The 506 API is a robust platform designed to facilitate seamless interactions through a chat endpoint and several GET endpoints. This API is ideal for developers looking to integrate advanced conversational capabilities and data retrieval functions into their applications.
\n","schema":"https://schema.getpostman.com/json/collection/v2.0.0/collection.json","toc":[],"owner":"30708536","collectionId":"08a15832-ba16-450f-a9ca-823b1c39f271","publishedId":"2sA3kbfyBa","public":true,"customColor":{"top-bar":"FFFFFF","right-sidebar":"303030","highlight":"004673"},"publishDate":"2024-09-16T14:10:20.000Z"},"item":[{"name":"API Endpoints","item":[{"name":"LLM Endpoints","item":[{"name":"models","id":"8a38b7cd-7986-41a6-a946-53c7b29bb176","protocolProfileBehavior":{"disableBodyPruning":true},"request":{"auth":{"type":"noauth","isInherited":false},"method":"GET","header":[{"key":"api-organization-id","value":"This endpoint returns the available LLM models of the 506.ai Platform.
\nThe response JSON contains the id and name of all models which are available. The id should be used in the model -> id field of the chat and systemMedia endpoints.
\nResponse Codes:
\n204 - No Content = the returned list of models is empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint returns the roleId, title, createdAt and updatedAt of all roles that are shared with the API user. The roleId can be used in the different chat calls.
\nResponse Codes:
\n204 - No Content = the returned list of roles is empty
\n401 - Unauthorized = wrong organization id or api key
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint provides detailed information about all assistants shared with the API user. The response includes fields such as assistantId, name, description, model details, as well as information about selected files and data collections associated with each assistant. The assistantId is a key identifier that can be used in subsequent chat-related API calls. Additionally, the image field returns the assistant image as a base64 encoded string if one was selected during configuration.
Response Codes:
\n204 - No Content = the returned list of roles is empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = API user permission error
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe mediaContext endpoint provides flexible content retrieval functionality that adapts to different use cases through its mode-based operation. This POST endpoint serves as an optional but powerful tool for maintaining contextual information across your application.
\n❗️The selectedAssistantId field is optional. When this endpoint is called with an assistantId, the selectedMode field will be disregarded, as the assistant's configuration will be applied automatically on the backend.
The request body requires a mode parameter that defines the operation type, along with conditional parameters based on your selected mode.
\nWhen operating in BASIC mode, the endpoint requires you to specify your desired files and/or dataCollections in the request body. This mode returns the complete, unprocessed content of all specified files, making it ideal for scenarios where you need access to entire documents.
\nThe QA mode also requires files and/or dataCollections parameters in your request. However, unlike BASIC mode, this approach performs intelligent content processing through vector search technology. The endpoint returns selected content chunks that are most relevant to your query context, making it perfect for question-answering scenarios where precision matters more than completeness.
\nSEARCH mode operates independently and requires neither files nor dataCollections parameters - these should remain empty in your request. This mode performs comprehensive search operations and returns discovered content that matches your search criteria.
\nThe real power of this endpoint becomes apparent when integrating with chat functionality. The returned content can be seamlessly incorporated into your chat endpoint's message array structure. By assigning the role=\"system\" to the mediaContext content within your messages array, you maintain consistent context throughout conversations.
\nYour message flow might follow this pattern: a user message initiates the conversation, followed by system content from mediaContext to provide context, then an assistant response, creating a natural conversational flow that maintains awareness of your media content.
\nThis optional endpoint enhances your application's ability to maintain contextual awareness, whether you're building comprehensive document review systems, intelligent Q&A platforms, or sophisticated search interfaces.
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint returns the answer from the LLM as a text-event stream without references.
\n❗️The API includes a custom header api-ai-disclaimer with the value \"Treat with caution, content is AI-generated\" to clearly indicate that responses are AI-generated content and should be handled appropriately.
❗️There are two examples provided, each demonstrating different configurations. For examples related to assistants, please refer to the chat (assistant) example. You can find all examples conveniently located under the Example section on the right side within the drop-down menu.
\nShort description of the body JSON fields:
\nmodel -> id = the id returned by the models endpoint
\nmessage = the chat messages, use role = user for the user sent content and role = assistant for the answer from the LLM. The fields references and sources, in this version of the API, should always be an empty array. User sent message are role = user, responses from the LLM are role = assistant and responses from the systemMedia endpoint are role = system.
\nroleId = should be an empty string, or the id of a shared role with the API user.
\ntemperature = defines the creativity of the used LLM model. The value is between 0.0 and 1.0. With a low value the model is more constant and with a higher value it is more random
\nselectedMode = possible values are BASIC, QA and SEARCH.
\nBASIC mode performs a standard call to the LLM. If files are selected it will use the whole content of the file as input.
