OpenAI Wrapper

Power your OpenAI chat completions with an intuitive Desearch wrapper that automates search.

Desearch provides a AI search and retrieval ecosystem designed for building RAG (Retrieval-Augmented Generation) applications. By leveraging multi-source information from the X and web, Desearch delivers precise, current data to enhance your LLM outputs.

Effective RAG implementation requires more than basic search functionality. It demands intelligent query generation, multi-source integration, and contextual understanding. The Desearch OpenAI wrapper addresses these challenges with a single line of code, transforming any OpenAI chat completion into a comprehensive Desearch-powered RAG system.

Get Started

1. Installation

Install the Desearch and OpenAI Python libraries:

pip install openai desearch-py

2. Set Up Clients

Import and initialize the Desearch and OpenAI clients with your API keys:

from openai import OpenAI
from desearch_py import Desearch

openai = OpenAI(api_key='YOUR_OPENAI_API_KEY')

desearch = Desearch(api_key='YOUR_DESEARCH_API_KEY')

3. Enhance Your OpenAI Client

Use the Desearch.wrap method to enhance your existing OpenAI client with advanced RAG capabilities:

desearch_openai = desearch.wrap(openai)

4. Make Enhanced API Calls

The wrapped client maintains the familiar OpenAI interface while automatically augmenting completions with relevant search results from multiple sources:

completion = desearch_openai.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "What are the latest developments in quantum computing?"}]
)

print(completion.choices[0].message.content)

5. Example Output

Here's a summary of the latest developments in quantum computing:

1. **IBM's Quantum Computing Milestone**
   - IBM announced their new 1,000+ qubit quantum processor, demonstrating significant progress in quantum hardware scaling.
   - The system includes advanced error correction techniques that improve computational stability.

2. **Google's Quantum Supremacy Update**
   - Researchers at Google AI Quantum published results showing their quantum computer performed a calculation in minutes that would take traditional supercomputers thousands of years.
   - The team has made substantial improvements to their error mitigation techniques, bringing practical quantum advantage closer to reality.

3. **Quantum Machine Learning Breakthroughs**
   - Recent research has demonstrated quantum algorithms that can potentially offer exponential speedups for specific machine learning tasks.
   - These advancements could revolutionize areas like drug discovery, materials science, and complex systems modeling.

These developments suggest we're approaching a critical threshold where quantum computing may begin delivering practical advantages for specific computational problems, though general-purpose quantum computing still faces significant challenges.

6. Complete Code Example

Here's a comprehensive example you can copy into a Python script or Jupyter notebook to test the Desearch wrapper:

from openai import OpenAI
from desearch_py import Desearch
from desearch_py.protocol import ToolEnum, ModelEnum

# Initialize clients
openai = OpenAI(api_key='YOUR_OPENAI_API_KEY')
desearch = Desearch(api_key='YOUR_DESEARCH_API_KEY')

# Enhance the OpenAI client
desearch_openai = desearch.wrap(openai)

# Create a completion with RAG capabilities
completion = desearch_openai.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "What are the latest breakthroughs in nuclear fusion?"}]
)

# Display the enhanced response
print(completion.choices[0].message.content)

7. Handling Multiple Questions

Here's how to process multiple questions efficiently with the Desearch wrapper:

# Define a list of questions
questions = [
    "What progress has been made in CRISPR gene editing recently?",
    "How are autonomous vehicles handling extreme weather conditions?",
]

# Process each question with enhanced context
for question in questions:
    completion = desearch_openai.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": question}]
    )
    
    print(f"Question: {question}")
    print(f"Answer: {completion.choices[0].message.content}")
    print("-" * 50)

Information Source Selection

Specify which information sources to include using the tools parameter:

from desearch_py import ToolEnum

completion = desearch_openai.chat.completions.create(
    model="gpt-4",
    messages=messages,
    desearch_tools=[DesearchTool.web, DesearchTool.twitter, DesearchTool.hacker_news]
)

Time Relevance

Filter results by date using the date_filter parameter:

from desearch_py import DateFilterEnum

completion = desearch_openai.chat.completions.create(
    model="gpt-4",
    messages=messages,
    date_filter=DateFilterEnum.PAST_WEEK
)

Response Format

Customize how results are returned with the result_type parameter:

from desearch_py import ResultTypeEnum

completion = desearch_openai.chat.completions.create(
    model="gpt-4",
    messages=messages,
    result_type=ResultTypeEnum.LINKS_WITH_FINAL_SUMMARY
)

Model Selection

Choose which Desearch model to use for processing:

from desearch_py import ModelEnum

completion = desearch_openai.chat.completions.create(
    model="gpt-4",
    messages=messages,
    desearch_model=ModelEnum.HORIZON
)