# AIVAX AI model inference, document indexing and search. Helpful links: - API endpoint: https://inference.aivax.net/ - Reference for agents: https://inference.aivax.net/apidocs/llms.txt - Documentation/guides: https://docs.aivax.net/ - Terms of service: https://docs.aivax.net/docs/en/legal/terms-of-service - Privacy Policy: https://docs.aivax.net/docs/en/legal/privacy-policy API Version: 1 # AIVAX Overview AIVAX is an AI orchestration platform that unifies inference, knowledge retrieval, and agentic tools into a single API. With it, you create intelligent assistants based on your knowledge base, connected to external tools and ready to operate across multiple channels — all without managing model infrastructure. ## What does AIVAX solve? Building AI assistants that go beyond generic responses requires integrating multiple services: model hosting, semantic search, tool calling, moderation, conversational memory, and communication channels. Each piece adds complexity, cost, and points of failure. AIVAX consolidates all this into a single layer. You configure an **AI Gateway**, connect your knowledge base, enable tools, and publish — without provisioning servers, training models, or maintaining separate embedding pipelines. ## Platform services ### AI Gateways The AI Gateway is the core of AIVAX. It works as a configurable agent engine that aggregates the language model, MCP servers, tools, skills, instructions, RAG context and other platform resources behind one inference endpoint. With a gateway you define: - **System instructions** that guide the agent's behavior - **Language model** — use one of the models hosted by AIVAX or bring your own (BYOK) via an OpenAI‑compatible endpoint - **Agent capabilities** such as MCP servers, built‑in tools, protocol functions and skills - **Inference parameters** such as temperature, top_p and prefill - **Processing pipelines** that chain RAG, moderation, context truncation and post‑processing The gateway exposes an endpoint compatible with the OpenAI API (`chat/completions`), which means immediate integration with existing SDKs in Python, Node.js, C# and any language that supports the OpenAI interface. **Use cases:** customer support assistants, internal copilots, sales agents, educational tutors, and any application that needs a parametrized and grounded AI model. ### RAG (Retrieval‑Augmented Generation) The RAG service allows your agent to answer based on real documents instead of relying solely on the model's pre‑trained knowledge. - **Collections**: semantic knowledge libraries that store up to tens of thousands of documents per collection, automatically indexed with embeddings - **Documents**: individual pieces of collection knowledge, such as article chunks, manual sections, product descriptions or other RAG items - **Semantic search**: retrieve the most relevant passages for each user query, with minimum relevance control and result count - **Multiple embedding strategies**: from full query rewrite to concatenation with conversational context, allowing precision vs. cost tuning - **Native integration**: the RAG pipeline injects retrieved documents directly into the model context, without additional code **Use cases:** intelligent FAQs, technical manual search, corporate knowledge bases, product catalogs, legal and regulatory documentation. ### Tools and Function Calling AIVAX offers a complete ecosystem of tools that turn your agent from a conversational model into an agentic system capable of performing actions. #### Built‑in tools Ready‑to‑use tools, without additional configuration: | Tool | What it does | |---|---| | **Web search** | Retrieves up‑to‑date information on the internet | | **Advanced search** | Automated browser navigation for deep searches | | **Code execution** | Runs JavaScript in a sandboxed environment for calculations and processing | | **URL context** | Analyzes up to 5 URLs simultaneously, extracting relevant content | | **Image generation** | Creates images at various quality levels | | **X/Twitter search** | Retrieves real‑time posts | | **Document generation** | Creates PDFs from HTML | | **Page hosting** | Publishes HTML pages for reports and landing pages | | **Memory** | Stores context between conversations for personalized experiences | #### External integrations - **MCP (Model Context Protocol)**: connect external MCP servers to expand the agent's capabilities with your own tools - **Protocol functions**: define custom functions with webhook callbacks, with encrypted nonce authentication - **Shell**: virtual bash environment with standard Unix tools for agents that need to manipulate data **Use cases:** research agents that query the web, assistants that generate PDF reports, bots that perform financial calculations, systems that integrate with internal APIs via webhooks. ### Skills Skills are sets of specialized instructions loaded on demand. Unlike system instructions (which are always active), skills are activated only when the model identifies them as relevant to the current conversation. This allows creating versatile agents with deep knowledge across multiple areas without consuming unnecessary tokens. **Use cases:** support agent that switches between technical and commercial tone based on