> ## Documentation Index > Fetch the complete documentation index at: https://docs.larynth.com/llms.txt > Use this file to discover all available pages before exploring further. # Create a draft version ## OpenAPI ````yaml /api-reference/openapi.json post /v1/assistants/{id}/versions openapi: 3.1.0 info: description: >- Larynth control-plane API v1. Errors are RFC 9457 problem+json with a stable type URI per code (x-larynth-error-codes). Operations carry x-larynth-min-role, x-larynth-auth, x-larynth-scope, x-larynth-idempotency and x-larynth-sdk. title: Larynth API version: 1.0.0 servers: - url: https://api-us.larynth.com description: US region security: [] tags: - description: Dashboard login and session. Excluded from SDKs. name: auth x-larynth-sdk: false - description: Tenant settings and members. Dashboard session only. name: tenants x-larynth-sdk: true - description: API keys. Dashboard session only. name: api-keys x-larynth-sdk: true - description: Audit log. Dashboard session only. name: audit x-larynth-sdk: true - description: Models and voices available to assistants. name: catalog x-larynth-sdk: true - description: Assistants and their immutable published versions. name: assistants x-larynth-sdk: true - description: Phone numbers and inbound routing. name: numbers x-larynth-sdk: true - description: Contacts and CSV imports. name: contacts x-larynth-sdk: true - description: Consent, revocations, do-not-call and pre-dial decisions. name: compliance x-larynth-sdk: true - description: Outbound campaigns. name: campaigns x-larynth-sdk: true - description: Calls, turns and events. name: calls x-larynth-sdk: true - description: Encrypted call recordings. name: recordings x-larynth-sdk: true - description: Prepaid balance, ledger, top-ups and rates (micro-USD). name: billing x-larynth-sdk: true - description: Webhook endpoints and deliveries (Standard Webhooks signatures). name: webhooks x-larynth-sdk: true - description: >- Knowledge bases built from files and websites, attached to assistant versions. name: knowledge x-larynth-sdk: true - description: Live updates over WebSocket. Dashboard session only. name: realtime x-larynth-sdk: true - description: Platform operator endpoints. Excluded from SDKs. name: admin x-larynth-sdk: false paths: /v1/assistants/{id}/versions: post: tags: - assistants summary: Create a draft version operationId: createAssistantVersion parameters: - in: path name: id required: true schema: format: uuid type: string requestBody: content: application/json: schema: $ref: '#/components/schemas/AssistantVersionSpec' required: true responses: '201': content: application/json: schema: $ref: '#/components/schemas/AssistantVersion' description: Created '400': $ref: '#/components/responses/Problem400' '401': $ref: '#/components/responses/Problem401' '403': $ref: '#/components/responses/Problem403' '404': $ref: '#/components/responses/Problem404' '409': $ref: '#/components/responses/Problem409' '422': $ref: '#/components/responses/Problem422' '429': $ref: '#/components/responses/Problem429' '500': $ref: '#/components/responses/Problem500' security: - session: [] - apiKey: - assistants:write components: schemas: AssistantVersionSpec: additionalProperties: false properties: denoise: type: boolean fillers: items: type: string maxItems: 20 type: array greeting_text: maxLength: 1000 minLength: 1 type: string interruption_profile: $ref: '#/components/schemas/InterruptionProfile' knowledge_base_ids: description: >- Knowledge bases the assistant can search with the search_knowledge tool; each must be ready when the version is published items: type: string maxItems: 5 type: array uniqueItems: true language: pattern: ^[a-z]{2,3}(-[A-Z]{2})?$ type: string latency: $ref: '#/components/schemas/LatencySettings' llm_fallback_model_ids: items: type: string maxItems: 3 type: array llm_model_id: description: Required for cascade format: uuid type: string llm_params: additionalProperties: {} type: object max_call_seconds: default: 1800 format: int64 maximum: 14400 minimum: 60 type: integer mode: enum: - cascade - s2s type: string post_call: $ref: '#/components/schemas/PostCallSettings' description: Post-call summary, data extraction and success evaluation s2s_model_id: description: Required for s2s format: uuid type: string s2s_params: additionalProperties: {} type: object stt_model_id: description: Required for cascade format: uuid type: string stt_params: additionalProperties: {} type: object system_prompt: maxLength: 32000 type: string tools: items: $ref: '#/components/schemas/AssistantTool' maxItems: 32 type: array