> ## 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. # Evaluate decision models (TypeSafe-compatible) > **Beta API.** Choose telnyx/decision-flash for the lowest cost and latency, or telnyx/decision-pro for decisions that require long context, including inputs beyond Jev’s 32k per-decision limit. Omitted model defaults to telnyx/decision-flash. Evaluate shared context using named choice, noul (yes/no), and score questions. Returns TypeSafe System One-compatible answer shapes, the selected public model alias, and token usage. See the [decision model guide](https://developers.telnyx.com/docs/inference/decision-models) for examples and compatibility limits. The supported request subset requires instructions for every question, string descriptions for criteria (or null for choice descriptions), 1–64 questions, and 2–64 options for choice and score questions. The model field accepts only telnyx/decision-flash or telnyx/decision-pro. Unsupported model values and unknown fields are rejected. The endpoint is synchronous and does not stream. Use the TypeSafe Python SDK with base_url set to https://api.telnyx.com/v2/ai/typesafe and a Telnyx API key. The SDK appends /v1/systemone; explicitly set model to a supported Telnyx alias because its own default model is not supported. Compatibility covers this operation and the documented request subset; it does not include TypeSafe model listing. Scores describe relative preference, not calibrated correctness. ## OpenAPI ````yaml /openapi/source/external/inference/inference-decision-models.json post /ai/typesafe/v1/systemone openapi: 3.1.0 info: version: 2.0.0 title: Telnyx API x-latency-category: responsive x-endpoint-cost: light description: SIP trunking, SMS, MMS, Call Control and Telephony Data Services. contact: email: support@telnyx.com servers: - url: https://api.telnyx.com/v2 description: Version 2.0.0 of the Telnyx API security: - bearerAuth: [] tags: - name: Decision Models description: >- Beta API for evaluating shared context with typed questions and structured answers using Flash or Pro. paths: /ai/typesafe/v1/systemone: post: tags: - Decision Models summary: Evaluate decision models (TypeSafe-compatible) description: >- **Beta API.** Choose telnyx/decision-flash for the lowest cost and latency, or telnyx/decision-pro for decisions that require long context, including inputs beyond Jev’s 32k per-decision limit. Omitted model defaults to telnyx/decision-flash. Evaluate shared context using named choice, noul (yes/no), and score questions. Returns TypeSafe System One-compatible answer shapes, the selected public model alias, and token usage. See the [decision model guide](https://developers.telnyx.com/docs/inference/decision-models) for examples and compatibility limits. The supported request subset requires instructions for every question, string descriptions for criteria (or null for choice descriptions), 1–64 questions, and 2–64 options for choice and score questions. The model field accepts only telnyx/decision-flash or telnyx/decision-pro. Unsupported model values and unknown fields are rejected. The endpoint is synchronous and does not stream. Use the TypeSafe Python SDK with base_url set to https://api.telnyx.com/v2/ai/typesafe and a Telnyx API key. The SDK appends /v1/systemone; explicitly set model to a supported Telnyx alias because its own default model is not supported. Compatibility covers this operation and the documented request subset; it does not include TypeSafe model listing. Scores describe relative preference, not calibrated correctness. operationId: create_typesafe_systemone requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/DecisionModelRequest' example: model: telnyx/decision-flash state: Our production calls are failing. Every customer is affected. questions: team: type: choice instructions: Choose the team that should handle this incident. criteria: billing: Payments and refunds technical_support: Service faults and technical problems sales: New purchases production_incident: type: noul instructions: Does the message describe an active production incident? urgency: type: score instructions: Rate operational urgency. criteria: - Low - Normal - High - Critical responses: '200': description: >- All questions evaluated successfully. The response is a complete JSON object, not a stream. content: application/json: schema: $ref: '#/components/schemas/DecisionModelResponse' example: model: telnyx/decision-flash answers: team: type: choice choice: technical_support probabilities: billing: 0.0024719786572585446 technical_support: 0.9972674768333701 sales: 0.00026054450937119613 confidence: 0.9820522416354122 production_incident: type: noul noul: 0.9991959141264766 urgency: type: score score: 2.9974242859177265 legend: '0': Low '1': Normal '2': High '3': Critical probabilities: '0': 0.0003348349100603869 '1': 0.00003529133987660637 '2': 0.0015006266723396524 '3': 0.9981292470777234 confidence: 0.9894200002646713 usage: input_tokens: 267 output_tokens: 4 '401': description: Unauthorized. Supply a valid Telnyx API key. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Authentication required. '413': description: Request body is too large or incomplete. Reduce or fix the request. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Request body too large. '422': description: >- Invalid schema or input token limit exceeded. Correct or split the request before retrying. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Question "team" requires instructions. '429': description: Too many requests. Reduce concurrency and retry with backoff. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Too many requests. headers: Retry-After: description: When present, wait this many seconds before retrying. schema: type: integer minimum: 0 '502': description: The scoring service failed. Retry with bounded backoff. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Scoring service failed. '503': description: >- The decision model service is unavailable. Retry with bounded backoff. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Decision model service unavailable. headers: Retry-After: description: When present, wait this many seconds before retrying. schema: type: integer minimum: 0 '504': description: >- Evaluation deadline exceeded. Reduce request size or concurrency before retrying. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Evaluation deadline exceeded. '529': description: >- The decision model service is at capacity. Retry with bounded backoff. content: application/json: schema: $ref: '#/components/schemas/DecisionModelError' example: error: message: Decision model service at capacity. headers: Retry-After: description: When present, wait this many seconds before retrying. schema: type: integer minimum: 0 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.ai.typesafe.v1.systemone({ questions: {}, state: 'Our production calls are failing. Every customer is affected.', }); console.log(response.model); - 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.ai.typesafe.v1.systemone( questions={}, state="Our production calls are failing. Every customer is affected.", ) print(response.model) - 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\tresponse, err := client.AI.Typesafe.V1.Systemone(\n\t\tcontext.TODO(),\n\t\ttelnyx.AITypesafeV1SystemoneParams{\n\t\t\tQuestions: \"questions\",\n\t\t\tState: \"Our production calls are failing. Every customer is affected.\",\n\t\t},\n\t)\n\tif err != nil {\n\t\tpanic(err.Error())\n\t}\n\tfmt.Printf(\"%+v\\n\", response.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.ai.typesafe.v1.V1SystemoneParams; public final class Main { private Main() {} public static void main(String[] args) { TelnyxClient client = TelnyxOkHttpClient.fromEnv(); V1SystemoneParams params = V1SystemoneParams.builder() .questions(null) .state("Our production calls are failing. Every customer is affected.") .build(); var response = client.ai().typesafe().v1().systemone(params); } } - lang: Ruby source: >- require "telnyx" telnyx = Telnyx::Client.new(api_key: "My API Key") response = telnyx.ai.typesafe.v1.systemone(questions: {}, state: "Our production calls are failing. Every customer is affected.") puts(response) - lang: PHP source: >- ai->typesafe->v1->systemone( questions: [], state: 'Our production calls are failing. Every customer is affected.', ); var_dump($response); } catch (APIException $e) { echo $e->getMessage(); } - lang: CLI source: |- telnyx ai:typesafe:v1 systemone \ --api-key 'My API Key' \ --state 'Our production calls are failing. Every customer is affected.' \ --questions '{}' components: schemas: DecisionModelRequest: type: object description: >- Decision Models beta request. Choose a public model alias; omitted model defaults to telnyx/decision-flash. Telnyx manages the underlying models behind these aliases. properties: model: type: string enum: - telnyx/decision-flash - telnyx/decision-pro default: telnyx/decision-flash description: >- Public model alias. telnyx/decision-flash offers the lowest cost and latency; telnyx/decision-pro supports decisions that require long context, including inputs beyond Jev’s 32k per-decision limit. Applies to every question in the request. Other values are rejected. example: telnyx/decision-flash state: $ref: '#/components/schemas/DecisionModelContent' description: Shared context evaluated by every question. example: Our production calls are failing. Every customer is affected. questions: type: object description: >- Between 1 and 64 named questions. Each key identifies the corresponding answer. minProperties: 