client profile, legal assistant that loads area‑specific rules, tutor that adapts teaching style per discipline. ### Structured responses (JSON Healing) The structured response service ensures that the model's output conforms to a defined JSON Schema — even for models that do not natively support structured outputs. When the model's response is invalid, AIVAX runs automatic validation and correction cycles, providing feedback to the model until the JSON complies. It works with any LLM and preserves the reasoning phase in reasoning models. **Use cases:** extracting data from free text, conversational forms, data pipelines that consume model output, integrations with systems requiring strict format. ### Chat Clients AIVAX offers ready‑to‑use chat clients to connect your agent to end users: - **Web chat**: embeddable widget with visual customization, session management and conversation persistence - **Telegram**: native integration with special commands and user identification - **WhatsApp**: integration via Z‑Api, Evolution API or Kapso with conversation‑based sessions A chat client is the bridge between an AI Gateway and a controlled external environment, such as an iframe widget, WhatsApp or Telegram. Each chat client session represents one active conversation context, including its access key, metadata, messages, expiration and usage state. **Use cases:** website customer service, WhatsApp support, informational bots on Telegram, query interfaces embedded in web applications. ### Conversations Conversations provide observability for transactional and session-based interactions. They record messages, usage resources, metadata and timing information so teams can monitor agent behavior, inspect production exchanges, export conversation data and analyze how gateways are being used. **Use cases:** quality monitoring, support audits, usage analysis, debugging unexpected answers and exporting conversation history for offline review. ### Batch Processing Batch workflows define reusable processing models for large asynchronous workloads. A workflow describes how each input should be handled, while each batch job is a concrete execution created from that workflow. Batch jobs are divided into batch items. Each item represents one unit of work with its own input, output, state, confidence, processing timestamp and cost tracking, allowing the platform to retry failures, monitor progress and process large queues safely. **Use cases:** bulk classification, document analysis, dataset enrichment, evaluation runs and other workloads where many independent inputs must be processed by the same AI workflow. ### Workers Workers are remote hooks triggered by inference pipeline events. They allow intercepting and controlling the processing flow at specific points, such as when receiving a message or before sending a response. **Use cases:** data validation before processing, context enrichment with external system information, custom logging, conditional flow control. ### Moderation Security layer that analyzes the model's responses with an auxiliary model before delivering them to the user, ensuring compliance with content policies. **Use cases:** public‑facing applications, regulated environments, children's or educational products. ## How the services integrate AIVAX services are not isolated pieces — they form a cohesive pipeline where each component amplifies the others: ``` User → Chat Client → AI Gateway │ ┌───────────┼───────────┐ ▼ ▼ ▼ RAG Tools Skills (documents) (actions) (instructions) └───────────┼───────────┘ ▼ Processed response (moderation, JSON healing) ▼ User ``` 1. The user message arrives via **chat client** or direct **API** 2. The **AI Gateway** orchestrates the processing: loads instructions, queries the knowledge base via **RAG**, makes **tools** available and activates **skills** as needed 3. The model generates the response, which goes through validation (**JSON healing** if configured) and **moderation** 4. The response is delivered to the user through the same originating channel This flow is fully configurable. You can use only the gateway with RAG for a simple FAQ, or combine all services for a complex agent with multiple tools and sub‑agents via MCP. ## Getting started in minutes Creating an agent in AIVAX involves three steps: 1. **Create an AI Gateway** with the instructions and desired model 2. **Connect a collection** of documents (optional, for RAG) 3. **Integrate** via OpenAI‑compatible API, chat widget or messaging channel ```python from openai import OpenAI client = OpenAI( base_url="https://inference.aivax.net/v1", api_key="your_api_key" ) response = client.chat.completions.create( model="my-agent:50c3", messages=[ {"role": "user", "content": "What are the return policies?"