\nQA mode uses the vector search on selected files and/or data collections before sending the message to the LLM.
\nSEARCH mode leverages real-time web data to enhance responses before processing through the language model.
\nselectedFiles (optional) = The uniqueTitles of the selectedFiles which should be used for the call.
\nselectedDataCollections (optional) = The ids of the selected data collections which should be used for the call. Data collections are only possible in QA mode!
\nselectedAssistantId (optional) = Use this field to specify a pre-configured assistant for the chat endpoint. When an assistantId is provided, only the messages and selectedAssistantId fields are required. You may optionally include additional files and data collections via the selectedFiles and selectedDataCollections fields. All other fields will be ignored, as the assistant’s configuration will be loaded automatically from the backend.
Important: If you include files or data collections in selectedFiles or selectedDataCollections, ensure that the API user has the necessary access permissions to these media elements.
Response Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files, data collections or roles
\n413 - Payload to large = message greater than context window of model
\n429 - Too many requests = to many requests to azure open ai in a short period, request limit
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nOptional - Turns the 506 internal system prompts on or off. The default value is true.
\n","type":"text/plain"},"key":"internalSystemPrompt","value":"true"}],"variable":[]}},"response":[{"id":"d5c9580c-7a1e-4cc0-bebc-a1b3ccccf13b","name":"chat","originalRequest":{"method":"POST","header":[{"key":"Content-Type","value":"application/json","type":"text"},{"key":"api-organization-id","value":"This endpoint provides responses from the language model as a text-event stream. At the end of the stream, you'll receive references formatted as a valid JSON object, which begins with the label \"References:\". Each reference includes the unique title, title, and associated chunks from the file. If the answer is sourced from multiple files, the references will detail each file's unique title, title, and related chunks accordingly. You can use the uniqueTitle and chunk number (from chunks array) to make a GET chunk call, allowing you to retrieve the specific content of that chunk.
\nTo obtain a mapping of the inline references to their corresponding chunks in the file, you can utilize the citation parameter by setting its value to true. This will result in a response that includes a cites array for each file, with the same number of entries as the chunk array. The first entry in the cites array corresponds to the inline reference number of the content, while the first entry in the chunk array represents the related chunk, and this pattern continues for the subsequent entries. The references and cites are only returned in QA mode!
Please note that the citation parameter works only when the internalSystemPrompt parameter is true. Additionally, there is a second example showcasing the citation parameter set to true. You can find this option in the Example request section by selecting it from the dropdown menu on the right side, labeled as chatReferences (citation).
❗️The API includes a custom header api-ai-disclaimer with the value \"Treat with caution, content is AI-generated\" to clearly indicate that responses are AI-generated content and should be handled appropriately.
❗️There are three examples provided, each demonstrating different configurations. For examples related to assistants, please refer to the chatReferences (assistant) example. You can find all examples conveniently located under the Example section on the right side within the drop-down menu.
\nShort description of the body JSON fields:
\nmodel -> id = the id returned by the models endpoint
\nmessage = the chat messages, use role = user for the user sent content and role = assistant for the answer from the LLM. The fields references and sources, in this version of the API, should always be an empty array. User sent message are role = user, responses from the LLM are role = assistant and responses from the systemMedia endpoint are role = system.
\nroleId = should be an empty string, or the id of a shared role with the API user.
\ntemperature = defines the creativity of the used LLM model. The value is between 0.0 and 1.0. With a low value the model is more constant and with a higher value it is more random
\nselectedMode = possible values are BASIC, QA and SEARCH.
\nBASIC mode performs a standard call to the LLM. If files are selected it will use the whole content of the file as input.
\nQA mode uses the vector search on selected files and/or data collections before sending the message to the LLM.
\nSEARCH mode leverages real-time web data to enhance responses before processing through the language model.
\nselectedFiles (optional) = The uniqueTitles of the selectedFiles which should be used for the call.
\nselectedDataCollections (optional) = The ids of the selected data collections which should be used for the call. Data collections are only possible in QA mode!
\nselectedAssistantId (optional) = Use this field to specify a pre-configured assistant for the chat endpoint. When an assistantId is provided, only the messages and selectedAssistantId fields are required. You may optionally include additional files and data collections via the selectedFiles and selectedDataCollections fields. All other fields will be ignored, as the assistant’s configuration will be loaded automatically from the backend.
Important: If you include files or data collections in selectedFiles or selectedDataCollections, ensure that the API user has the necessary access permissions to these media elements.
Response Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files, data collections or roles
\n413 - Payload to large = message greater than context window of model
\n429 - Too many requests = to many requests to azure open ai in a short period, request limit
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nOptional - Turns the 506 internal system prompts on or off. The default value is true.
\n","type":"text/plain"},"key":"internalSystemPrompt","value":"true"},{"description":{"content":"Optional - Returns also the cites for mapping the inline references with the chunks. Only if internalSystemPrompt is true.