1 maxProperties: 64 additionalProperties: $ref: '#/components/schemas/DecisionModelQuestion' required: - state - questions additionalProperties: false example: model: telnyx/decision-flash state: Our production calls are failing. Every customer is affected. questions: team: type: choice instructions: Choose the team that should handle this incident. criteria: billing: Payments and refunds technical_support: Service faults and technical problems sales: New purchases production_incident: type: noul instructions: Does the message describe an active production incident? urgency: type: score instructions: Rate operational urgency. criteria: - Low - Normal - High - Critical DecisionModelResponse: type: object description: >- A complete synchronous evaluation. Answers are returned directly without a data wrapper. properties: model: type: string description: >- Public model alias used to evaluate the request. Returns telnyx/decision-flash when model was omitted. The underlying model is managed by Telnyx. readOnly: true enum: - telnyx/decision-flash - telnyx/decision-pro example: telnyx/decision-flash answers: type: object description: >- Answers keyed by exactly the question IDs in the request. Each answer type matches its question. minProperties: 1 maxProperties: 64 additionalProperties: $ref: '#/components/schemas/DecisionModelAnswer' usage: $ref: '#/components/schemas/DecisionModelUsage' required: - model - answers - usage additionalProperties: false example: model: telnyx/decision-flash answers: team: type: choice choice: technical_support probabilities: billing: 0.0024719786572585446 technical_support: 0.9972674768333701 sales: 0.00026054450937119613 confidence: 0.9820522416354122 production_incident: type: noul noul: 0.9991959141264766 urgency: type: score score: 2.9974242859177265 legend: '0': Low '1': Normal '2': High '3': Critical probabilities: '0': 0.0003348349100603869 '1': 0.00003529133987660637 '2': 0.0015006266723396524 '3': 0.9981292470777234 confidence: 0.9894200002646713 usage: input_tokens: 267 output_tokens: 4 DecisionModelError: type: object description: A decision model service error. properties: error: type: object description: Error details. properties: message: type: string description: Human-readable explanation of the failure. required: - message additionalProperties: false required: - error additionalProperties: false DecisionModelContent: description: >- Text, a JSON object, or an array containing text-based context. Text-only conversation histories are supported; image and audio inputs are not supported. oneOf: - type: string - type: object additionalProperties: true - type: array items: {} DecisionModelQuestion: description: A choice, yes/no, or ordered-score question. oneOf: - $ref: '#/components/schemas/DecisionModelChoiceQuestion' - $ref: '#/components/schemas/DecisionModelNoulQuestion' - $ref: '#/components/schemas/DecisionModelScoreQuestion' discriminator: propertyName: type mapping: choice: $ref: '#/components/schemas/DecisionModelChoiceQuestion' noul: $ref: '#/components/schemas/DecisionModelNoulQuestion' score: $ref: '#/components/schemas/DecisionModelScoreQuestion' DecisionModelAnswer: description: An answer whose type matches its question. oneOf: - $ref: '#/components/schemas/DecisionModelChoiceAnswer' - $ref: '#/components/schemas/DecisionModelNoulAnswer' - $ref: '#/components/schemas/DecisionModelScoreAnswer' discriminator: propertyName: type mapping: choice: $ref: '#/components/schemas/DecisionModelChoiceAnswer' noul: $ref: '#/components/schemas/DecisionModelNoulAnswer' score: $ref: '#/components/schemas/DecisionModelScoreAnswer' DecisionModelUsage: type: object description: Token usage for the completed evaluation. properties: input_tokens: type: integer minimum: 0 description: >- Input tokens processed, including shared-context preparation and question evaluation. This can exceed the token count of the unique input text. output_tokens: type: integer minimum: 0 description: >- Output tokens used for the evaluation, including shared-context preparation. required: - input_tokens - output_tokens additionalProperties: false DecisionModelChoiceQuestion: type: object description: Select one of the supplied options. properties: type: type: string description: Question type. enum: - choice instructions: $ref: '#/components/schemas/DecisionModelContent' description: >- Required instructions describing what to decide about the shared state. criteria: type: object description: >- Between 2 and 64 option keys mapped to description strings or null. A null description uses the option key as its text. minProperties: 2 maxProperties: 64 additionalProperties: type: - string - 'null' required: - type - instructions - criteria additionalProperties: false example: type: choice instructions: Choose the team that