} ] ) print(response.choices[0].message.content) ``` No proprietary SDKs, no lock‑in. The same integration that works with OpenAI works with AIVAX. ## Pricing model AIVAX operates with a **pre‑pay** model combined with subscription plans. To compare differences between Free, Pro and Max, see the [AIVAX pricing page](https://aivax.net/pricing), which is kept as an up‑to‑date source for plans. Inference costs are passed through directly from providers without markup — you pay the same list price you would using Google or OpenAI directly. For details on balance billing, storage and tools, see the [pricing page](https://aivax.net/docs/en/pricing). ## Next steps - [Getting started](https://aivax.net/docs/en/getting-started) — set up your account and make your first call - [Authentication](https://aivax.net/docs/en/authentication) — obtain and use your API key - [AI Gateways](https://aivax.net/docs/en/inference/ai-gateway) — configure your first agent - [Collections](https://aivax.net/docs/en/rag/collections) — import documents for RAG - [Tools](https://aivax.net/docs/en/tools/builtin-tools) — explore the available tools ## Inference - [Model listing](http://inference.aivax.net/apidocs/llms.txt?endpoint=Modellisting): Returns OpenAI-compatible model entries for AIVAX integrated models and AI Gateways owned by the authenticated account. - [Inference (chat completions)](http://inference.aivax.net/apidocs/llms.txt?endpoint=Inferencechatcompletions): Creates an OpenAI-compatible chat completion using an integrated model or an AI Gateway. Supports streaming, tool calls, multimodal messages... ## Hook Schedules - [Get Schedule](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetSchedule): Returns one pending, activated or failed one-time hook activation schedule for a web chat client. - [Cancel Schedule](http://inference.aivax.net/apidocs/llms.txt?endpoint=CancelSchedule): Deletes one hook activation schedule from a web chat client. - [Get Recurring Task](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetRecurringTask): Returns one recurring hook activation task, including cron expression, activation counts and next activation time. - [Cancel Recurring Task](http://inference.aivax.net/apidocs/llms.txt?endpoint=CancelRecurringTask): Deletes one recurring hook activation task from a web chat client. - [List Hook Schedules](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListHookSchedules): Lists one-time hook schedules, recurring hook tasks and recently inactive users for a web chat client. ## Web Chat - [Delete Web Chat Client Integration](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteWebChatClientIntegration): Removes one messaging integration configuration from a web chat client. - [Edit Web Chat Client Integrations](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditWebChatClientIntegrations): Updates integrations for a web chat client. integrationType is required. The selected integration payload may be partial; it is shallow-merg... - [View Web Chat Client](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewWebChatClient): Returns the web chat client's gateway link, rate limits, UI/client parameters and messaging integration settings. - [Edit Web Chat Client](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditWebChatClient): Updates an existing web chat client. The request may contain only the fields to change; omitted fields keep their current values. clientPara... - [Delete Web Chat Client](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteWebChatClient): Deletes a web chat client owned by the authenticated account. - [List Web Chat Clients](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListWebChatClients): Lists web chat clients owned by the authenticated account with active session counts and linked gateway summaries. - [Create Web Chat Client](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateWebChatClient): Creates a web chat client linked to an AI Gateway for browser chat sessions and optional messaging integrations. ## Web Chat Sessions - [Delete Web Chat Session](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteWebChatSession): Deletes a web chat session from a client, ending access through its session key. - [Create Web Chat Session](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateWebChatSession): Creates a web chat session for a client, or refreshes the existing session when the same tag already exists. - [List Web Chat Sessions](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListWebChatSessions): Lists up to 300 sessions for a web chat client, including context preview, tag, timestamps, message count and token count. ## Skills - [Import Skills (JSONL)](http://inference.aivax.net/apidocs/llms.txt?endpoint=ImportSkillsJSONL): Uploads and imports skills from a JSONL file. Each line must be a JSON object with 'slug' and 'instructions' at minimum. See the example for... - [Export Skills (JSONL)](http://inference.aivax.net/apidocs/llms.txt?endpoint=ExportSkillsJSONL): Exports all skills for the authenticated account in JSONL format. Each line is a JSON object with 'slug', 'description', 'instructions', and... - [Clear Skills](http://inference.aivax.net/apidocs/llms.txt?endpoint=ClearSkills): Deletes every skill owned by the authenticated account. - [Get Skill](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetSkill): Returns a skill's slug, instructions, description and tool/instruction-source options. - [Update Skill](http://inference.aivax.net/apidocs/llms.txt?endpoint=UpdateSkill): Updates an existing skill. The request may contain only the fields to change; omitted fields keep their current values. options is shallow-m... - [Delete Skill](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteSkill): Deletes