\n","type":"text/plain"},"key":"citation","value":"true"}],"variable":[]}},"response":[{"id":"6a13d488-4835-463a-877d-af545ba22164","name":"chatReferences","originalRequest":{"method":"POST","header":[{"key":"Content-Type","value":"application/json","type":"text"},{"key":"api-organization-id","value":"This endpoint provides a straightforward way to receive a single text answer from the LLM in JSON format, without streaming. The response includes the answer in the \"content\" field, along with a \"references\" section that details each relevant file. Each reference contains a uniqueTtitle, the title of the file, and the associated chunks. You can use the uniqueTitle and chunk number (from chunks array) to make a GET chunk call, allowing you to retrieve the specific content of that chunk.
\nTo obtain a mapping of the inline references to their corresponding chunks in the file, you can utilize the citation parameter by setting its value to true. This will result in a response that includes a cites array for each file, with the same number of entries as the chunk array. The first entry in the cites array corresponds to the inline reference number of the content, while the first entry in the chunk array represents the related chunk, and this pattern continues for the subsequent entries. The references and cites are only returned in QA mode!
Please note that the citation parameter works only when the internalSystemPrompt parameter is true. Additionally, there is a second example showcasing the citation parameter set to true. You can find this option in the Example request section by selecting it from the dropdown menu on the right side, labeled as chatNoStream (citation).
❗️The API includes a custom header api-ai-disclaimer with the value \"Treat with caution, content is AI-generated\" to clearly indicate that responses are AI-generated content and should be handled appropriately.
❗️There are three examples provided, each demonstrating different configurations. For examples related to assistants, please refer to the chatNoStream (assistant) example. You can find all examples conveniently located under the Example section on the right side within the drop-down menu.
\nShort description of the body JSON fields:
\nmodel -> id = the id returned by the models endpoint
\nmessage = the chat messages, use role = user for the user sent content and role = assistant for the answer from the LLM. The fields references and sources, in this version of the API, should always be an empty array. User sent message are role = user, responses from the LLM are role = assistant and responses from the systemMedia endpoint are role = system.
\nroleId = should be an empty string, or the id of a shared role with the API user.
\ntemperature = defines the creativity of the used LLM model. The value is between 0.0 and 1.0. With a low value the model is more constant and with a higher value it is more random
\nselectedMode = possible values are BASIC, QA and SEARCH.
\nBASIC mode performs a standard call to the LLM. If files are selected it will use the whole content of the file as input.
\nQA mode uses the vector search on selected files and/or data collections before sending the message to the LLM.
\nSEARCH mode leverages real-time web data to enhance responses before processing through the language model.
\nselectedFiles (optional) = The uniqueTitles of the selectedFiles which should be used for the call.
\nselectedDataCollections (optional) = The ids of the selected data collections which should be used for the call. Data collections are only possible in QA mode!
\nselectedAssistantId (optional) = Use this field to specify a pre-configured assistant for the chat endpoint. When an assistantId is provided, only the messages and selectedAssistantId fields are required. You may optionally include additional files and data collections via the selectedFiles and selectedDataCollections fields. All other fields will be ignored, as the assistant’s configuration will be loaded automatically from the backend.
Important: If you include files or data collections in selectedFiles or selectedDataCollections, ensure that the API user has the necessary access permissions to these media elements.
Response Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files, data collections or roles
\n413 - Payload to large = message greater than context window of model
\n429 - Too many requests = to many requests to azure open ai in a short period, request limit
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nOptional - Turns the 506 internal system prompts on or off. The default value is true.
\n","type":"text/plain"},"key":"internalSystemPrompt","value":"true"},{"description":{"content":"Optional - Returns also the cites for mapping the inline references with the chunks. Only if internalSystemPrompt is true.
\n","type":"text/plain"},"key":"citation","value":"true"}],"variable":[]}},"response":[{"id":"3acf53cc-579e-4ff1-9cf1-9a2cccfe8e44","name":"chatNoStream","originalRequest":{"method":"POST","header":[{"key":"Content-Type","value":"application/json","type":"text"},{"key":"api-organization-id","value":"This endpoint generates a structured JSON output based on the provided content input. You can leverage existing files and data collections within the 506.ai Platform to create your structured output. When files or data collections are selected, their content is automatically added to the content field. If you prefer to work exclusively with files and/or data collections, simply leave the content field empty.
