Deep Research

Smart Search SDK supports advanced deep research using the Sonar and Assafelovic engines, with both POST and WebSocket interfaces. These methods allow asynchronous, long-form, or real-time research execution.

Initialization

Environment Setup

Make sure your environment variables are configured correctly in the .env file:

Update your environment configuration file as shown below:

SMARTSEARCH_BASE_URL=<your-api-url>
SMARTSEARCH_TOKEN=<your-auth-token>

Web Socket Initialization

Before using the WebSocket Function, ensure that you have initialized a DummyWebSocket or your own WebSocket implementation. The WebSocket is crucial for receiving real-time data from the engine. Here's an outline of what you need to do:

• Initialize a WebSocket instance: Create an instance of DummyWebSocket or your custom WebSocket class.

• Pass the WebSocket to the research client: When invoking the deep_research method, include this WebSocket instance in the input_data.

Here is the sample code for Dummy WebSocket. Ensure your web socket align with this structure.

class DummyWebSocket:
    def __init__(self):
        self.sent_messages = []

    async def send(self, message):
        print(f"[DummyWebSocket] Sent: {message}")
        self.sent_messages.append(message)

    async def recv(self):
        print("[DummyWebSocket] Receiving dummy message...")
        return json.dumps(
            {
                "status": "success",
                "message": "This is a dummy response.",
                "data": {"query": "Simulated Deep Research Result"},
            }
        )

    async def close(self):
        print("[DummyWebSocket] Closed.")

---

Initialize the Deep Research Client

from smartsearch.deep_research.client import SmartSearchDeepResearchClient

client = SmartSearchDeepResearchClient(
    base_url=os.getenv("SMARTSEARCH_BASE_URL"),
    cache_manager=cache_manager
)

---

Authenticate with Token

await client.authenticate(token=os.getenv("SMARTSEARCH_TOKEN"))

Must be called once before making any search requests.

---

Usage

1. Sonar Deep Research (POST)

Perform deep research via the Sonar engine using a standard POST request.

from smartsearch.deep_research.client import ResearchType

result = await client.deep_research(input_data={
    "type": ResearchType.POST_SONAR,
    "query": "Detailed Analysis"
})

Parameters:

Key	Type	Description
type	enum	Use ResearchType.POST_SONAR
query	string	The topic or question to analyze

---

1. Sonar Deep Research (WebSocket)

Enable real-time result streaming using Sonar via WebSocket.

from smartsearch.deep_research.client import ResearchType
websocket = DummyWebSocket()

await client.deep_research(input_data={
    "type": ResearchType.WS_SONAR,
    "query": "Real-time Analysis",
    "websocket": websocket
})

Parameters:

Key	Type	Description
type	enum	Use ResearchType.WS_SONAR
query	string	Query for research
websocket	DummyWebSocket	An async-compatible WebSocket handler

---

1. Assafelovic Deep Research (POST)

Run in-depth analysis using the Assafelovic engine through POST.

from smartsearch.deep_research.client import ResearchType

result = await client.deep_research(input_data={
    "type": ResearchType.POST_ASSAFELOVIC,
    "query": "In-depth Analysis"
})

Parameters:

Key	Type	Description
type	enum	Use ResearchType.POST_ASSAFELOVIC
query	string	Query or topic to be researched

---

1. Assafelovic Deep Research (WebSocket)

Use WebSocket to stream live responses from the Assafelovic engine.

from smartsearch.deep_research.client import ResearchType
websocket = DummyWebSocket()

await client.deep_research(input_data={
    "type": ResearchType.WS_ASSAFELOVIC,
    "query": "Live Analysis",
    "websocket": websocket
})

Parameters:

Key	Type	Description
type	enum	Use ResearchType.WS_ASSAFELOVIC
query	string	Research query
websocket	DummyWebSocket	WebSocket implementation to receive live results

---

Full Code Snippet

Sonar Deep Research (POST)

from smartsearch.deep_research.client import ResearchType, SmartSearchDeepResearchClient

client = SmartSearchDeepResearchClient(
    base_url=os.getenv("SMARTSEARCH_BASE_URL"),
    cache_manager=cache_manager
)
await client.authenticate(token=os.getenv("SMARTSEARCH_TOKEN"))

result = await client.deep_research(input_data={
    "type": ResearchType.POST_SONAR,
    "query": "Detailed Analysis"
})

Sonar Deep Research (WebSocket)

from smartsearch.deep_research.client import ResearchType, SmartSearchDeepResearchClient

websocket = DummyWebSocket()

client = SmartSearchDeepResearchClient(
    base_url=os.getenv("SMARTSEARCH_BASE_URL"),
    cache_manager=cache_manager
)
await client.authenticate(token=os.getenv("SMARTSEARCH_TOKEN"))

await client.deep_research(input_data={
    "type": ResearchType.WS_SONAR,
    "query": "Real-time Analysis",
    "websocket": websocket
})

Assafelovic Deep Research (POST)

from smartsearch.deep_research.client import ResearchType, SmartSearchDeepResearchClient

client = SmartSearchDeepResearchClient(
    base_url=os.getenv("SMARTSEARCH_BASE_URL"),
    cache_manager=cache_manager
)
await client.authenticate(token=os.getenv("SMARTSEARCH_TOKEN"))

result = await client.deep_research(input_data={
    "type": ResearchType.POST_ASSAFELOVIC,
    "query": "In-depth Analysis"
})

Assafelovic Deep Research (WebSocket)

from smartsearch.deep_research.client import ResearchType, SmartSearchDeepResearchClient

websocket = DummyWebSocket()

client = SmartSearchDeepResearchClient(
    base_url=os.getenv("SMARTSEARCH_BASE_URL"),
    cache_manager=cache_manager
)
await client.authenticate(token=os.getenv("SMARTSEARCH_TOKEN"))

await client.deep_research(input_data={
    "type": ResearchType.WS_ASSAFELOVIC,
    "query": "Live Analysis",
    "websocket": websocket
})