a skill owned by the authenticated account. - [List Skills](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListSkills): Lists skills owned by the authenticated account with slug, description and allowed-tool count. - [Create Skill](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateSkill): Creates a reusable instruction package that can be enabled on AI Gateways and can limit which tools are visible. ## Shell I/O - [List Directory](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListDirectory): Lists files and directories inside a shell sandbox scoped by account and X-Shell-User-Id. - [Get File Public Address](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetFilePublicAddress): Creates a temporary public download URL for one file in the shell sandbox. - [Get File Details](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetFileDetails): Returns size, timestamps, MIME type and text-detection metadata for one file in the shell sandbox. - [Download File](http://inference.aivax.net/apidocs/llms.txt?endpoint=DownloadFile): Downloads one file from the account and X-Shell-User-Id scoped shell sandbox. - [Upload File](http://inference.aivax.net/apidocs/llms.txt?endpoint=UploadFile): Writes the request body to a file inside the shell sandbox, creating or replacing the file. Maximum request size is 100 MB. - [Delete File](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteFile): Deletes one file from the shell sandbox. - [Create Directory](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateDirectory): Creates a directory path inside the shell sandbox. - [Delete Directory](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteDirectory): Recursively deletes a non-root directory from the shell sandbox. ## RAG - [Rerank documents](http://inference.aivax.net/apidocs/llms.txt?endpoint=Rerankdocuments): Orders up to 1,024 documents using the smart or lexical reranker. Query and document text may contain at most 32,768 characters combined. Ea... - [Semantic search](http://inference.aivax.net/apidocs/llms.txt?endpoint=Semanticsearch): Searches one or more RAG collections with semantic similarity and returns matching documents, scores, metadata and optional referenced docum... - [Reflex search](http://inference.aivax.net/apidocs/llms.txt?endpoint=Reflexsearch): Semantic search across supplied document strings. - [Answer generation](http://inference.aivax.net/apidocs/llms.txt?endpoint=Answergeneration): Searches one or more RAG collections, then asks a model to answer using the retrieved documents and returns both the answer and source resul... ## Web Chat Client - [Download Session Transcript](http://inference.aivax.net/apidocs/llms.txt?endpoint=DownloadSessionTranscript): Downloads the session messages, context, and metadata as a JSON file. - [Retry Last Message](http://inference.aivax.net/apidocs/llms.txt?endpoint=RetryLastMessage): Removes all messages after the last user message, allowing the user to retry the conversation from that point. - [Send Prompt](http://inference.aivax.net/apidocs/llms.txt?endpoint=SendPrompt): Sends a prompt to the AI gateway and returns the completion result synchronously. - [Get Session Info](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetSessionInfo): Retrieves information about the current session, including messages, client configuration, and culture settings. ## Notifications - [Count Unread Notifications](http://inference.aivax.net/apidocs/llms.txt?endpoint=CountUnreadNotifications): Counts notifications that have not been read by the authenticated account. - [List Notifications](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListNotifications): Lists notifications for the authenticated account from newest to oldest. ## Memories - [Get Memory](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetMemory): Returns one persisted memory with full content, external user id, format and expiration time. - [List Memories](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListMemories): Lists persisted memories stored for the authenticated account, optionally filtered by text and memory format. - [Delete Memories](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteMemories): Deletes either one persisted memory by id or all memories for an external user id and format. ## Information - [Get Monthly Usage Report](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetMonthlyUsageReport): Returns monthly usage insights and top-cost model, resource and SKU tables for the authenticated account. - [Get Account Balance](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetAccountBalance): Returns billing and quota status for the authenticated account, including balance, last-24-hour usage, plan, storage usage, subscription-mod... ## Generations - [Generate speech](http://inference.aivax.net/apidocs/llms.txt?endpoint=Generatespeech): Synthesizes speech from text. Set raw to true to receive binary audio; otherwise the response contains base64-encoded audio in the standard ... - [Generate images](http://inference.aivax.net/apidocs/llms.txt?endpoint=Generateimages): Generates one to four images for each supplied prompt using the selected image model. Optional reference images are