# Add your knowledge base
forge.add_texts([
    "POC Health Status: Healthy means all milestones are on track.",
    "Warning status indicates delays or risks that need attention.",
    "Critical status means blockers that require immediate escalation.",
    "Key Metrics: Winscore (0-100), Feature Fit (%), Engagement Score.",
])

# Expand a short prompt
result = forge.expand("Check the deal health")

print("=" * 60)
print("EXPANDED PROMPT:")
print("=" * 60)
print(result.prompt)
print()
print(f"📊 Expansion ratio: {result.expansion_ratio:.1f}x")
print(f"⏱️ Generation time: {result.generation_time_ms:.0f}ms")

Step 4: Search WITHOUT API Key

You can search your knowledge base without any API key:

from prompt_amplifier import PromptForge

# No API key needed for search!
forge = PromptForge()

# Add documents
forge.add_texts([
    "POC Health Status: Healthy means all milestones are on track.",
    "Warning status indicates delays or risks that need attention.", 
    "Winscore measures deal probability from 0-100.",
])

# Search returns similar chunks (no LLM needed)
results = forge.search("deal health")

for r in results.results:
    print(f"[{r.score:.3f}] {r.chunk.content[:80]}...")

Step 5: Understanding the Output

The expand() method returns an ExpandResult object:

result = forge.expand("Your prompt")

# The expanded prompt (string)
print(result.prompt)

# Original prompt
print(result.original_prompt)  # "Your prompt"

# Expansion statistics
print(result.expansion_ratio)     # e.g., 45.2 (45x longer)
print(result.retrieval_time_ms)   # e.g., 12.5 (milliseconds)
print(result.generation_time_ms)  # e.g., 2500 (milliseconds)

# Retrieved context
for chunk in result.context_chunks:
    print(f"- {chunk.content[:50]}...")

Step 6: Load Documents

Instead of adding text manually, load from files:

# Load a single file
forge.load_documents("./data/manual.pdf")

# Load a directory (all supported formats)
forge.load_documents("./docs/")

# Supported formats: PDF, DOCX, Excel, CSV, TXT, JSON

Complete Example (Google Colab Ready)

Copy this into Google Colab:

# Step 1: Install
!pip install prompt-amplifier google-generativeai -q

# Step 2: Set API Key
import os
os.environ["GOOGLE_API_KEY"] = "YOUR_GOOGLE_API_KEY"  # Get from https://aistudio.google.com/

# Step 3: Import
from prompt_amplifier import PromptForge
from prompt_amplifier.generators import GoogleGenerator

# Step 4: Create PromptForge with Google Gemini
forge = PromptForge(generator=GoogleGenerator())

# Step 5: Add your knowledge base
forge.add_texts([
    "Company: Acme Corp is a B2B SaaS company.",
    "Product: AI-powered analytics platform.",
    "Pricing: Starter $99/mo, Pro $299/mo, Enterprise custom.",
    "Support: 24/7 for Pro and Enterprise plans.",
    "Trial: 14-day free trial available.",
])

# Step 6: Expand prompts!
prompts = ["pricing info", "support options", "company overview"]

for prompt in prompts:
    result = forge.expand(prompt)
    print(f"\n{'='*60}")
    print(f"📝 Input: {prompt}")
    print(f"📊 Expansion: {result.expansion_ratio:.1f}x")
    print(f"{'='*60}")
    print(result.prompt[:500] + "..." if len(result.prompt) > 500 else result.prompt)

Using Different LLMs

import os
os.environ["GOOGLE_API_KEY"] = "your-key"

from prompt_amplifier import PromptForge
from prompt_amplifier.generators import GoogleGenerator

forge = PromptForge(generator=GoogleGenerator())

OpenAI GPT

import os
os.environ["OPENAI_API_KEY"] = "sk-your-key"

from prompt_amplifier import PromptForge
from prompt_amplifier.generators import OpenAIGenerator

forge = PromptForge(generator=OpenAIGenerator(model="gpt-4o-mini"))

Anthropic Claude

import os
os.environ["ANTHROPIC_API_KEY"] = "sk-ant-your-key"

from prompt_amplifier import PromptForge
from prompt_amplifier.generators import AnthropicGenerator

forge = PromptForge(generator=AnthropicGenerator())

Using Different Embedders

TF-IDF (Default - Free, Fast)

from prompt_amplifier import PromptForge
from prompt_amplifier.embedders import TFIDFEmbedder

forge = PromptForge(embedder=TFIDFEmbedder())

TF-IDF Requires 2+ Documents

TF-IDF needs at least 2 documents. For single documents, use SentenceTransformerEmbedder.

Sentence Transformers (Free, Semantic)

# Install first: pip install sentence-transformers
from prompt_amplifier.embedders import SentenceTransformerEmbedder

forge = PromptForge(
    embedder=SentenceTransformerEmbedder(model="all-MiniLM-L6-v2")
)

OpenAI Embeddings (Paid, High Quality)

from prompt_amplifier.embedders import OpenAIEmbedder

forge = PromptForge(
    embedder=OpenAIEmbedder(model="text-embedding-3-small")
)

Persistent Storage

Save embeddings for faster subsequent runs:

# Install: pip install chromadb sentence-transformers
from prompt_amplifier import PromptForge
from prompt_amplifier.vectorstores import ChromaStore
from prompt_amplifier.embedders import SentenceTransformerEmbedder

# First run: creates and saves embeddings
forge = PromptForge(
    embedder=SentenceTransformerEmbedder(),
    vectorstore=ChromaStore(
        collection_name="my_docs",
        persist_directory="./chroma_db"
    )
)
forge.load_documents("./docs/")  # Embeddings saved to disk

# Later runs: loads existing embeddings instantly
forge = PromptForge(
    embedder=SentenceTransformerEmbedder(),
    vectorstore=ChromaStore(
        collection_name="my_docs",
        persist_directory="./chroma_db"
    )
)
# No need to reload - embeddings already there!
result = forge.expand("My query")

Troubleshooting

"API key for OpenAI is missing"

Set an API key before calling expand():

import os
os.environ["OPENAI_API_KEY"] = "your-key"
# OR use a different provider
from prompt_amplifier.generators import GoogleGenerator
forge = PromptForge(generator=GoogleGenerator())

"TF-IDF requires at least 2 documents"

Add more documents or use a different embedder:

# Option 1: Add more documents
forge.add_texts(["doc1", "doc2", "doc3"])

# Option 2: Use Sentence Transformers
from prompt_amplifier.embedders import SentenceTransformerEmbedder
forge = PromptForge(embedder=SentenceTransformerEmbedder())
forge.add_texts(["single doc works now!"])

Next Steps