\nRequest Configuration
The endpoint accepts several configuration fields to customize the output generation:
Mandatory Fields:
model - Specifies which AI model to use for generation
content - The text content to process (can be empty if using only files/data collections)
selectedFiles - Files from the 506.ai Platform to include (can be empty)
selectedDataCollections - Data collections to include (can be empty)
selectedMode - The processing mode for generation
jsonStructure - Defines the structure of the output JSON
Defining the JSON Structure:
The jsonStructure object determines how your output will be formatted. It requires two mandatory fields:
name - A descriptive name for your structure
schema - An array defining the properties (must not be empty)
Each schema entry includes:
\ndescription - Explains what the field represents
properties - Defines the actual data structure
Supported Data Types and Properties
Available Data Types:
string - Text values
number - Numeric values with decimals
boolean - true/false values
integer - Whole numbers
object - Nested structures
array - Lists of items
Supported string properties:
A regular expression that the string must match. — \"pattern\": \"^@[a-zA-Z0-9_]+$\"
A categorical object that the string must match. — \"enum\": [\"°F\", \"°C”, \"K\"]
Supported number properties:
The number must be less than or equal to this value. — \"maximum\": 80
The number must be greater than or equal to this value. — \"minimum\": -10
Supported array properties:
\n\"items\": {\n \"properties\": {\n \"name\": {\n \"type\": \"string\",\n \"description\": \"The name of the location, for Example Linz, Wels, Wien\"\n }\n }\n}\n\nThis flexible configuration allows you to create precisely structured outputs tailored to your specific data requirements, whether you're working with direct content input, existing files, or data collections within the 506.ai Platform.
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to create the data collection
\n406 - Not Acceptable = not allowed model id
\n429 - Too many requests = to many requests in a short period, request limit (20 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint is dedicated to classifying content using the Mistral moderation model. It processes requests without streaming, meaning all data is sent and processed in a single, complete operation. Request Details:
\nInput Format: JSON
\nRequired Fields:
\nmodel->id: The identifier of the moderation model to use.
content: The text content to be classified.
Output: JSON object containing the classification result for the submitted content.
\nThis concise, non-streaming endpoint offers a straightforward way to perform content moderation, ensuring rapid and reliable feedback by leveraging the specified model.
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n406 - Not Acceptable = not allowed model id
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform
\nThis endpoint is an additional feature and does not have access to the 506.ai Platform data collections. If you would like to activate the mistral moderation model, please get in touch with our dedicated 506 support team.
\n","urlObject":{"protocol":"https","port":"3003","path":["api","v1","public","moderation"],"host":["In this guide, we will explore the fundamental endpoints available for interacting with Large Language Models (LLMs). These endpoints allow you to harness the power of LLMs for various applications, enhancing your projects with advanced language processing capabilities.
\nModels: Use this endpoint to retrieve the list of available LLM models. This feature enables you to select the most suitable model for your specific needs, ensuring optimal performance and results.
\nRoles: This endpoint provides access to the configured roles within 506.ai Platform. By understanding the different roles, you can tailor your interactions and leverage the capabilities of the LLM more effectively.
\nAssistants: This endpoint allows access to the assistants configured within the 506.ai Platform client. Use it to retrieve information about the available assistants and interact with them as needed.
\nChat: Engage in dynamic conversations with your chosen LLM through this endpoint. The chat functionality allows for interactive dialogues, making it easier to obtain information, generate content, or explore ideas.
\nThis endpoint returns the id, name, createdAt and updatedAt of the availabe shared data collections. The id can be used in the chat or mediaByDataCollection calls.
\nResponse Codes:
\n204 - No Content = the returned list of data collections is empty
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint returns the id, name, createdAt, updatedAt and dataCollectionDeletable of all data collections where the API user is owner, admin or user. The id can be used in the chat, mediaByDataCollection or deleteDataCollection calls.
\nResponse Codes:
\n204 - No Content = the returned list of data collections is empty
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint returns the paginated media elements of the shared data collection with the given dataCollectionId. The call uses pagination where the page can be set with the parameter pageNumber (default value = 1) and the amount of elements per page can be set with the parameter pageSize (default value = 5).
\nThe available data collections can be fetched with the dataCollections call.
\nResponse Codes:
\n204 - No Content = the returned page of media elements is empty
\n400 - Bad Request = one or more mandatory fields are missing or empty or pageNumber = 0 or pageSize = 0
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe number of the returned page (integer). The minimum value is 1. Default value is set to 1.
\n","type":"text/plain"},"key":"pageNumber","value":"The amount of returned media elements (integer) per page. The minimum vlalue is 1. Default value is set to 5.
\n","type":"text/plain"},"key":"pageSize","value":"This endpoint returns the paginated media elements of the shared data collection with the given dataCollectionId. The call uses pagination where the page can be set with the parameter pageNumber (default value = 1) and the amount of elements per page can be set with the parameter pageSize (default value = 5).
\nYou can easily retrieve the available data collections using the dataCollections V2 call. This updated version introduces a new field called \"status,\" which indicates the current state of the file. Please note that if the uploaded files are larger, it may take some time for all the chunks to be fully embedded. There are three possible status values you might encounter:
\nUNREADY: This means the file is still in the embedding process and is not yet ready for use.