supported when the model... - [Describe media](http://inference.aivax.net/apidocs/llms.txt?endpoint=Describemedia): Resolves each OpenAI-compatible multimodal content part into text. Audio, images, video, PDF and other supported files are processed indepen... ## Conversations - [Export Conversations (JSONL)](http://inference.aivax.net/apidocs/llms.txt?endpoint=ExportConversationsJSONL): Exports conversations from the authenticated account as JSONL for offline analysis, audits, and monitoring workflows. - [Export Conversation (JSON)](http://inference.aivax.net/apidocs/llms.txt?endpoint=ExportConversationJSON): Exports a single conversation as a JSON file using the same export options available for the JSONL export. - [View Conversation](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewConversation): Returns full observability details for one stored conversation, including messages, request metadata, tools, response schema, usage object, ... - [Delete Conversation](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteConversation): Deletes a stored conversation from the authenticated account. - [List Conversations](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListConversations): Lists recent conversations for observability, monitoring, and analysis. The list includes message previews, model, token usage, external use... ## Collections - [Update Collection Vectors](http://inference.aivax.net/apidocs/llms.txt?endpoint=UpdateCollectionVectors): Marks every document vector in the collection as outdated so the indexing job rebuilds embeddings. - [Export Collection RAG Transactions](http://inference.aivax.net/apidocs/llms.txt?endpoint=ExportCollectionRAGTransactions): Streams the selected RAG transaction view as JSONL. - [View Collection RAG Transaction](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewCollectionRAGTransaction): Returns one RAG transaction with cost, timing, scores, reranker metadata, and matched documents. - [List Collection RAG Transactions](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListCollectionRAGTransactions): Lists recent, low-quality, or high-quality RAG transactions for one collection within the account plan retention period. - [Reset Collection](http://inference.aivax.net/apidocs/llms.txt?endpoint=ResetCollection): Deletes all documents from a collection while keeping the collection record and configuration. - [Export Collection](http://inference.aivax.net/apidocs/llms.txt?endpoint=ExportCollection): Streams all documents in the collection as JSONL lines containing docid, text, reference and tags. - [Edit Collection](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditCollection): Updates a collection. The request may contain only the fields to change; omitted fields keep their current values. Send null for optional fi... - [Get Collection Details](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetCollectionDetails): Returns a collection's name, creation time, indexing state, discovered tags, contextual tags and RAG usage statistics. - [Delete Collection](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteCollection): Deletes the collection and its stored documents for the authenticated account. - [List Collections](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListCollections): Lists RAG collections owned by the authenticated account with document counts by indexing state. - [Create Collection](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateCollection): Creates an empty RAG collection that can receive documents and later be searched or attached to an AI Gateway. ## Documents - [Get Document](http://inference.aivax.net/apidocs/llms.txt?endpoint=GetDocument): Returns the full stored document, including contents, metadata, tags, timestamps and indexing state. - [Delete Document](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteDocument): Deletes one document from a collection. When include-references is present, deletes documents in the same collection that share the selected... - [Browse Documents](http://inference.aivax.net/apidocs/llms.txt?endpoint=BrowseDocuments): Lists up to 400 documents from a collection, optionally filtered by text, tag, reference, content, name or id. - [Index Documents (JSONL)](http://inference.aivax.net/apidocs/llms.txt?endpoint=IndexDocumentsJSONL): Uploads documents from a JSONL file. Existing documents are matched by docId: changed text queues reindexing; changed text with non-null __m... - [Create or Update Document](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateorUpdateDocument): Creates or updates a single document matched by name. name is required to identify the document. Existing documents may be updated with a pa... ## Batch - [Retry Batch Job Items](http://inference.aivax.net/apidocs/llms.txt?endpoint=RetryBatchJobItems): Moves selected failed or low-confidence non-running items back to pending and starts the job when at least one item is retried. - [View Batch Job Item](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewBatchJobItem): Returns one batch item's input, output, state, validation result, confidence and cost. - [Edit Batch Job Item](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditBatchJobItem): Changes a non-running