openapi: 3.0.0 info: title: Universal Chat API version: 1.0.0 servers: - url: https://api.openai.com/v1 paths: /openai/deployments/{Deployment_Name}/chat/completions/{api_version}: post: security: - ApiKeyAuth: [] parameters: - in: path name: api_version schema: type: string required: true description: The API version to use for this operation. This follows the YYYY-MM-DD format. example: '2023-03-15' requestBody: required: true content: application/json: schema: type: object properties: model: type: string description: ID of the model to use. messages: type: array minItems: 1 description: A list of messages comprising the conversation so far. items: $ref: '#/components/schemas/Message' functions: type: array items: $ref: '#/components/schemas/Function' description: A list of functions the assistant may generate JSON inputs for. function_call: description: 'Controls how the assistant responds to function calls. "none" means the assistant does not call a function, and responds to the end-user. "auto" means the assistant can pick between an end-user or calling a function. Specifying a particular function via `{"name": "my_function"}` forces the assistant to call that function. "none" is the default when no functions are present. "auto" is the default if functions are present.' oneOf: - type: string enum: - none - auto - type: object properties: name: type: string description: The name of the function to call. required: [name] addons: type: array items: $ref: '#/components/schemas/Addon' description: A list of addons the assistant may use. stream: type: boolean description: 'If set, partial message deltas will be sent. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a `data: [DONE]` message' default: false temperature: type: number description: What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. minimum: 0 maximum: 2 default: 1 top_p: type: number description: An alternative to sampling with temperature, called nucleus sampling, where the assistant considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. minimum: 0 maximum: 1 default: 1 'n': type: integer description: How many chat completion choices to generate for each input message. minimum: 1 maximum: 128 default: 1 stop: oneOf: - type: string nullable: true - type: array minItems: 1 maxItems: 4 items: type: string description: Up to 4 sequences where the assistant will stop generating further tokens. max_tokens: type: integer description: The maximum number of tokens to generate by the assistant. default: infinity presence_penalty: type: number description: Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics. minimum: -2 maximum: 2 default: 0 frequency_penalty: type: number description: Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim. minimum: -2 maximum: 2 default: 0 logit_bias: type: object additionalProperties: true description: |- Modify the likelihood of specified tokens appearing in the completion. Accepts a json object that maps tokens (specified by their token ID in the tokenizer) to an associated bias value from -100 to 100. Mathematically, the bias is added to the logits generated by the model prior to sampling. The exact effect will vary per model, but values between -1 and 1 should decrease or increase likelihood of selection; values like -100 or 100 should result in a ban or exclusive selection of the relevant token. default: null user: type: string description: A unique identifier representing end-user. required: - messages examples: GPT World: value: messages: - role: user content: What is chat gpt? GPT World Streaming: value: stream: true messages: - role: user content: What is chat gpt? Weather-forecast Addon: value: model: gpt-4 messages: - role: user content: Temperature tomorrow in New York? addons: - name: weather-forecast Bing Search: value: messages: - role: user content: Find something in bing IMF Demo. The first user message: value: messages: - role: user content: Give me NGDP indicator for Poland IMF Demo. Control activation: value: messages: - role: user content: Give me NGDP indicator for Poland - role: assistant content: |- I see that you already have the NGDP indicator for Poland in your current data state. Here's the updated form based on your request: Form: Query Params { Period: [2000, 2022] Country: 964 Indicator: NGDP Frequency: A } Is there anything else you'd like to change or add? - role: user custom_content: completed_control: label: Get Data responses: '200': description: Success content: application/json: schema: $ref: '#/components/schemas/Response' examples: GPT World: value: id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion created: '1687222194' choices: - index: 0 message: role: assistant content: ChatGPT is an advanced conversational AI model built on the GPT-3.5 architecture. It is a Large Language Model (LLM) that can engage in context-aware, human-like interactions. It is a suitable testbed for creativity and has demonstrated remarkable capabilities in understanding and generating human-like text. It heavily relies on in-context examples and has a strong bias toward using or not using certain tools. custom_content: attachments: - index: 0 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 1 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 2 type: text/plain title: 'HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face' data: By leveraging the strong language capability of ChatGPT and abundant AI models in Hugging Face, HuggingGPT is able to cover numerous sophisticated AI tasks in different modalities and domains and achieve impressive results in language, vision, speech, and