Use Desearch AI Search with OpenAI - Integrations

Use Desearch AI Search to fetch fresh web results and X posts, then pass them to an OpenAI chat completion as context. This page uses the async desearch-py client and the openai Python package.

Get Started

1. Installation

bash

pip install openai desearch-py==1.2.1

2. Set Up Clients

python

from openai import OpenAI
from desearch_py import Desearch

openai_client = OpenAI(api_key="YOUR_OPENAI_API_KEY")
desearch = Desearch(api_key="YOUR_DESEARCH_API_KEY")

3. Search with Desearch

ai_search is async. It calls POST /desearch/ai/search and returns search (web results) and tweets (X posts) for the tools you pass.

python

import asyncio
from desearch_py import Desearch

async def fetch_sources(question: str):
    async with Desearch(api_key="YOUR_DESEARCH_API_KEY") as desearch:
        return await desearch.ai_search(
            prompt=question,
            tools=["web", "twitter"],
            date_filter=None,
            result_type="ONLY_LINKS",
            count=10,
        )

4. Pass Results to OpenAI

Turn the results into a context block and send it with the user question:

python

def to_context(result) -> str:
    data = result if isinstance(result, dict) else result.model_dump()
    lines = []
    for item in data.get("search") or []:
        lines.append(f"- {item.get('title')}: {item.get('link')}\n  {item.get('snippet', '')}")
    for tweet in data.get("tweets") or []:
        lines.append(f"- X post: {tweet.get('url')}\n  {tweet.get('text', '')}")
    return "\n".join(lines)

question = "What are the latest developments in quantum computing?"
result = asyncio.run(fetch_sources(question))

completion = openai_client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "Answer using the sources below. Cite links.\n\n" + to_context(result)},
        {"role": "user", "content": question},
    ],
)
print(completion.choices[0].message.content)

ai_search can return a typed object or a plain dict, so to_context handles both.

5. Complete Code Example

Copy this into a script or notebook:

python

import asyncio
from openai import OpenAI
from desearch_py import Desearch

openai_client = OpenAI(api_key="YOUR_OPENAI_API_KEY")

async def fetch_sources(question: str):
    async with Desearch(api_key="YOUR_DESEARCH_API_KEY") as desearch:
        return await desearch.ai_search(
            prompt=question,
            tools=["web", "twitter"],
            date_filter=None,
            result_type="ONLY_LINKS",
            count=10,
        )

def to_context(result) -> str:
    data = result if isinstance(result, dict) else result.model_dump()
    lines = []
    for item in data.get("search") or []:
        lines.append(f"- {item.get('title')}: {item.get('link')}\n  {item.get('snippet', '')}")
    for tweet in data.get("tweets") or []:
        lines.append(f"- X post: {tweet.get('url')}\n  {tweet.get('text', '')}")
    return "\n".join(lines)

def answer(question: str) -> str:
    result = asyncio.run(fetch_sources(question))
    completion = openai_client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": "Answer using the sources below. Cite links.\n\n" + to_context(result)},
            {"role": "user", "content": question},
        ],
    )
    return completion.choices[0].message.content

print(answer("What are the latest breakthroughs in nuclear fusion?"))

6. Handling Multiple Questions

python

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

for question in questions:
    print(f"Question: {question}")
    print(f"Answer: {answer(question)}")
    print("-" * 50)

The snippets below use await, so run them inside an async function like fetch_sources above.

Information Source Selection

Pick which sources to search with the tools parameter of ai_search:

python

result = await desearch.ai_search(
    prompt=question,
    tools=["web", "twitter"],
    date_filter=None,
)

Supported tools: web, twitter. Other values are unsupported and may be rejected (for example, youtube returns 422).

Time Relevance

Filter results by date with the date_filter parameter. On AI Search, date filtering applies to X results.

python

result = await desearch.ai_search(
    prompt="What are people saying about quantum computing this week?",
    tools=["twitter"],
    date_filter="PAST_WEEK",
)

Response Format

Control how results are returned with the result_type parameter:

python

result = await desearch.ai_search(
    prompt=question,
    tools=["web", "twitter"],
    date_filter=None,
    result_type="LINKS_WITH_FINAL_SUMMARY",
)

With LINKS_WITH_FINAL_SUMMARY, the response also includes a written summary in completion.