\nREADY: This indicates that the file is fully prepared and ready to be utilized.
\nERROR: This status signifies that an error has occurred, preventing the file from being used.
\nResponse Codes:
\n204 - No Content = the returned page of media elements is empty
\n400 - Bad Request = one or more mandatory fields are missing or empty or pageNumber = 0 or pageSize = 0
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint provides the content of a specified chunk (chunkNr) for a particular file (uniqueTitle). The returned JSON response also includes the page number (pageNr) where the chunk can be found in the original file.
\nPlease note: It is essential that the uniqueTitle value is URL encoded!
\n🆕 Optional parameter assistantId: Include this parameter if you want to retrieve the chunk associated with a chat where an assistant was involved.
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty or chunkNr < 1
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe uniqueTitle (url encoded) of the file to get the chunk content.
\n","type":"text/plain"},"key":"uniqueTitle","value":"The chunk number (integer) of the file.
\n","type":"text/plain"},"key":"chunkNr","value":"Optional - The used assistantId.
\n","type":"text/plain"},"key":"assistantId","value":"This endpoint allows you to retrieve the original file using the specified uniqueTitle. The response will be provided as a byte array. The .docx format, along with the transcripts of audio and video files, will be stored internally as .pdf documents and returned in that format. All other file types will be provided in their original formats. You can retrieve the uniqueTitle of the files using the mediaByDataCollection call.
Please ensure that the uniqueTitle is URL-encoded for proper processing. Additionally, keep in mind that only files up to a maximum size of 50MB can be retrieved.
🆕 Optional parameter assistantId: Include this parameter if you want to retrieve the original file associated with a chat where an assistant was involved.
\nResponse Codes:
\n400 - Bad Request = uniqueTitle is missing or empty or file not available
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions on files or data collections
\n413 - Payload too large = file to retrieve is bigger than 50MB
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe uniqueTitle (url encoded) of the file to get the original representation.
\n","type":"text/plain"},"key":"uniqueTitle","value":"Optional - The used assistantId.
\n","type":"text/plain"},"key":"assistantId","value":"The key for the files.
\n","value":null}]},"url":"https://This endpoint is designed for uploading files to a data collection. To use it, please send your files in the request body as form-data, using the key media. The ID of the data collection should be included as a query parameter named folderId. Additionally, the replace parameter indicates whether existing files should be replaced if they are already present.
The following file types are supported for upload:
\n*.txt
\n*.docx
\n*.pptx
\n*.xlsx
\n*.csv
\n*.json
\n*.xml
\n*.eml
\n*.msg
\n*.md
\n*.rtf
\n*.py
\n*.java
\n*.js
\nAudio (M4A, MPEG, MP3, WAV, AIFF, AAC, OGG, FLAC)
\nVideo (MP4)
\nUpon completion of your request, the response will be in JSON format, providing detailed information for each file, including its status code. This will help you determine whether a file was successfully uploaded, already exists, or is too large.
\nPlease be aware of the following limitations:
\nA maximum of 3000 chunks is allowed per request. If one or more files exceed this limit, they will receive their own status code. For instance, if you upload ten files and nine are successful while one exceeds the chunk limit, you will need to resend that specific file in a separate request. If a single file exceeds the 3000 chunks on its own, it cannot be uploaded.
\nThere is also a limit of 300MB per request, which means you can upload one file up to 300MB or multiple files with a combined total of 300MB.
\nLastly, please note that there is a rate limit of 10 requests per minute.
\nFile Response Codes:
\n400 = no file part in request or no files selected
\n409 = file already exists
\n413 = chunk limit exeeded
\n415 = unsupported file type
\n422 = file too large
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to upload or replace file(s) in the given dataCollectionId
\n429 - Too many requests = to many requests in a short period, request limit (10 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe identifier for the data collection where the files will be stored.
\n","type":"text/plain"},"key":"dataCollectionId","value":"Set to true to replace already existing files with the same name but false should be used as default.
\n","type":"text/plain"},"key":"replace","value":"false"}],"variable":[]}},"response":[{"id":"2d0b773d-4aa8-4674-b7fb-413b586c2618","name":"uploadMedia","originalRequest":{"method":"POST","header":[{"key":"api-organization-id","value":"The key for the files.
\n","value":null}]},"url":"https://The extractText endpoint is designed to efficiently extract text from one or more files. This functionality is particularly useful for applications that require text analysis, extraction, or content processing. The request body must be sent as form-data. The key for the files should be media.
The following file types are supported for upload:
\n*.txt
\n*.docx
\n*.pptx
\n*.xlsx
\n*.csv
\n*.json
\n*.xml
\n*.eml
\n*.msg
\n*.md
\n*.rtf
\n*.py
\n*.java
\n*.js
\nFor OCR, please use the dedicated extractTextOcr endpoint, and for video or audio files, use the transcribeMedia endpoint.
