> ## Documentation Index
> Fetch the complete documentation index at: https://developers.telnyx.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Start research task

> Starts a deep research task that runs multiple searches, reads sources, and synthesizes an answer with citations.

## Synchronous mode (default)

When `background` is `false` or omitted, the request blocks until the research completes and returns the answer with citations. This can take up to 120 seconds depending on `research_effort`.

## Asynchronous mode

When `background` is `true`, the request returns immediately with a `task_id` and `status: pending`. Poll `GET /web_search/research/{task_id}` to check when the research completes and retrieve the answer.



## OpenAPI

````yaml /openapi/source/external/web-search/web-search.json post /web_search/research
openapi: 3.1.0
info:
  version: 1.0.0
  title: Telnyx API
  x-latency-category: interactive
  x-endpoint-cost: light
  description: >-
    Real-time web search, page content retrieval, and deep research with
    citations, exposed through the Telnyx API Gateway and powered by the
    `telnyx-agent-web-search` service.


    ## Authentication


    All endpoints require a Telnyx API key sent as a Bearer token in the
    `Authorization` header. The API Gateway validates the key and forwards the
    verified identity to the backend service as `Telnyx-Auth-Rev2`.


    ## Rate Limits


    Default rate limit: 60 requests per minute per API key.


    ## Error Responses


    Errors follow the format `{ "error": { "message": string, "details": object
    } }`.
  contact:
    email: support@telnyx.com
servers:
  - url: https://api.telnyx.com/v2
    description: Telnyx API v2
security:
  - bearerAuth: []
tags:
  - name: Web Search
    description: Real-time web search returning structured, LLM-ready JSON results.
  - name: Contents
    description: Page content retrieval for URLs.
  - name: Research
    description: Deep research with citations and async task polling.
paths:
  /web_search/research:
    post:
      tags:
        - Research
      summary: Start research task
      description: >-
        Starts a deep research task that runs multiple searches, reads sources,
        and synthesizes an answer with citations.


        ## Synchronous mode (default)


        When `background` is `false` or omitted, the request blocks until the
        research completes and returns the answer with citations. This can take
        up to 120 seconds depending on `research_effort`.


        ## Asynchronous mode


        When `background` is `true`, the request returns immediately with a
        `task_id` and `status: pending`. Poll `GET
        /web_search/research/{task_id}` to check when the research completes and
        retrieve the answer.
      operationId: CreateWebSearchResearch
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ResearchRequest'
            example:
              query: >-
                Compare the performance of RAG vs fine-tuning for
                domain-specific QA
              research_effort: standard
              max_sources: 20
              background: false
      responses:
        '200':
          description: >-
            Research response. Shape depends on `background`:


            - **Synchronous** (`background` false/unset): returns `answer` +
            `citations`.

            - **Asynchronous** (`background` true): returns `task_id` +
            `status`.
          content:
            application/json:
              schema:
                type: object
                properties:
                  data:
                    oneOf:
                      - $ref: '#/components/schemas/ResearchResponseSync'
                      - $ref: '#/components/schemas/ResearchResponseAsync'
              examples:
                sync:
                  summary: Synchronous response (background=false)
                  value:
                    data:
                      answer: RAG and fine-tuning serve different purposes...
                      citations:
                        - url: https://arxiv.org/abs/2401.15884
                          title: >-
                            Retrieval-Augmented Generation for
                            Knowledge-Intensive NLP Tasks
                          snippet: >-
                            We show that RAG models produce more factually
                            grounded responses...
                async:
                  summary: Asynchronous response (background=true)
                  value:
                    data:
                      task_id: bf3026a5-dd57-44dd-b922-200041be3a4b
                      status: pending
        '400':
          $ref: '#/components/responses/BadRequest'
        '401':
          $ref: '#/components/responses/Unauthorized'
        '500':
          $ref: '#/components/responses/InternalServerError'
        '502':
          $ref: '#/components/responses/ProviderError'
        '504':
          $ref: '#/components/responses/ProviderTimeout'
      x-codeSamples:
        - lang: JavaScript
          source: |-
            import Telnyx from 'telnyx';

            const client = new Telnyx({
              apiKey: process.env['TELNYX_API_KEY'], // This is the default and can be omitted
            });

            const response = await client.webSearch.research.create({
              query: 'Compare the performance of RAG vs fine-tuning for domain-specific QA',
            });

            console.log(response.data);
        - lang: Python
          source: |
            import os
            from telnyx import Telnyx

            client = Telnyx(
                api_key=os.environ.get("TELNYX_API_KEY"),  # This is the default and can be omitted
            )
            response = client.web_search.research.create(
                query="Compare the performance of RAG vs fine-tuning for domain-specific QA",
            )
            print(response.data)
        - lang: Go
          source: "package main\n\nimport (\n\t\"context\"\n\t\"fmt\"\n\n\t\"github.com/team-telnyx/telnyx-go\"\n\t\"github.com/team-telnyx/telnyx-go/option\"\n)\n\nfunc main() {\n\tclient := telnyx.NewClient(\n\t\toption.WithAPIKey(\"My API Key\"),\n\t)\n\tresearch, err := client.WebSearch.Research.New(\n\t\tcontext.TODO(),\n\t\ttelnyx.WebSearchResearchNewParams{\n\t\t\tQuery: \"Compare the performance of RAG vs fine-tuning for domain-specific QA\",\n\t\t},\n\t)\n\tif err != nil {\n\t\tpanic(err.Error())\n\t}\n\tfmt.Printf(\"%+v\\n\", research.Data)\n}\n"
        - lang: Java
          source: >-
            package com.telnyx.sdk.example;


            import com.telnyx.sdk.client.TelnyxClient;

            import com.telnyx.sdk.client.okhttp.TelnyxOkHttpClient;

            import
            com.telnyx.sdk.models.webSearch.research.ResearchCreateParams;


            public final class Main {
                private Main() {}

                public static void main(String[] args) {
                    TelnyxClient client = TelnyxOkHttpClient.fromEnv();

                    ResearchCreateParams params = ResearchCreateParams.builder()
