Chat Completions API

OpenAI-compatible chat completions with built-in reasoning, tool use, and streaming.

POSThttps://api.apodex.ai/v1/chat/completions

POST/v1/chat/completions

https://api.apodex.ai

Overview

The Chat Completions API is the primary interface for interacting with Apodex models. It follows the OpenAI Chat Completions format, so you can use the OpenAI SDK or any compatible client. The API is served by the frontier gateway.

Base URL

https://api.apodex.ai

Auth

Bearer API key

Format

OpenAI-compatible

This page is the contract for the deep research tiers. The core models apodex-1.1 and apodex-1.1-mini share this same /v1/chat/completions path, but their semantics differ deliberately — stream default, failure surface, sampling parameters and billing unit all change. For the core models, see OpenAI-compatible API.

Models

Pass the model id below in the model parameter when creating a chat completion.

Models	When to use	Context window	Max completion	Pricing
Deep Research `apodex-1-1-deep-research`	When the answer exists and needs to be found fast.	128k	64k	$5.00 / $20.00
Deep Solve `apodex-1-1-deep-solve`	When the answer requires inference, judgment, or tradeoff analysis.	128k	64k	$5.00 / $25.00
Deep DiscoverPreview `apodex-1-1-deep-discover`	When the problem is hard, high-stakes, novel, or easy to get wrong.	128k	256k	$10.00 / $100.00

Input / Output / 1M tokens

Apodex 1.0 models and GET /v1/models

Models	When to use	Context window	Max completion	Pricing
Deep Research `apodex-1-0-deep-research`	When the answer exists and needs to be found fast.	256k	16k	$10.00 / $40.00
Deep Solve `apodex-1-0-deep-solve`	When the answer requires inference, judgment, or tradeoff analysis.	256k	16k	$10.00 / $50.00
Deep DiscoverPreview `apodex-1-0-deep-discover`	When the problem is hard, high-stakes, novel, or easy to get wrong.	128k	256k	$10.00 / $100.00

List via API

The same data is available programmatically — useful for clients that want to enumerate or check model capabilities at runtime.

Example

curl https://api.apodex.ai/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

Response

{
  "object": "list",
  "data": [
    {
      "id": "apodex-1-1-deep-research",
      "object": "model",
      "created": 1700000000,
      "owned_by": "apodex",
      "context_length": 131072,
      "max_completion_tokens": 65536
    },
    {
      "id": "apodex-1-1-deep-solve",
      "object": "model",
      "created": 1700000000,
      "owned_by": "apodex",
      "context_length": 131072,
      "max_completion_tokens": 65536
    },
    {
      "id": "apodex-1-1-deep-discover",
      "object": "model",
      "created": 1700000000,
      "owned_by": "apodex",
      "context_length": 131072,
      "max_completion_tokens": 262144
    }
  ]
}

Create Chat Completion

Send a conversation to the model and receive a completion. Supports streaming (SSE) and non-streaming modes.

Headers

Header	Value	Description
`Authorization` Required	`Bearer YOUR_API_KEY`	Your Apodex API key
`Content-Type` Required	`application/json`	Must be application/json

Body parameters

Parameter	Description
`model` stringRequired	Model ID to use, e.g. "apodex-1-1-deep-research"
`messages` arrayRequired	Array of message objects with "role" (system/user/assistant) and "content" fields
`stream` booleandefault true	Whether to stream the response via SSE. Defaults to true — note this differs from OpenAI's default of false. Always pass stream explicitly when using OpenAI SDKs.
`max_tokens` integer	Maximum number of tokens to generate in the completion
`mcp_servers` array	Array of MCP server configs ({name, url, headers?, access_token?, oauth?}) for external tool access

model

stringRequired

Model ID to use, e.g. "apodex-1-1-deep-research"

messages

arrayRequired

Array of message objects with "role" (system/user/assistant) and "content" fields

stream

booleandefault true

Whether to stream the response via SSE. Defaults to true — note this differs from OpenAI's default of false. Always pass stream explicitly when using OpenAI SDKs.

max_tokens

integer

Maximum number of tokens to generate in the completion

mcp_servers

array

Array of MCP server configs ({name, url, headers?, access_token?, oauth?}) for external tool access

Request

curl -X POST https://api.apodex.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "apodex-1-1-deep-research",
    "messages": [
      {
        "role": "user",
        "content": "What are the latest trends in AI?"
      }
    ],
    "stream": true
  }'

Streaming Response

When stream: true (the default), the response is delivered as Server-Sent Events (SSE). Each line is prefixed with data: followed by a JSON chunk. The stream ends with data: [DONE]. The stream has two phases:

Phase 1: Reasoning

The model emits reasoning steps via delta.reasoning_steps. Each step has a type (e.g. thinking, web_search) and a content field with the step details.