should handle this incident. criteria: billing: Payments and refunds technical_support: Service faults and technical problems sales: New purchases DecisionModelNoulQuestion: type: object description: >- Evaluate a yes/no question. Omit criteria to use Yes and No descriptions. properties: type: type: string description: Question type. enum: - noul instructions: $ref: '#/components/schemas/DecisionModelContent' description: >- Required instructions describing what to decide about the shared state. criteria: $ref: '#/components/schemas/DecisionModelNoulCriteria' required: - type - instructions additionalProperties: false example: type: noul instructions: Does the message describe an active production incident? DecisionModelScoreQuestion: type: object description: Rate the state against an ordered rubric. properties: type: type: string description: Question type. enum: - score instructions: $ref: '#/components/schemas/DecisionModelContent' description: >- Required instructions describing what to decide about the shared state. criteria: type: array description: >- Between 2 and 64 description strings in ascending score order. Indices start at zero. minItems: 2 maxItems: 64 items: type: string required: - type - instructions - criteria additionalProperties: false example: type: score instructions: Rate operational urgency. criteria: - Low - Normal - High - Critical DecisionModelChoiceAnswer: type: object description: A selected option and the distribution across all supplied option keys. properties: type: type: string description: Answer type. enum: - choice choice: type: string description: >- The option key with the highest relative score. Ties favor the first option in request order. probabilities: $ref: '#/components/schemas/DecisionModelProbabilities' confidence: $ref: '#/components/schemas/DecisionModelConfidence' required: - type - choice - probabilities - confidence additionalProperties: false example: type: choice choice: technical_support probabilities: billing: 0.0024719786572585446 technical_support: 0.9972674768333701 sales: 0.00026054450937119613 confidence: 0.9820522416354122 DecisionModelNoulAnswer: type: object description: A yes/no score with no separate confidence or probabilities fields. properties: type: type: string description: Answer type. enum: - noul noul: type: number description: >- Score of the positive outcome. Values near 1 favor yes; values near 0 favor no. This is a number, not a Boolean, and is not calibrated correctness. minimum: 0 maximum: 1 required: - type - noul additionalProperties: false example: type: noul noul: 0.9991959141264766 DecisionModelScoreAnswer: type: object description: An expected rating over the ordered criteria. properties: type: type: string description: Answer type. enum: - score score: type: number description: >- Expected zero-based criterion index: sum(index * probability). Ranges from 0 to N-1 for N criteria; fractional values are valid. minimum: 0 maximum: 63 legend: type: object description: >- Criterion descriptions keyed by stringified zero-based indices, such as "0", "1", and "2". minProperties: 2 maxProperties: 64 additionalProperties: type: string probabilities: $ref: '#/components/schemas/DecisionModelProbabilities' description: Relative scores keyed by the same stringified indices as legend. confidence: $ref: '#/components/schemas/DecisionModelConfidence' required: - type - score - legend - probabilities - confidence additionalProperties: false example: type: score score: 2.9974242859177265 legend: '0': Low '1': Normal '2': High '3': Critical probabilities: '0': 0.0003348349100603869 '1': 0.00003529133987660637 '2': 0.0015006266723396524 '3': 0.9981292470777234 confidence: 0.9894200002646713 DecisionModelNoulCriteria: type: object description: >- Optional descriptions for the positive and negative outcomes. Descriptions must be strings. properties: 'true': type: string description: Description of the positive outcome. default: 'Yes' 'false': type: string description: Description of the negative outcome. default: 'No' required: [] additionalProperties: false default: 'true': 'Yes' 'false': 'No' DecisionModelProbabilities: type: object description: >- Relative scores normalized across the supplied options, summing approximately to 1. These are not calibrated probabilities of correctness. minProperties: 2 maxProperties: 64 additionalProperties: type: number minimum: 0 maximum: 1 DecisionModelConfidence: type: number description: >- Normalized entropy confidence: 1 - H(p) / ln(N), where H(p) = -sum(p * ln(p)) and N is the number of options. Zero indicates a uniform distribution; one indicates concentration on one option. This is neither the winning probability nor calibrated correctness. minimum: 0 maximum: 1 securitySchemes: bearerAuth: type: http scheme: bearer ````