Upon completion of your request, the response will be in JSON format, providing the content of each file, including its status code.
\nPlease be aware of the following limitations:
\nThere is also a limit of 300MB per request, which means you can upload one file up to 300MB or multiple files with a combined total of 300MB.
\nLastly, please note that there is a rate limit of 15 requests per minute.
\nFile Response Codes:
\n400 = no file part in request or no files selected
\n415 = unsupported file type
\n422 = file has no text (for example Screenshot, no OCR in this endpoint)
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to upload file(s)
\n429 - Too many requests = to many requests in a short period, request limit (15 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe key for the files.
\n","value":null}]},"url":"https://The extractTextOcr endpoint provides efficient text extraction from scanned documents using advanced OCR technology. This endpoint is perfect for digitizing printed or handwritten content from your PDF files.
\nSend your request using form-data format with files attached under the media key.
Currently supports:
\nThe endpoint returns a JSON response containing:
\nExtracted text content from each processed file
\nIndividual status codes for each file processed
\nFile Size: Maximum 300MB per request
\nUpload a single file up to 300MB, or
\nMultiple files with combined size not exceeding 300MB
\nRate Limiting: 10 requests per minute
\nFile Response Codes:
\n400 = no file part in request or no files selected
\n422 = unsupported file type
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to upload file(s)
\n429 - Too many requests = to many requests in a short period, request limit (10 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThe key for the files.
\n","value":null}]},"url":"https://The transcribeMedia endpoint provides efficient text extraction from audio and video files.
\nSubmit your request using form-data with files attached under the media key.
Audio Formats:
\nM4A, MPEG, MP3, WAV
\nAIFF, AAC, OGG, FLAC
\nVideo Formats:
\nThe API returns transcription results in JSON format, including:
\nExtracted text content for each processed file
\nIndividual status codes per file
\nFile Size: Maximum 300MB per request
\nSingle file up to 300MB, or
\nMultiple files with combined size ≤ 300MB
\nRate Limiting: 10 requests per minute
\nThis endpoint offers flexible file processing while maintaining reasonable resource constraints for optimal performance.
\nFile Response Codes:
\n400 = no file part in request or no files selected
\n415 = unsupported file type
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to upload file(s)
\n429 - Too many requests = to many requests in a short period, request limit (10 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint allows you to create a new data collection with a specified name. The API user will be designated as the owner of this collection. Additionally, all admin users from the 506.ai Platform will automatically be granted admin access to the data collection, enabling them to manage the folders through the 506.ai Platform client effectively. The response in JSON format contains the details of the newly created folder.
\nPlease note that there is a rate limit for this endpoint, allowing up to 20 requests per minute.
\nShort description of the body JSON fields:
\nResponse Codes:
\n400 - Bad Request = one or more mandatory fields are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to create the data collection
\n429 - Too many requests = to many requests in a short period, request limit (20 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint provides the functionality to delete one or more files. You can remove files for which the API user is the owner, as well as files within data collections where the API user holds admin privileges. Upon successful deletion, the response will be a JSON array containing details about all the files that have been deleted. File which are not in READY state cannot be deleted!
\nResponse Codes:
\n400 - Bad Request = uniqueTitles are missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to delete one or more files
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nThis endpoint provides a convenient way to delete a data collection using its ID. To initiate the deletion, simply include the ID as a path parameter. It's important to remember that only data collections owned by the API user can be deleted.
\nAdditionally, the dataCollections V2 response includes a field called dataCollectionDeletable, which indicates whether a specific data collection can be deleted via the API.
Please be aware that there is a rate limit in place for this endpoint, allowing up to 20 requests per minute.
\nResponse Codes:
\n400 - Bad Request = data collection ID in the path is missing or empty
\n401 - Unauthorized = wrong organization id or api key
\n403 - Forbidden = no permissions to delete the data collection
\n429 - Too many requests = to many requests in a short period, request limit (20 per minute)
\n500 - Internal Server Error = an internal error in the backend of the 506.ai Platform occurred
\nIn this guide, we will explore the various media endpoints available for managing data collections and media files. Here’s a brief overview of the key functionalities:
\ndataCollections V2: This endpoint retrieves data collections created by the API user and those that have been shared with them.
\nmediaByDataCollections V2: With this endpoint, users can retrieve all files associated with a specific data collection by providing the corresponding dataCollectionId.
chunk: This functionality enables users to obtain a specific chunk of a file by specifying its uniqueTitle and chunkNr.
originalMedia: This endpoint can access the original version of a file by providing its uniqueTitle.
uploadMedia: This endpoint facilitates the uploading of files to a designated data collection by using the dataCollectionId.
createDataCollection: This endpoint creates a new data collection by specifying a name.