Phase 2: Final Answer

After reasoning completes, the final answer streams via delta.content, token by token, just like standard OpenAI streaming.

SSE Stream Example

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"role":"assistant"}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"reasoning_steps":[{"type":"thinking","thought":"Let me analyze this question..."}]}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"reasoning_steps":[{"type":"web_search","web_search":{"search_keywords":["latest AI trends 2026"]}}]}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":"Here are"}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":" the latest"}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":{"prompt_tokens":12,"completion_tokens":256,"total_tokens":268,"completion_tokens_details":{"reasoning_tokens":45},"num_search_queries":1}}

data: [DONE]

Non-Streaming Response

Set stream: false to receive the full response as a single JSON object. The response waits until the model finishes generating before returning.

Response

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1712345678,
  "model": "apodex-1-1-deep-research",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here are the latest trends in AI...",
        "reasoning_steps": [
          {
            "type": "thinking",
            "thought": "Let me analyze this question..."
          },
          {
            "type": "web_search",
            "web_search": {
              "search_keywords": ["latest AI trends 2026"],
              "search_results": [{"title": "...", "url": "...", "snippet": "..."}]
            }
          }
        ]
      },
      "finish_reason": "stop"
    }
  ],
  "search_results": [{"title": "...", "url": "...", "snippet": "..."}],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 256,
    "total_tokens": 268,
    "completion_tokens_details": {
      "reasoning_tokens": 45
    },
    "num_search_queries": 1
  }
}

Finish Reason & Usage

The final chunk (streaming) or the response (non-streaming) includes a finish_reason and a usage object.

finish_reason	Description
`stop`	The model completed normally
`error`	The workflow failed due to an internal error. An `error` object is included in the chunk.
`cancelled`	The request was cancelled (client disconnect or explicit cancel)

usage field	Description
`prompt_tokens`	Tokens in the input prompt
`completion_tokens`	Tokens generated in the completion
`total_tokens`	Sum of prompt + completion tokens
`completion_tokens_details.reasoning_tokens`	Hidden reasoning tokens used by the model to produce the answer. Counted as part of `completion_tokens` and billed as output. OpenAI-compatible nested shape.
`num_search_queries`	Number of billed `fetch_url_content` invocations during reasoning. Omitted when zero.

Reasoning Step Types

During the reasoning phase, the model can emit the following step types:

Type	Description
`thinking`	Internal reasoning and analysis
`web_search`	Searching the web for information
`fetch_url_content`	Fetching and reading content from a URL
`execute_python`	Executing Python code in a sandboxed environment
`execute_command`	Running a shell command
`tool_call`	Calling an MCP tool or external function

Capabilities & Limits

Capability	Status
Context window	256k tokens per request, shared between input and output.
External tool use (MCP)	Supported via the `mcp_servers` request parameter; currently in private beta. Contact us for access.
Custom function calling	OpenAI-style `tools` / `tool_choice` are not supported. Use `mcp_servers` for external tools.
Structured output	`response_format` / `json_schema` are not supported.
Prompt caching	`cache_control` is not supported.
Multimodal input	Image and document inputs are supported on the platform but not yet exposed via the public API. Coming soon.

OpenAI SDK Compatibility

Point any OpenAI SDK at the Apodex base URL and use your Apodex API key. The API is fully compatible with the OpenAI Chat Completions format, with one behavioral difference: stream defaults to true. Always pass stream explicitly — use stream=False for a plain JSON response; omitting it will make the SDK's non-streaming call receive SSE and fail to parse.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.apodex.ai/v1",
    api_key="YOUR_API_KEY",
)

stream = client.chat.completions.create(
    model="apodex-1-1-deep-research",
    messages=[
        {"role": "user", "content": "What are the latest trends in AI?"}
    ],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta
    if delta.content:
        print(delta.content, end="", flush=True)

Error Responses

The API returns standard HTTP error codes with a JSON error body.

Status	Code	Description
`400`	`bad_request`	Invalid request body or missing required fields
`401`	`unauthorized`	Missing or invalid API key
`402`	`insufficient_balance`	Account balance is too low to process the request
`429`	`rate_limited`	Too many requests. Retry after the Retry-After header value.
`503`	`service_unavailable`	The service is temporarily overloaded or down for maintenance

Error Response Format

{
  "error": {
    "code": "unauthorized",
    "message": "Invalid API key provided.",
    "type": "authentication_error"
  }
}