                        .query("Compare the performance of RAG vs fine-tuning for domain-specific QA")
                        .build();
                    var response = client.webSearch().research().create(params);
                }
            }
        - lang: Ruby
          source: >-
            require "telnyx"


            telnyx = Telnyx::Client.new(api_key: "My API Key")


            response = telnyx.web_search.research.create(query: "Compare the
            performance of RAG vs fine-tuning for domain-specific QA")


            puts(response)
        - lang: PHP
          source: >-
            <?php


            require_once dirname(__DIR__) . '/vendor/autoload.php';


            use Telnyx\Client;

            use Telnyx\Core\Exceptions\APIException;


            $client = new Client(apiKey: getenv('TELNYX_API_KEY') ?: 'My API
            Key');


            try {
              $response = $client->webSearch->research->create(
                query: 'Compare the performance of RAG vs fine-tuning for domain-specific QA',
              );

              var_dump($response);
            } catch (APIException $e) {
              echo $e->getMessage();
            }
        - lang: CLI
          source: |-
            telnyx web-search:research create \
              --api-key 'My API Key' \
              --query 'Compare the performance of RAG vs fine-tuning for domain-specific QA'
components:
  schemas:
    ResearchRequest:
      type: object
      required:
        - query
      properties:
        query:
          type: string
          minLength: 1
          maxLength: 2000
          description: The research question or topic.
          example: Compare the performance of RAG vs fine-tuning for domain-specific QA
        research_effort:
          type: string
          enum:
            - lite
            - standard
            - deep
          description: Research depth level. `lite` is fastest, `deep` is most thorough.
          example: standard
        max_sources:
          type: integer
          minimum: 1
          maximum: 50
          description: Maximum number of sources to use.
          example: 20
        background:
          type: boolean
          description: >-
            When `true`, the research runs asynchronously. The response returns
            a `task_id` immediately instead of waiting for the result. Poll `GET
            /web_search/research/{task_id}` to check status.
          example: false
    ResearchResponseSync:
      type: object
      description: Synchronous research response (when `background` is false or unset).
      required:
        - answer
      properties:
        answer:
          type: string
          description: The synthesized research answer.
          example: RAG and fine-tuning serve different purposes...
        citations:
          type: array
          items:
            $ref: '#/components/schemas/ResearchCitation'
          description: Sources cited in the answer.
    ResearchResponseAsync:
      type: object
      description: Asynchronous research response (when `background` is true).
      required:
        - task_id
        - status
      properties:
        task_id:
          type: string
          description: >-
            Unique identifier for the research task. Use this to poll the
            status.
          example: bf3026a5-dd57-44dd-b922-200041be3a4b
        status:
          type: string
          enum:
            - pending
            - running
            - completed
            - failed
          description: Current status of the research task.
          example: pending
    ResearchCitation:
      type: object
      required:
        - url
        - title
      properties:
        url:
          type: string
          format: uri
          description: Source URL.
        title:
          type: string
          description: Title of the source page.
        snippet:
          type: string
          description: Relevant excerpt from the source (if available).
    WebSearchError:
      type: object
      properties:
        error:
          type: object
          required:
            - message
          properties:
            message:
              type: string
              description: Human-readable error message.
            details:
              type: object
              additionalProperties: true
              description: Additional error details (e.g. validation field errors).
    GatewayError:
      type: object
      description: >-
        Standard Telnyx JSON:API error envelope returned by the API Gateway for
        authentication failures (401).
      required:
        - errors
      properties:
        errors:
          type: array
          items:
            type: object
            required:
              - code
              - title
            properties:
              code:
                type: string
                description: Telnyx error code.
              title:
                type: string
                description: Error title.
              detail:
                type: string
                description: Human-readable error detail.
              source:
                type: object
                properties:
                  pointer:
                    type: string
              meta:
                type: object
                properties:
                  url:
                    type: string
                    format: uri
  responses:
    BadRequest:
      description: Invalid request — validation error or invalid parameters.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/WebSearchError'
          example:
            error:
              message: Validation error
              details: {}
    Unauthorized:
      description: >-
        Unauthorized — missing or invalid API key.


        The API Gateway returns this response before the request reaches the
        backend service. The error format follows the standard Telnyx JSON:API
        error envelope with `errors[]`, not the backend-level `WebSearchError`
        shape.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/GatewayError'
          example:
            errors:
              - code: '10009'
                title: Authentication failed
                detail: Could not find any usable credentials in the request.
                meta:
                  url: https://developers.telnyx.com/docs/overview/errors/10009
    InternalServerError:
      description: Internal server error.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/WebSearchError'
          example:
            error:
              message: Internal server error
    ProviderError:
      description: The upstream search provider returned an error.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/WebSearchError'
          example:
            error:
              message: Provider request failed
    ProviderTimeout:
      description: The upstream search provider timed out.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/WebSearchError'
          example:
            error:
              message: Provider request timed out
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: Telnyx API key

````