\ndeleteDataCollection: A data collection can be deleted by providing the relevant id.
These endpoints collectively enhance the user experience by providing comprehensive tools for managing media collections and files effectively.
\n","_postman_id":"6a735265-9509-4ccb-90aa-e755b93074ab"}],"id":"728afb96-32af-4ba2-940d-63fc64dd7623","description":"To gain access to the API, an admin must first create an API key via the 506.ai Platform client. This process involves the following steps:
\nAPI Key Generation:
\nNavigate to the admin settings page within the 506.ai Platform client.
\nLocate the section API and generate a new API key.
\nOrganization ID:
\nRequest Headers
\nFor each API request, two critical pieces of information must be included in the header fields:
\napi-organization-id: This header field should contain the organization ID.
\napi-key: This header field should contain the API key generated by the admin.
\nThe 506 API ensures secure communication and data integrity through the use of API keys and organization IDs, making sure that only authorized users can access and interact with the API.
\nBy following the above steps for authentication and utilizing the provided endpoints, developers can efficiently integrate the 506 API into their applications, enhancing functionality with advanced chat capabilities and reliable data access.
\nIn this guide, we'll explore the various API calls available and the sequence in which they could be executed. Whether you're a seasoned developer or just getting started, this step-by-step walkthrough will help you understand how to effectively interact with our API.
\nThe initial API call, known as the POST chat call, is crafted to enable straightforward chat interactions. It requires only the model ID as an additional configuration. This call lays the groundwork for starting a conversation between the user and the system.
\nFirst, make a call to the GET models endpoint. In the response JSON, you will find the available LLM models. Below is an example of what the response JSON might look like:
\nGET models
\n[\n {\n \"id\": \"gpt-4o\",\n \"name\": \"GPT-4 Omni\"\n },\n {\n \"id\": \"gpt-4o-mini\", // used id for upcoming chat call\n \"name\": \"GPT-4o Mini \"\n }\n]\n\nFirst, obtain the value from the id field of the model you wish to use. Then, include this value as the input for the model > id field in the body of the POST request to the chat endpoint. Below is an example of the JSON body:
\nPOST chat
\n{\n \"model\": {\n \"id\": \"gpt-4o-mini\" // used id from models call above\n },\n \"messages\": [\n {\n \"role\": \"user\",\n \"content\": \"Wer bist du?\",\n \"references\": [],\n \"sources\": []\n }\n ],\n \"roleId\": \"\",\n \"temperature\": 0.2,\n \"selectedMode\": \"BASIC\"\n}\n\nTo use a specific role that you've configured in CompanyGPT during your chat session, you'll need to make a GET request to retrieve the available roles. Once you have the response, you can use the roleId of the desired role as an input in the body of your chat call. Below is an example of the JSON response from the GET roles call:
\nGET roles
\n[\n {\n \"roleId\": \"2984997d-892d-49ac-af5b-0ee43087dcc1\",\n \"title\": \"The Creative Copywriter\",\n \"createdAt\": \"2024-04-23T13:03:02.022Z\",\n \"updatedAt\": \"2024-10-17T13:12:22.912Z\"\n },\n {\n \"roleId\": \"06a21d40-e4ee-4219-b4b6-6b1064010051\", // used roleId for\n \"title\": \"The Data Analyst\", // upcoming chat call \n \"createdAt\": \"2024-04-23T13:07:29.838Z\",\n \"updatedAt\": \"2024-08-27T13:35:27.838Z\"\n }\n]\n\nHere is an example of the body JSON of the POST chat endpoint with the used Role “The Data Analyst”:
\nPOST chat
\n{\n \"model\": {\n \"id\": \"gpt-4o-mini\"\n },\n \"messages\": [\n {\n \"role\": \"user\",\n \"content\": \"Wer ist Einstein?\",\n \"references\": [],\n \"sources\": []\n }\n ],\n \"roleId\": \"06a21d40-e4ee-4219-b4b6-6b1064010051\", // used roleId from roles \n \"temperature\": 0.2, // call above\n \"selectedMode\": \"BASIC\"\n}\n\nTo request one or more data collections that you've set up in CompanyGPT, you need to use the GET dataCollections call. Simply include the ID of the desired data collections in the body of your chat call. Below is an example of the response JSON from the GET dataCollections call:
\nGET dataCollections
\n[\n {\n \"id\": \"0eb75a35-ad02-4506-98e8-ddc88f422a25\", // used id for upcoming chat\n \"name\": \"Test Collection1\", // call\n \"createdAt\": \"2024-03-14T13:18:09.002Z\",\n \"updatedAt\": \"2024-10-28T10:27:44.783Z\"\n },\n {\n \"id\": \"cf49aac9-2362-4576-b485-7039e5fa51b7\",\n \"name\": \"Test Collection2\",\n \"createdAt\": \"2024-03-25T08:31:32.006Z\",\n \"updatedAt\": \"2024-08-26T12:37:51.951Z\"\n }\n]\n\nHere is an example of the JSON body for the POST chat call using the data collection \"Test Collection1\". It's important to set the selectedMode to QA if you want to query data collections:
\nPOST chat
\n{\n \"model\": {\n \"id\": \"gpt-4o-mini\"\n },\n \"messages\": [\n {\n \"role\": \"user\",\n \"content\": \"Welche Prompts gibt es?\",\n \"references\": [],\n \"sources\": []\n }\n ],\n \"roleId\": \"06a21d40-e4ee-4219-b4b6-6b1064010051\",\n \"temperature\": 0.2,\n \"selectedDataCollections\":[ // used id from dataCollections\n \"0eb75a35-ad02-4506-98e8-ddc88f422a25\" // call above\n ],\n \"selectedMode\": \"QA\" // selected mode QA\n}\n\nTo request one or more individual files, you need to follow these steps:
\nFirst, make a request to the GET dataCollections endpoint to retrieve the available data collections.