tts_model_id: description: Required for cascade format: uuid type: string tts_params: additionalProperties: {} type: object tts_voice_id: description: Required for cascade format: uuid type: string turn_profile: $ref: '#/components/schemas/TurnProfile' required: - mode - language - system_prompt - greeting_text type: object AssistantVersion: additionalProperties: false properties: assistant_id: format: uuid type: string created_at: format: date-time type: string created_by: type: string denoise: type: boolean fillers: items: type: string maxItems: 20 type: array greeting_text: maxLength: 1000 minLength: 1 type: string id: format: uuid type: string interruption_profile: $ref: '#/components/schemas/InterruptionProfile' knowledge_base_ids: description: >- Knowledge bases the assistant can search with the search_knowledge tool; each must be ready when the version is published items: type: string maxItems: 5 type: array uniqueItems: true language: pattern: ^[a-z]{2,3}(-[A-Z]{2})?$ type: string latency: $ref: '#/components/schemas/LatencySettings' llm_fallback_model_ids: items: type: string maxItems: 3 type: array llm_model_id: description: Required for cascade format: uuid type: string llm_params: additionalProperties: {} type: object max_call_seconds: default: 1800 format: int64 maximum: 14400 minimum: 60 type: integer mode: enum: - cascade - s2s type: string post_call: $ref: '#/components/schemas/PostCallSettings' description: Post-call summary, data extraction and success evaluation published_at: format: date-time type: string s2s_model_id: description: Required for s2s format: uuid type: string s2s_params: additionalProperties: {} type: object stt_model_id: description: Required for cascade format: uuid type: string stt_params: additionalProperties: {} type: object system_prompt: maxLength: 32000 type: string tools: items: $ref: '#/components/schemas/AssistantTool' maxItems: 32 type: array tts_model_id: description: Required for cascade format: uuid type: string tts_params: additionalProperties: {} type: object tts_voice_id: description: Required for cascade format: uuid type: string turn_profile: $ref: '#/components/schemas/TurnProfile' version: format: int64 type: integer required: - id - assistant_id - version - created_by - created_at - mode - language - system_prompt - greeting_text type: object InterruptionProfile: additionalProperties: false properties: enabled: type: boolean false_interruption_ms: format: int64 maximum: 10000 minimum: 0 type: integer ignore_words: items: type: string maxItems: 100 type: array keep_echo_words: description: >- Count caller words that repeat the agent's own reply (line echo) as barge-in words type: boolean min_speech_ms: format: int64 maximum: 10000 minimum: 0 type: integer min_words: format: int64 maximum: 20 minimum: 0 type: integer vad_only_min_speech_ms: description: >- Barge-in on voice activity alone, before the STT has heard a word, needs this much speech. Unset is 1000 format: int64 maximum: 10000 minimum: 0 type: integer type: object LatencySettings: additionalProperties: false properties: force_finalize: description: >- Ask the STT provider to end its turn when the turn is committed on a non-STT signal. Unset uses the engine node default type: boolean short_first_chunk: description: >- Cut the first TTS chunk early and send it as its own utterance for a faster first audio. Unset uses the engine node default type: boolean speculative_tts: description: >- Synthesise the first chunk of a speculative reply privately and play it only if the turn is committed. Unset uses the engine node default type: boolean speculative_tts_min_prob: description: >- Minimum eager end-of-turn probability for speculative TTS. Unset uses the engine node default format: double maximum: 1 minimum: 0 type: number type: object PostCallSettings: additionalProperties: false properties: extraction_fields: items: $ref: '#/components/schemas/ExtractionField' maxItems: 30 type: array success_criteria: description: When set, the analysis also returns successful and reason maxLength: 2000 type: string summary_enabled: description: Unset is true type: boolean summary_prompt: description: Overrides the default summary instruction maxLength: 2000 type: string type: object AssistantTool: additionalProperties: false properties: class: enum: - fast - slow - terminal type: string config: additionalProperties: {} type: object description: maxLength: 2000 minLength: 1 type: string destinations: description: >- Transfer tools only (kind transfer, name transfer_call): where the agent may transfer the caller items: $ref: '#/components/schemas/TransferDestination' maxItems: 20 type: array endpoint_ref: maxLength: 200 type: string json_schema: additionalProperties: {} type: object kind: enum: - end_call - opt_out - http_request - transfer type: string name: pattern: ^[a-z][a-z0-9_]{0,63}$ type: string timeout_ms: default: 8000 format: int64 maximum: 30000 minimum: 100 type: integer required: - name - kind - description - json_schema - class type: object TurnProfile: additionalProperties: false properties: commit_prob: format: double maximum: 1 minimum: 0 type: number commit_without_words: description: >- With STT end-of-turn primary, commit turns the STT heard no words in instead of discarding them as noise or echo type: boolean eager_prob: format: double maximum: 1 minimum: 0 type: number hold_max_ms: description: >- Silence a model end-of-turn commit needs at commit_prob; confident predictions commit sooner. Unset is 500 format: int64 maximum: 5000 minimum: 0 type: integer hold_min_ms: description: >- Silence a model end-of-turn commit needs at probability 1. Unset is 200 format: int64 maximum: 5000 minimum: 0 type: integer max_silence_ms: format: int64 maximum: 30000 minimum: 0 type: integer min_speech_ms: format: int64 maximum: 10000 minimum: 0 type: integer stop_silence_ms: format: int64 maximum: 10000 minimum: 0 type: integer use_stt_eot: type: boolean type: object Problem: additionalProperties: true properties: code: description: >- Stable machine-readable error code; compliance denials carry the dial.* rule id type: string detail: type: string errors: items: $ref: '#/components/schemas/ErrorDetail' type: array instance: type: string retry_at: description: Next moment the request may succeed (calling windows) format: date-time type: string status: format: int64 type: integer title: type: string type: description: Stable URI identifying the error code type: string required: - type - title - status - code type: object ExtractionField: additionalProperties: false properties: description: maxLength: 500 type: string enum: description: Allowed values; required for type enum, rejected otherwise items: type: string maxItems: 20 type: array name: pattern: ^[a-z][a-z0-9_]{0,39}$ type: string required: description: The model must return a non-null value type: boolean type: enum: - string - number - boolean - enum - date type: string required: - name - type type: object TransferDestination: additionalProperties: false properties: description: description: When to use this destination; shown to the agent maxLength: 500 type: string e164: description: Phone number; set exactly one of e164 and sip_uri pattern: ^\+[1-9][0-9]{1,14}$ type: string mode: description: >- cold hands the caller over directly; warm puts the caller on hold, briefs the human, then bridges enum: - cold - warm type: string name: description: The agent picks a destination by this name maxLength: 64 minLength: 1 type: string ring_timeout_seconds: description: >- How long to ring the human before returning to the caller. Unset is 25 format: int64 maximum: 60 minimum: 0 type: integer sip_uri: description: 'sip: or sips: URI; cold transfers only, by SIP REFER' maxLength: 256 type: string whisper: description: >- Warm only: what the agent says to the human before bridging. Variables {{summary}} and {{destination}} maxLength: 1000 type: string required: - name - mode type: object ErrorDetail: additionalProperties: false properties: location: description: >- Where the error occurred, e.g. 'body.items[3].tags' or 'path.thing-id' type: string message: description: Error message text type: string value: description: The value at the given location type: object responses: Problem400: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Bad Request Problem401: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Unauthorized Problem403: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Forbidden Problem404: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Not Found Problem409: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Conflict Problem422: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Unprocessable Entity Problem429: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Too Many Requests Problem500: content: application/problem+json: schema: $ref: '#/components/schemas/Problem' description: Internal Server Error securitySchemes: session: description: Dashboard session cookie in: cookie name: lx_session type: apiKey apiKey: bearerFormat: lx__ description: API key; the listed scope is required scheme: bearer type: http ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.