> ## Documentation Index > Fetch the complete documentation index at: https://docs.syllable.ai/llms.txt > Use this file to discover all available pages before exploring further. # Start Experiment > Start an experiment. (Only an experiment in 'draft' status can be started.) ## OpenAPI ````yaml https://spec.speakeasy.com/syllable/syllable/syllable-sdk-oas-with-code-samples post /api/v1/experiments/{experiment_id}/start openapi: 3.1.0 info: title: SyllableSDK description: > # Syllable Platform SDK Syllable SDK gives you the power of awesome AI agentry. 🚀 ## Overview The Syllable SDK provides a comprehensive set of tools and APIs to integrate powerful AI capabilities into your communication applications. Whether you're building phone agents, chatbots, virtual assistants, or any other AI-driven solutions, Syllable SDK has got you covered. ## Features - **Agent Configuration**: Create and manage agents that can interact with users across various channels. - **Channel Management**: Configure channels like SMS, web chat, and more to connect agents with users. - **Custom Messages**: Set up custom messages that agents can deliver as greetings or responses. - **Conversations**: Track and manage conversations between users and agents, including session management. - **Tools and Workflows**: Leverage tools and workflows to enhance agent capabilities, such as data processing and API calls. - **Data Sources**: Integrate data sources to provide agents with additional context and information. - **Insights and Analytics**: Analyze conversations and sessions to gain insights into user interactions. - **Permissions and Security**: Manage permissions to control access to various features and functionalities. - **Language Support**: Define language groups to enable multilingual support for agents. - **Outbound Campaigns**: Create and manage outbound communication campaigns to reach users effectively. - **Session Labels**: Label sessions with evaluations of quality and descriptions of issues encountered. - **Incident Management**: Track and manage incidents related to agent interactions. version: 0.0.3 servers: - url: https://api.syllable.cloud description: API server security: [] tags: - name: agents description: >- Operations related to agent configuration. When a user interacts with the Syllable system, they do so by communicating with an agent. An agent is linked to a prompt, a custom message, and one or more channel targets to define its behavior and capabilities. For more information, see [Console docs](https://docs.syllable.ai/workspaces/Agents). - name: agents.test description: >- Operations for testing agents with live text. These endpoints allow sending messages to an agent and receiving its responses. - name: channels description: >- Operations related to channel configuration. A channel is an organization-level point of communication, like a phone number or a web chat. A channel can be associated with an agent by creating a channel target linking them. - name: channels.twilio description: Operations related to Twilio channel configuration. - name: channels.twilio.numbers description: >- Operations related to setting up phone numbers in Twilio for use in channels. - name: channels.targets description: >- Operations related to channel target configuration. A channel target links a channel to an agent, allowing users to communicate with the agent through that channel. For more information, see [Console docs](https://docs.syllable.ai/Resources/Channels). - name: conversations description: >- Operations related to conversations. A conversation is a record of messages between a user and an agent, and is composed of one or more sessions. - name: custom_messages description: >- Operations related to custom message configuration. A custom message is a pre-configured message delivered by an agent as a greeting at the beginning of a conversation. Multiple agents can use the same custom mesasage. A custom message has one or more rules defined, which allow for different messages to be dynamically selected and delivered at runtime based on the current time and either date or day of the week. For more information, see [Console docs](https://docs.syllable.ai/Resources/Messages). - name: dashboards description: >- Operations related to dashboards. Currently the API/SDK only supports fetching basic information about dashboards. - name: data_sources description: >- Operations related to data sources. A data source is a blob of text that can be made available to an agent's general info tools to provide more context to the agent when generating its responses. For more information, see [Console docs](https://docs.syllable.ai/Resources/DataSources). - name: directory description: Operations related to directory - name: events description: >- Operations related to events. An event represents a specific occurrence during a session. Currently the API/SDK only supports fetching logged events. - name: incidents description: Operations related to incidents. - name: insights description: >- Operations related to insights results. An insight is a tool that processes conversation data to extract information and generate reports. - name: insights.folders description: >- Operations related to insights upload folders. An insight folder is used to upload call recordings for insight workflow analysis. - name: insights.tools description: >- Operations related to insights tool configurationss. An insight is a tool that processes conversation data to extract information and generate reports. - name: insights.workflows description: >- Operations related to insights workflows. An workflow is series of tool invocations that processes conversation data to extract information and generate reports. - name: language_groups description: >- Operations related to language groups. A language group is a collection of language, voice, and DTMF configuration that can be linked to an agent to define the languages and voices it supports. For more information, see [Console docs](https://docs.syllable.ai/Resources/LanguageGroups). - name: organizations description: Operations related to organizations. - name: outbound.campaigns description: Operations related to outbound message campaigns - name: outbound.batches description: Operations related to outbound campaign batches - name: permissions description: >- Operations related to permissions. A permission is a specific capability or access level granted to a user within the Syllable system. Permissions are used to control access to various features and functionalities. - name: prompts description: >- Operations related to prompts. A prompt defines the behavior of an agent by delivering instructions to the LLM about how the agent should behave. A prompt can be linked to one or more agents. A prompt can also be linked to tools to allow an agent using the prompt to use them. For more information, see [Console docs](https://docs.syllable.ai/Resources/Prompts). - name: roles description: >- Operations related to roles. A role is a collection of permissions that can be assigned to users to control their access to various features within the Syllable system. - name: services description: >- Operations related to service configuration. A service is a collection of tools. You can specify an authentication method and values on a service, and any