View Source LangChain.Tools.DeepResearchClient (LangChain v0.4.0)

HTTP client for OpenAI Deep Research API.

This module handles the three main operations for OpenAI Deep Research:

  1. Creating a research request
  2. Checking request status
  3. Retrieving completed results

Uses the same authentication and HTTP client patterns as other OpenAI integrations in LangChain, reusing the existing :openai_key configuration.

Configuration

The endpoint can be customized for alternative providers (Azure, OpenRouter, etc.):

# In config.exs or runtime.exs
config :langchain, :deep_research_endpoint, "https://custom-endpoint.com/v1/responses"

If not configured, defaults to OpenAI's standard endpoint.

Summary

Functions

check_status(request_id, endpoint \\ nil)

Checks the status of a deep research request.

create_research(query, options \\ %{})

Creates a new deep research request.

get_results(request_id, endpoint \\ nil)

Retrieves the results of a completed deep research request.

Functions

Link to this function

check_status(request_id, endpoint \\ nil)

View Source 
@spec check_status(String.t(), String.t() | nil) ::
  {:ok, map()} | {:error, String.t()}

Checks the status of a deep research request.

Parameters

  • request_id: The ID returned from create_research/2
  • endpoint: Optional custom endpoint URL

Returns

  • {:ok, status_map} containing status information
  • {:error, reason} on failure
Link to this function

create_research(query, options \\ %{})

View Source 
@spec create_research(String.t(), map()) :: {:ok, String.t()} | {:error, String.t()}

Creates a new deep research request.

Parameters

  • query: The research question or topic
  • options: Optional parameters including:
    • :model - The model to use (defaults to "o3-deep-research-2025-06-26")
    • :system_message - Optional guidance for research approach
    • :max_tool_calls - Maximum number of tool calls to make
    • :summary - Summary mode: "auto" or "detailed" (defaults to "auto")
    • :include_code_interpreter - Include code interpreter tool (defaults to true)
    • :endpoint - Custom endpoint URL (defaults to "https://api.openai.com/v1/responses")

Returns

  • {:ok, request_id} on success
  • {:error, reason} on failure
Link to this function

get_results(request_id, endpoint \\ nil)

View Source 
@spec get_results(String.t(), String.t() | nil) :: {:ok, map()} | {:error, String.t()}

Retrieves the results of a completed deep research request.

Parameters

  • request_id: The ID of the completed research request
  • endpoint: Optional custom endpoint URL

Returns

  • {:ok, result_map} containing the research findings and metadata
  • {:error, reason} on failure