other challenging tasks, which paves a new way towards advanced artificial intelligence 2. reference_url: https://arxiv.org/abs/2303.17580 stages: - index: 0 name: Paraphrasing content: '**New phrase:** What is chat gpt?' status: completed - index: 1 name: Search for relevant documents attachments: - index: 0 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 1 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 2 type: text/plain title: 'HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face' data: By leveraging the strong language capability of ChatGPT and abundant AI models in Hugging Face, HuggingGPT is able to cover numerous sophisticated AI tasks in different modalities and domains and achieve impressive results in language, vision, speech, and other challenging tasks, which paves a new way towards advanced artificial intelligence 2. reference_url: https://arxiv.org/abs/2303.17580 status: completed finish_reason: stop statistics: usage_per_model: - name: text-embedding-ada-002 prompt_tokens: '23' total_tokens: '23' - name: gpt-4 prompt_tokens: '123' completion_tokens: '17' total_tokens: '140' Weather-forecast Addon: value: id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion created: '1687222194' choices: - index: 0 message: role: assistant content: |- The weather forecast for tomorrow, June 21, 2023, in New York is as follows: 01:00: Overcast clouds, temperature 20.59°C, feels like 20.3°C, humidity 61%, wind speed 4.97 m/s 22:00: Light rain, temperature 19.11°C, feels like 18.64°C, humidity 60%, wind speed 6.68 m/s custom_content: stages: - index: 0 name: Weather forecast content: |- ##### Addon request: `weather-forecast("New York", "2023-06-21")` ##### Addon response: Date & Time: 2023-06-21 01:00:00 Weather Description: overcast clouds Temperature: 293.74 K Feels Like: 293.45 K Humidity: 61 % Wind Speed: 4.97 m/s Date & Time: 2023-06-21 04:00:00 Weather Description: overcast clouds Temperature: 292.64 K Feels Like: 292.26 K Humidity: 62 % Wind Speed: 4.37 m/s status: completed finish_reason: stop usage: prompt_tokens: 0 completion_tokens: 0 total_tokens: 0 Bing Search: value: id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion created: '1687222194' choices: - index: 0 message: role: assistant content: I found some breaking news on CNN regarding a missing Titanic submersible. The submersible reportedly has about 40 hours of air left, and efforts to locate the vessel have not yet been successful. The sub, which disappeared on Sunday, has five people on board. The Coast Guard is uncertain if they have enough time to rescue those onboard before the oxygen runs out. The five members aboard the submersible include a Pakistani billionaire father and son, a British explorer, a French diver, and the CEO of the company that operates the submersible. custom_content: stages: - index: 0 name: Browsing content: I found some breaking news on CNN regarding a missing Titanic submersible. The submersible reportedly has about 40 hours of air left, and efforts to locate the vessel have not yet been successful. The sub, which disappeared on Sunday, has five people on board. The Coast Guard is uncertain if they have enough time to rescue those onboard before the oxygen runs out. The five members aboard the submersible include a Pakistani billionaire father and son, a British explorer, a French diver, and the CEO of the company that operates the submersible. attachments: - index: 0 title: Searched Bing data: 'Searched Bing: "[latest news](https://www.bing.com/search?q=latest%20news)"' reference_url: https://www.bing.com/search?q=latest%20news - index: 1 title: Clicked on data: 'Clicked on: "[edition.cnn.com](https://edition.cnn.com/)"' reference_url: https://edition.cnn.com/ - index: 2 title: Reading content data: Reading content - index: 3 title: Finished browsing data: Finished browsing status: completed finish_reason: stop usage: prompt_tokens: 0 completion_tokens: 0 total_tokens: 0 IMF Demo. The first user message: value: id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion created: '1687222194' choices: - index: 0 message: role: assistant content: |- I see that you already have the NGDP indicator for Poland in your current data state. Here's the updated form based on your request: Form: Query Params { Period: [2000, 2022] Country: 964 Indicator: NGDP Frequency: A } Is there anything else you'd like to change or add? custom_content: controls: - index: 0 label: Get Data finish_reason: stop usage: prompt_tokens: 0 completion_tokens: 0 total_tokens: 0 IMF Demo. Control activation: value: id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion created: '1687222194' choices: - index: 0 message: role: assistant content: '' custom_content: attachments: - index: 0 title: Gross domestic product, current prices type: image/png url: https://_.com/72153376-0fca-11ee-be56-0242ac120002.png - index: 1 title: Excel input data: |- | | DATA_CODE | SERIES_CODE | | - | --------- | ----------- | | 0 | WEO | 964.NGDP.A | - index: 2 title: SDMX URL data: SDMX reference_url: https://apim-imfeid-dev-01.azure-api.net/uatsdmx/api/v1/workspaces/default:StatGPT/registry/sdmx/2.1/data/WEO/964.NGDP.A?startPeriod=2000&endPeriod=2022 - index: 3 title: Code snippet data: |- ```python provider = sdmx.Request('IMF_RES') data_msg = provider.data('WEO', key={'FREQUENCY': ['A'], 'COUNTRY': ['964', '964'], 'INDICATOR': ['NGDP', 'NGDP']}, params={'startPeriod': '2000', 'endPeriod': '2022'})``` finish_reason: stop usage: prompt_tokens: 0 completion_tokens: 0 total_tokens: 0 text/event-stream: schema: type: array items: $ref: '#/components/schemas/StreamResponse' examples: GPT World Streaming: value: - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: role: assistant finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: custom_content: stages: - index: 0 name: Paraphrasing status: null finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: custom_content: stages: - index: 0 content: '**New phrase:**' status: null finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: custom_content: stages: - index: 0 content: ' What is chat gpt?' status: completed finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: custom_content: stages: - index: 1 name: Search for relevant documents status: null finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: custom_content: stages: - index: 1 attachments: - index: 0 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 1 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf status: null finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: custom_content: stages: - index: 1 attachments: - index: 2 type: text/plain title: 'HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face' data: By leveraging the strong language capability of ChatGPT and abundant AI models in Hugging Face, HuggingGPT is able to cover numerous sophisticated AI tasks in different modalities and domains and achieve impressive results in language, vision, speech, and other challenging tasks, which paves a new way towards advanced artificial intelligence 2. reference_url: https://arxiv.org/abs/2303.17580 status: completed finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222194' choices: - index: 0 delta: content: ChatGPT is an advanced conversational AI model built on the GPT-3.5 architecture. finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222195' choices: - index: 0 delta: content: ' It is a Large Language Model (LLM) that can engage in context-aware, human-like interactions. It is a suitable testbed for creativity and has demonstrated remarkable capabilities in understanding and generating human-like text. It heavily relies on in-context examples and has a strong bias toward using or not using certain tools.' finish_reason: null - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222195' choices: - index: 0 delta: custom_content: attachments: - index: 0 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 1 type: text/plain title: 'GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo' data: This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant in- teractions including word problems, story de- scriptions, multi-turn dialogue, and code. We openly release the collected data, data cura- tion procedure, training code, and final model weights to promote open research and repro- ducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU. reference_url: https://s3.amazonaws.com/static.nomic.ai/gpt4all/2023_GPT4All_Technical_Report.pdf - index: 2 type: text/plain title: 'HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face' data: By leveraging the strong language capability of ChatGPT and abundant AI models in Hugging Face, HuggingGPT is able to cover numerous sophisticated AI tasks in different modalities and domains and achieve impressive results in language, vision, speech, and other challenging tasks, which paves a new way towards advanced artificial intelligence 2. reference_url: https://arxiv.org/abs/2303.17580 finish_reason: stop - id: 1d84aa54-e476-405d-9713-386bdfc85993 object: chat.completion.chunk created: '1687222196' statistics: usage_per_model: - index: 0 name: text-embedding-ada-002 prompt_tokens: 23 total_tokens: 23 - index: 1 name: gpt-4 prompt_tokens: 123 completion_tokens: 17 total_tokens: 140 '401': description: Invalid Authentication content: application/json: schema: type: object properties: error: $ref: '#/components/schemas/Error' '429': description: Rate limit reached content: application/json: schema: type: object properties: error: $ref: '#/components/schemas/Error' '500': description: The server had an error while processing your request content: application/json: schema: type: object properties: error: $ref: '#/components/schemas/Error' '503': description: The engine is currently overloaded, please try again later content: application/json: schema: type: object properties: error: $ref: '#/components/schemas/Error' components: securitySchemes: ApiKeyAuth: type: apiKey in: header name: Api-Key description: |- All API requests must include the API key in the `Api-Key` HTTP header as follows: ```bash Api-Key: DIAL_API_KEY ``` schemas: FunctionCall: type: object description: |- The name and arguments of a function that should be called, as generated by the assistant. Only the `assistant` can set this field. properties: name: type: string description: The name of the function that the assistant decided to call. arguments: type: string description: The arguments for the function in JSON format. ResponseFunctionCall: type: object description: The name and arguments of a function that should be called, as generated by the assistant. properties: name: type: string description: The name of the function that the assistant decided to call. arguments: type: string description: The arguments for the function in JSON format. Function: type: object properties: name: type: string description: The name of the function to be called. Must be a-z, A-Z, 0-9, or contain underscores and dashes, with a maximum length of 64. description: type: string description: The description of what the function does. parameters: type: object description: The parameters the functions accepts, described as a JSON Schema