linked tools will use that auth information to generate headers for HTTP calls. - name: sessions description: >- Operations related to sessions. A session is a building block of a conversation. For more information, see [Console docs](https://docs.syllable.ai/workspaces/Sessions). - name: session_labels description: >- Operations related to labeling sessions with evaluations of quality and descriptions of issues the user encountered or other details. For more information, see [Console docs](https://docs.syllable.ai/workspaces/Sessions). - name: tools description: >- Operations related to tool configuration. A tool is a function that an agent can call to perform actions like accessing databases, making API calls, or processing data. For an agent to have access to a tool, the prompt associated with that agent should be linked to the tool and include instructions to use it. For more information, see [Console docs](https://docs.syllable.ai/Resources/Tools). paths: /api/v1/experiments/{experiment_id}/start: post: tags: - experiments summary: Start Experiment description: >- Start an experiment. (Only an experiment in 'draft' status can be started.) operationId: experiments_start parameters: - name: experiment_id in: path required: true schema: type: integer title: Experiment Id responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/ExperimentStartResponse' '400': description: Bad Request '404': description: Not Found '409': description: Conflict '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' security: - APIKeyHeader: [] x-codeSamples: - lang: typescript label: Typescript (SDK) source: |- import { SyllableSDK } from "syllable-sdk"; const syllableSDK = new SyllableSDK({ apiKeyHeader: process.env["SYLLABLESDK_API_KEY_HEADER"] ?? "", }); async function run() { const result = await syllableSDK.experiments.experimentsStart({ experimentId: 381743, }); console.log(result); } run(); - lang: python label: Python (SDK) source: |- import os from syllable_sdk import SyllableSDK with SyllableSDK( api_key_header=os.getenv("SYLLABLESDK_API_KEY_HEADER", ""), ) as ss_client: res = ss_client.experiments.experiments_start(experiment_id=381743) # Handle response print(res) components: schemas: ExperimentStartResponse: properties: experiment: $ref: '#/components/schemas/ExperimentResponse' description: The experiment, now running warnings: items: type: string type: array title: Warnings description: >- Things that weaken the result without invalidating it, such as a variant whose agent does not pin a prompt version, so that the prompt can be edited while the experiment runs type: object required: - experiment title: ExperimentStartResponse description: A started experiment. HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError ExperimentResponse: properties: id: type: integer title: Id description: The internal ID of the experiment examples: - 1 name: type: string title: Name description: The name of the experiment examples: - Shorter greeting target_id: anyOf: - type: integer - type: 'null' title: Target Id description: >- The channel target being split, or null if that target has since been deleted. A running experiment always has one. examples: - 1 variants: items: $ref: '#/components/schemas/DaoExperimentVariantDetail' type: array title: Variants description: The variants of the experiment, in a stable order description: anyOf: - type: string - type: 'null' title: Description description: The hypothesis under test status: $ref: '#/components/schemas/ExperimentStatus' description: draft, running, or stopped started_at: anyOf: - type: string format: date-time - type: 'null' title: Started At description: When the experiment started stopped_at: anyOf: - type: string format: date-time - type: 'null' title: Stopped At description: When the experiment stopped created_at: type: string format: date-time title: Created At description: When the experiment was created updated_at: type: string format: date-time title: Updated At description: When the experiment was last written last_updated_by: type: string title: Last Updated By description: Who last wrote it target: anyOf: - $ref: '#/components/schemas/DaoExperimentTargetSummary' - type: 'null' description: >- The channel target being split, resolved for display. Null when the target has since been deleted, and when the read did not load it. type: object required: - id - name - target_id - variants - status - created_at - updated_at - last_updated_by title: ExperimentResponse description: >- Response model for experiment operations. An experiment splits the traffic of one channel target between two or more agents, so that the agents can be compared on the same population. One variant is the control the others are measured against. Each variant is a weighted pointer to the agent that variant runs; the weights are relative and not percentages. ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type type: object required: - loc - msg - type title: ValidationError DaoExperimentVariantDetail: properties: id: type: integer title: Id description: The internal ID of the variant examples: - 1 name: type: string title: Name description: The name of the variant, which is the grouping key in reports examples: - b weight: type: integer title: Weight description: >- The relative weight of the variant, not a percentage. (E.g., if two variants both have a weight of 10, they are equally likely to be selected.) examples: - 10 is_control: type: boolean title: Is Control description: >- Whether this is the reference variant the others are measured against agent_id: type: integer title: Agent Id description: The agent this variant runs examples: - 1 agent_name: type: string title: Agent Name description: The name of that agent examples: - Front desk type: object required: - id - name - weight - is_control - agent_id - agent_name title: DaoExperimentVariantDetail description: A variant, with the name of the agent it runs resolved for display. ExperimentStatus: type: string enum: - draft - running - stopped title: ExperimentStatus description: >- Status of an experiment. The column is a varchar and not a database enum, so a new value here does not need a migration. DaoExperimentTargetSummary: properties: id: type: integer title: Id description: The internal ID of the channel target examples: - 1 target: type: string title: Target description: The target itself, such as a phone number examples: - '+19995551234' target_mode: $ref: '#/components/schemas/TargetModes' description: The mode of the channel target channel_id: type: integer title: Channel Id description: The channel the target belongs to examples: - 1 channel_name: type: string title: Channel Name description: The name of that channel examples: - Main line type: object required: - id - target - target_mode - channel_id - channel_name title: DaoExperimentTargetSummary description: >- The channel target an experiment splits, resolved for display. Enough to name the target on the experiment card without a second read: the Console shows the target itself and the channel it belongs to, and links to neither by ID. TargetModes: type: string enum: - voice - chat - sms - email - whatsapp title: TargetModes description: Available modes (communication methods) for channel targets. securitySchemes: APIKeyHeader: type: apiKey in: header name: Syllable-API-Key ````