> ## Documentation Index > Fetch the complete documentation index at: https://guide.omnia-voice.com/llms.txt > Use this file to discover all available pages before exploring further. # Create a call with inline configuration > Run a call without a saved agent — the whole configuration travels in the request. Nothing is stored and nothing is reused. Use it for per-user personas, generated agents, or testing a prompt without leaving a record behind. If the same configuration is used repeatedly, create a real agent instead: you gain version history, tool assignment, a phone number, corpora, and the dashboard. ## OpenAPI ````yaml /openapi.yaml post /calls/inline openapi: 3.0.0 info: title: Omnia Voice API description: > The Omnia Voice API allows you to programmatically manage voice agents, tools, phone numbers, and call logs. ## Authentication All API requests require an API key to be passed in the `X-API-Key` header. ## Rate Limiting API requests are rate limited to 1000 requests per hour per API key. version: 1.0.0 contact: email: support@omnia-voice.com servers: - url: https://api.omnia-voice.com/api/v1 description: Production - url: https://dashboard.omnia-voice.com/api/v1 description: Production (alias — same service) security: - ApiKeyAuth: [] tags: - name: Agents description: Manage voice agents - name: Tools description: Manage agent tools and actions - name: Phone Numbers description: Manage phone number assignments - name: Calls description: Create and manage voice calls - name: Call Logs description: View call history and details - name: Pricing description: Get pricing information and calculate call costs - name: Configuration description: Get configuration options (voices, languages) paths: /calls/inline: post: tags: - Calls summary: Create a call with inline configuration description: | Run a call without a saved agent — the whole configuration travels in the request. Nothing is stored and nothing is reused. Use it for per-user personas, generated agents, or testing a prompt without leaving a record behind. If the same configuration is used repeatedly, create a real agent instead: you gain version history, tool assignment, a phone number, corpora, and the dashboard. operationId: createInlineCall requestBody: required: true content: application/json: schema: type: object required: - systemPrompt - voice - language properties: systemPrompt: type: string description: | The **entire** prompt. There is no basePrompt/context/ customInstructions layering here — that belongs to saved agents. voice: type: string description: Voice name or ID. See `GET /voices`. example: Mark language: type: string description: Language code. See `GET /languages`. example: en model: type: string enum: - gemma - llama - glm default: llama temperature: type: number minimum: 0 maximum: 1 default: 0 greeting: type: string description: | Appended to the prompt as the opening line when the agent speaks first. Ignored when `firstSpeaker` is `user`. firstSpeaker: type: string enum: - agent - user default: agent interruptible: type: boolean default: false description: Applies to the agent's first message only. maxDuration: type: integer minimum: 60 maximum: 7200 default: 3000 description: Hard cap in seconds. inactivityTimeout: type: integer minimum: 10 maximum: 300 description: Seconds of silence before the agent reacts. inactivityAction: type: string enum: - hang_up - prompt recordingEnabled: type: boolean default: true connectionType: $ref: '#/components/schemas/ConnectionType' websocket: $ref: '#/components/schemas/WebSocketOptions' selectedTools: type: array items: $ref: '#/components/schemas/InlineSelectedTool' metadata: type: object additionalProperties: true responses: '201': description: Call session created content: application/json: schema: type: object properties: id: type: string websocketUrl: type: string description: Short-lived, single-use URL to stream audio to. status: type: string connectionType: type: string createdAt: type: string format: date-time '400': description: | Missing `systemPrompt` / `voice` / `language`; an invalid `model`; an unsupported `connectionType`; a `websocket` connection without both sample rates; or a `selectedTools` entry setting more than one of `toolId` / `toolName` / `temporaryTool`. '402': description: Insufficient credits '429': description: Rate limit exceeded components: schemas: ConnectionType: type: string enum: - twilio - telnyx - plivo - websocket - webrtc default: twilio description: | Which transport the returned `websocketUrl` is prepared for. `twilio` / `telnyx` / `plivo` — carrier media streams, **G.711 audio**. `websocket` — your own client; you choose the sample rates and codec. `webrtc` — browser transport, normally reached through the SDK. Any other value returns `400`. WebSocketOptions: type: object description: | Audio parameters. Only read when `connectionType` is `websocket` — telephony transports force G.711 at 8000 Hz regardless of what is sent here. required: - inputSampleRate - outputSampleRate properties: inputSampleRate: type: integer example: 16000 description: | Sample rate of the audio you send, in Hz. **Required** when `connectionType` is `websocket` — omitting it returns `400`. outputSampleRate: type: integer example: 16000 description: | Sample rate you want back, in Hz. **Required** when `connectionType` is `websocket`. codec: type: string enum: - pcm - g711 default: g711 description: '`pcm` for direct integrations, `g711` for telephony.' enableAudioBuffering: type: boolean description: Buffer audio to smooth jitter on unreliable links. twilioStream: type: boolean description: Format the stream for Twilio's `` verb. InlineSelectedTool: type: object description: | A tool for one inline call. Set **exactly one** of `toolId`, `toolName`, or `temporaryTool` — setting more than one returns `400`. properties: toolId: type: string description: A tool already saved in your workspace. toolName: type: string description: A system tool, e.g. `hangUp`. example: hangUp temporaryTool: type: object description: | A full tool definition scoped to this call only. Same shape as a saved tool, but using the runtime's array form for parameters rather than the keyed-object form. properties: modelToolName: type: string pattern: ^[a-zA-Z][a-zA-Z0-9_]*$ description: type: string timeout: $ref: '#/components/schemas/ToolTimeout' precomputable: type: boolean description: Runtime form of `readOnly`. dynamicParameters: type: array items: type: object properties: name: type: string location: $ref: '#/components/schemas/ParameterLocation' required: type: boolean schema: type: object staticParameters: type: array items: type: object automaticParameters: type: array items: type: object http: type: object properties: baseUrlPattern: type: string format: uri httpMethod: type: string enum: - GET - POST - PUT - PATCH - DELETE client: type: object description: Empty object marks this as a client-executed tool. ToolTimeout: type: string pattern: ^\d+(\.\d+)?s$ description: | Seconds with an `s` suffix, e.g. `"2.5s"` or `"10s"`. Must be greater than 0 and at most **40s**. Defaults to 2.5s when unset. example: 5s ParameterLocation: type: string description: | Where the value goes in the HTTP request. Both the short and the canonical long form are accepted on input; the long form is what is sent to the runtime. enum: - header - query - path - body - PARAMETER_LOCATION_HEADER - PARAMETER_LOCATION_QUERY - PARAMETER_LOCATION_PATH - PARAMETER_LOCATION_BODY securitySchemes: ApiKeyAuth: type: apiKey in: header name: X-API-Key description: API key for authentication ````