object. required: - name Addon: oneOf: - type: object properties: name: type: string description: The name of the system addon. - type: object properties: url: type: string description: The url to open ai plugin schema. Message: type: object description: The assistant message. properties: role: type: string description: The role of the messages author. One of `system`, `user`, `assistant` or `function`. content: type: string description: The contents of the message. `content` is required for all messages except `assistant` messages with function calls or `user` messages with control. custom_content: $ref: '#/components/schemas/CustomContent' name: type: string writeOnly: true description: |- The name of the author of this message. `name` is required if `role` is _function_, and it should be the name of the function whose response is in the content. May contain a-z, A-Z, 0-9, and underscores, with a maximum length of 64 characters. function_call: type: object writeOnly: true $ref: '#/components/schemas/FunctionCall' required: - role ResponseMessage: type: object description: The assistant message. properties: role: type: string description: The role of the messages author. Always is `assistant` for response. content: type: string description: The contents of the message. `content` is set for all messages except messages with function calls. custom_content: $ref: '#/components/schemas/ResponseCustomContent' function_call: type: object writeOnly: true $ref: '#/components/schemas/ResponseFunctionCall' required: - role Delta: type: object description: Description of new content of the message. properties: role: type: string description: The role of the messages author. Always is `assistant` for response. content: type: string description: The contents of the message. custom_content: $ref: '#/components/schemas/ResponseCustomContent' function_call: $ref: '#/components/schemas/ResponseFunctionCall' CustomContent: type: object description: |- The custom content of the message. - For `assistant`, the field should save `state` passed by the `assistant` in the response. - For `user`, only `attachments` and `completed_control` can be set. - For `system`, the field is not available. properties: attachments: type: array items: $ref: '#/components/schemas/Attachment' description: List of attachments. The field can be used by the `assistant` to attach additional output and by the `user` to attach additional input. completed_control: $ref: '#/components/schemas/CompletedControl' state: type: object additionalProperties: true description: |- The internal state of the assistant. This field can have an arbitrary set of fields with an arbitrary structure. Only `assistant` message can contain `state` field. ResponseCustomContent: type: object description: The custom content of the message. properties: attachments: type: array items: $ref: '#/components/schemas/ResponseAttachment' description: List of attachments. controls: type: array items: $ref: '#/components/schemas/Control' description: List of controls. These are suggested controls to user. stages: type: array readOnly: true items: $ref: '#/components/schemas/Stage' description: Intermediate stages the assistant did to get the response. state: type: object additionalProperties: true description: The internal state of the assistant. This field can have an arbitrary set of fields with an arbitrary structure. In case of streaming, the state is published fully in one chunk. Attachment: type: object properties: type: type: string default: text/markdown description: The content type of the attachment. Should be one of the MIME types. title: type: string description: The title of the attachment. data: type: string description: |- Should follow the format described in the MIME standard for `type`. It is **required** that the attachment has exactly one of the following fields: `data` or `url` (never both). url: type: string description: |- The content of `url` should follow the format described in the MIME standard for `type`. It is **required** that the attachment has exactly one of the following fields: `data` or `url` (never both). reference_type: type: string description: The content type of `reference_url`. Should be one of the MIME types. reference_url: type: string description: If `reference_type` is specified, the content of `reference_url` should follow the format described in the MIME standart for `reference_type`. required: - index ResponseAttachment: type: object properties: index: type: integer description: All attachments of the message must have unique index. The first attachment must have index 0, each subsequent one should set an index that is one greater than the previous one. type: type: string default: text/markdown description: The content type of the attachment. Should be one of the MIME types. title: type: string description: The title of the attachment. data: type: string description: |- Should follow the format described in the MIME standard for `type`. It is **required** that the attachment has exactly one of the following fields: `data` or `url` (never both). url: type: string description: |- The content of `url` should follow the format described in the MIME standard for `type`. It is **required** that the attachment has exactly one of the following fields: `data` or `url` (never both). reference_type: type: string description: The content type of `reference_url`. Should be one of the MIME types. reference_url: type: string description: If `reference_type` is specified, the content of `reference_url` should follow the format described in the MIME standart for `reference_type`. required: - index Control: type: object properties: index: type: integer description: All controls of the message must have unique index. The first control must have index 0, each subsequent one should set an index that is one greater than the previous one. label: type: string description: The label of the control element. schema: type: object description: Schema of the control defined using JSON Schema. required: - index - label CompletedControl: type: object description: The completed control. The control should be selected from the set of controls from the last assistant response. properties: label: type: string description: The label of the control element. value: type: object description: Should follow the schema described in the related control. required: - label Stage: type: object properties: index: type: integer description: All stages of the message must have unique index. The first stage must have index 0, each subsequent one should set an index that is one greater than the previous one. name: type: string description: The name of the stage. content: type: string description: The contents of the stage. attachments: type: array items: $ref: '#/components/schemas/ResponseAttachment' description: List of attachments to the stage. status: type: string nullable: true description: |- The status of execution of the stage. The possible statuses: | Status | Description | | - | - | | `null` | The stage is in progress | | completed | The stage is completed | | failed | The stage is failed | required: - index - name - status Response: type: object properties: id: type: string description: The ID of the response. object: type: string description: Object type. Always is "chat.completion" for non-streaming. created: type: integer description: The response timestamp. The time in seconds since the epoch. choices: type: array description: List of generated messages. Contains _n_ items. items: type: object properties: index: type: integer description: The index of the choice from _0_ to _n - 1_ message: type: object $ref: '#/components/schemas/ResponseMessage' finish_reason: type: string nullable: true description: |- The finish reason of generation of the choice. The possible reasons: | Reason | Description| | - | - | | `null` | The generation is in progress | | stop | Successful generation | | length | The generation was stopped due to exceeding the available number of tokens | |function_call|The assistant decided to call a function| |content_filter|Omitted content due to a flag from content filters| required: - index - message - finish_reason usage: type: object description: The field contains information about used models tokens for generation of the response. properties: prompt_tokens: type: integer description: The number of tokens in the request to the model. completion_tokens: type: integer description: The number of tokens in the response from the model. total_tokens: type: integer description: The sum of prompt and completion tokens. statistics: type: object $ref: '#/components/schemas/Statistics' Statistics: type: object description: The assistant work statistics. properties: usage_per_model: type: array description: Statistics of tokens used in models by the assistant. In case of streaming, the statistics is published fully in one chunk. items: type: object properties: index: type: integer description: All items of the `usage_per_model` must have unique index. The first item must have index 0, each subsequent one should set an index that is one greater than the previous one. model: type: string description: The model name. prompt_tokens: type: integer description: The number of tokens in the request to the model. completion_tokens: type: integer description: The number of tokens in the response from the model. total_tokens: type: integer description: The sum of prompt and completion tokens. StreamResponse: type: object properties: id: type: string description: The ID of the response. object: type: string description: Object type. Always is "chat.completion.chunk" for streaming. created: type: integer description: The response timestamp. The time in seconds since the epoch. choices: type: array description: List of generated chanks. items: type: object properties: index: type: integer description: The index of the choice from 0 to _n - 1_ delta: type: object $ref: '#/components/schemas/Delta' finish_reason: type: string description: |- The finish reason of generation of the choice. The possible reasons: | Reason | Description| | - | - | | stop | Successful generation | | length | The generation was stopped due to exceeding the available number of tokens | |function_call|The assistant decided to call a function| |content_filter|Omitted content due to a flag from content filters| |`null`|The generation of the choice is in progress| required: - index - message - finish_reason usage: type: object description: The field contains information about used models tokens for generation of the response. In case of streaming, the usage is published fully in one chunk. properties: prompt_tokens: type: integer description: The number of tokens in the request to the model. completion_tokens: type: integer description: The number of tokens in the response from the model. total_tokens: type: integer description: The sum of prompt and completion tokens. statistics: type: object $ref: '#/components/schemas/Statistics' Error: type: object properties: type: type: string nullable: false message: type: string nullable: false param: type: string nullable: true code: type: string nullable: true required: - type - message - param - code