batch item's priority or removes it when action is cancel. - [Delete Batch Job Item](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteBatchJobItem): Deletes one non-running batch item from a job. - [Import Batch Job Items](http://inference.aivax.net/apidocs/llms.txt?endpoint=ImportBatchJobItems): Imports batch job items from multipart/form-data. Mode 'lines' imports each non-empty file line; mode 'files' imports each uploaded text fil... - [List Batch Job Items](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListBatchJobItems): Lists batch job items with optional state, confidence and text filters. Results are capped to protect large jobs. - [Remove Batch Job Items](http://inference.aivax.net/apidocs/llms.txt?endpoint=RemoveBatchJobItems): Bulk-deletes non-running items from a job by mode: pending, finished, errors or all. - [Export Batch Job](http://inference.aivax.net/apidocs/llms.txt?endpoint=ExportBatchJob): Streams processed batch job items as JSONL, including metadata, input and output. Pending items are not exported. - [View Batch Job](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewBatchJob): Returns a batch job, aggregated summary, progression timeline and latest processed items. - [Edit Batch Job](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditBatchJob): Changes a job title or state. Setting state to Active starts processing; Paused or Finished stops further processing. - [Delete Batch Job](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteBatchJob): Deletes a batch job and every item under it. - [List Batch Jobs](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListBatchJobs): Lists jobs under a batch workflow with aggregated processing summary. - [Create Batch Job](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateBatchJob): Creates a paused job under a batch workflow. Import items, then start the job through the edit endpoint. - [View Batch Workflow](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewBatchWorkflow): Returns a batch workflow's title, timestamps and full processing options. - [Edit Batch Workflow](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditBatchWorkflow): Edits a batch workflow. The request may contain only the fields to change; omitted fields keep their current values. enabledTools is shallow... - [Delete Batch Workflow](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteBatchWorkflow): Deletes a batch workflow and all jobs and items created under it. - [List Batch Workflows](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListBatchWorkflows): Lists batch workflows owned by the authenticated account, including model, validation status and job count. - [Create Batch Workflow](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateBatchWorkflow): Creates a reusable batch workflow that defines item instructions, result schema, model, validation and retry behavior. - [List Account Batch Jobs](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListAccountBatchJobs): Lists recent batch jobs across all workflows owned by the authenticated account. ## Authentication - [Login](http://inference.aivax.net/apidocs/llms.txt?endpoint=Login): Authenticates an account using its login key and returns a 180-day access token along with basic account information. Rate limited to 10 att... ## AI Gateways - [Edit AI Gateway](http://inference.aivax.net/apidocs/llms.txt?endpoint=EditAIGateway): Edits an existing AI gateway. The request may contain only the fields to change; omitted fields keep their current values. When parameters i... - [View AI Gateway](http://inference.aivax.net/apidocs/llms.txt?endpoint=ViewAIGateway): Returns the gateway name, slug and full runtime parameters for one AI Gateway. - [Delete AI Gateway](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteAIGateway): Deletes one AI Gateway owned by the authenticated account. - [Create AI Gateway](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateAIGateway): Creates an AI Gateway runtime under the authenticated account. The gateway can combine model routing, instructions, RAG collections, built-i... - [List AI Gateways](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListAIGateways): Lists AI Gateways owned by the authenticated account with model, inference endpoint and slug summary fields. ## Accounts - [Roll Salt](http://inference.aivax.net/apidocs/llms.txt?endpoint=RollSalt): Generates a new random salt for the account, invalidating webhooks and integrations that depend on the previous salt. - [Update Account](http://inference.aivax.net/apidocs/llms.txt?endpoint=UpdateAccount): Replaces the authenticated account display name and account-level parameters. ## API Keys - [Delete API Key](http://inference.aivax.net/apidocs/llms.txt?endpoint=DeleteAPIKey): Revokes an API key owned by the authenticated account. - [List API Keys](http://inference.aivax.net/apidocs/llms.txt?endpoint=ListAPIKeys): Returns masked private and public API keys for the authenticated account, plus the current account salt used by webhook and integration sign... - [Create API Key](http://inference.aivax.net/apidocs/llms.txt?endpoint=CreateAPIKey): Creates a private or public API key for the authenticated account and returns the grant key once.