\nNext, use the GET mediaByDataCollection call to obtain the files within the specified data collection.
\nHere is an example of the JSON response from the GET dataCollections call:
\nGET dataCollections
\n[\n {\n \"id\": \"0eb75a35-ad02-4506-98e8-ddc88f422a25\", // used id for upcoming\n \"name\": \"Test Collection1\", // mediaByDataCollection call\n \"createdAt\": \"2024-03-14T13:18:09.002Z\",\n \"updatedAt\": \"2024-10-28T10:27:44.783Z\"\n }\n]\n\nTo retrieve the files, use the ID of the desired data collection as a path parameter in the GET mediaByDataCollection call. The pageNumber and pageSize parameters are used for pagination, as detailed in the endpoint documentation. Below is an example of how to use the GET mediaByDataCollection call:
\n/api/v1/public/mediaByDataCollection/0eb75a35-ad02-4506-98e8-ddc88f422a25?pageNumber=1&pageSize=2
\nHere an example of the response JSON:
\nGET mediaByDataCollection
\n{\n \"pageNumber\": 1,\n \"totalPages\": 17,\n \"pageSize\": 2,\n \"media\": [\n {\n \"title\": \"Test1.pdf\",\n \"uniqueTitle\": \"0eb75a35-ad02-4506-98e8-ddc88f422a25_Test1.pdf\", // used\n \"mediaType\": \"PDF\", // uniqueTitle for the upcoming\n \"mediaSize\": 3148, // chat call\n \"mediaPath\": \"\",\n \"tokenSize\": 997,\n \"originalSize\": 337740,\n \"url\": \"\",\n \"baseUrl\": \"\",\n \"createdAt\": \"2024-10-08T09:59:56.764Z\"\n },\n {\n \"title\": \"Test2.pdf\",\n \"uniqueTitle\": \"0eb75a35-ad02-4506-98e8-ddc88f422a25_Test2.pdf\",\n \"mediaType\": \"PDF\",\n \"mediaSize\": 1278,\n \"mediaPath\": \"\",\n \"tokenSize\": 480,\n \"originalSize\": 89577,\n \"url\": \"\",\n \"baseUrl\": \"\",\n \"createdAt\": \"2024-10-08T09:59:56.000Z\"\n }\n ],\n \"totalElements\": 34\n}\n\nIn the following step, utilize one or more uniqueTitles from the resulting files as input in the body of the POST chat endpoint. You have the option to use either the QA or BASIC mode. It's crucial to note that large files or multiple files in BASIC mode can exceed the context window. Therefore, it is advisable to use the QA mode. Below is an example of the body JSON:
\nPOST chat
\n{\n \"model\": {\n \"id\": \"gpt-4o-mini\"\n },\n \"messages\": [\n {\n \"role\": \"user\",\n \"content\": \"Was steht im Dokument?\",\n \"references\": [],\n \"sources\": []\n }\n ],\n \"roleId\": \"06a21d40-e4ee-4219-b4b6-6b1064010051\",\n \"temperature\": 0.2,\n \"selectedFiles\":[\n \"0eb75a35-ad02-4506-98e8-ddc88f422a25_Test1.pdf\" // used uniqueTitle from\n ], // mediaByDataCollection\n \"selectedMode\": \"QA\" // call above\n}\n\nThe only required value is the model ID. All other values, such as roleId, selectedFiles, or selectedDataCollections, are optional and can be used in various combinations. This flexibility allows you to tailor the configuration to your specific needs.
\nThis served as an introduction to the API's functionalities. Following the next update, the GET chunk call can be utilized to display the referenced chunk of the results in QA mode.
\n","_postman_id":"c12e9f0b-e25d-4223-9663-5cfda6efd1c7"}]}