Chat / LLM

FOTOhub provides two chat completion endpoints supporting multiple LLM providers. The OpenAI-compatible endpoint follows the standard OpenAI chat completions format, making it a drop-in replacement for existing integrations. The premium endpoint provides access to premium chat models with token-based billing for precise cost control.

Both endpoints support multi-turn conversations, system prompts, and return detailed billing information in every response. Streaming via Server-Sent Events is available on the OpenAI-compatible endpoint.

OpenAI SDK Compatible

The /v1/ai/chat/completions endpoint is fully compatible with the OpenAI Python and TypeScript SDKs. Simply change the base URL to https://apis.fotohub.app/v1/ai and use your FOTOhub API key. See the OpenAI Compatibility section below for drop-in usage examples.

OpenAI-Compatible Chat Completions

POST /v1/ai/chat/completions

Auth: API Key (Bearer token) | Billing: per token

Standard chat completions endpoint compatible with OpenAI SDKs. Supports Google Gemini, OpenAI GPT, and Anthropic Claude models.

Parameters

Parameter	Type	Required	Default	Description
`model`	string	No	`gemini-flash`	Model ID to use for completion. See pricing table below for available models.
`messages`	array	Yes	--	Array of message objects. Each message has a `role` (system, user, or assistant) and `content` (string or array for vision). System messages set behavior, user messages are inputs, assistant messages are for multi-turn context.
`temperature`	number	No	--	Accepted but ignored. See the compatibility note below.
`max_tokens`	integer	No	--	Accepted but ignored. See the compatibility note below.
`stream`	boolean	No	--	Accepted but ignored. This endpoint does not stream; see Streaming.

Only model and messages are read

This endpoint exists for drop-in OpenAI SDK compatibility, so it accepts the full OpenAI request body without erroring — but it only acts on model and messages. temperature, max_tokens, stream, top_p, frequency_penalty and presence_penalty are discarded silently: sending them is harmless, but they will not change the output. If you need sampling control or streaming, use the premium endpoint (/v1/ai/chat/claude) or the agent endpoint (/v1/ai/agent/stream) instead.

Pricing Table

Billed on the tokens you actually use, in credits, at the per-1M rates below. The charge is fractional: a short reply costs a fraction of one credit rather than rounding up to a whole one. When your included credits are exhausted the request is billed from your USD wallet instead.

Model	ID	Credits/1M in	Credits/1M out	Best for
Gemini Flash	`gemini-flash`	20	166.7	Fast responses, bulk tasks
Gemini Pro	`gemini-pro`	83	667	Balanced quality, general use
GPT-4o	`gpt-4o`	200	1000	Multimodal, vision, creative
Claude Sonnet	`claude-sonnet`	200	1000	Code, analysis, reasoning

Output tokens cost 4-8x input tokens, which is why the two columns are quoted separately -- a single blended number would misprice any workload that is not half prompt and half completion. A 25-token prompt with a 150-token answer on gemini-flash costs 25/1e6 * 20 + 150/1e6 * 166.7 = 0.0255 credits.

The exact amount charged comes back on every response in billing.usd_charged, with billing.basis: "tokens" confirming it was derived from real token counts.

Available Models

This endpoint accepts exactly the four IDs in the pricing table above. Any other value returns 400 Unknown chat model listing the supported set -- it is never silently substituted with a default, so a typo fails loudly instead of quietly serving you a different model.

GET /v1/models?category=text returns the same four. The premium endpoint (/v1/ai/chat/claude) serves a different and larger family of models with their own IDs in dot notation (claude-sonnet-4.6); those IDs are not accepted here.

Response

json

{
  "id": "chatcmpl-fh-abc123",
  "object": "chat.completion",
  "created": 1719849600,
  "model": "gemini-flash",
  "usd_charged": 0.0255,
  "billing": {
    "method": "wallet",
    "usd_charged": 0.0255,
    "usd_charged": 0,
    "basis": "tokens"
  },
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The Eiffel Tower is a wrought-iron lattice tower located on the Champ de Mars in Paris, France. It was constructed from 1887 to 1889 as the centerpiece of the 1889 World's Fair."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 150,
    "total_tokens": 175
  }
}

Premium Chat (Token-Based Billing)

POST /v1/ai/chat/claude

Auth: API Key (Bearer token) | Billing: token-based (per-token pricing)

Access to FOTOhub's premium chat models — the Claude-class and Nova-class families — with precise token-based billing and per-token cost breakdowns. Both endpoints bill on real token counts; the difference is that this one prices natively in USD and returns a cost_breakdown, while the OpenAI-compatible endpoint prices in credits.

Parameters

Parameter	Type	Required	Default	Description
`model`	string	No	`claude-sonnet-4.6`	Premium model ID. Must be one of the supported models listed in the pricing table below. Sending an unsupported ID returns `400 Unknown model`.
`messages`	array	Yes	--	Array of message objects with `role` (user or assistant) and `content` (string or array for vision). Note: the system prompt is passed via the separate `system` parameter, not as a message.
`temperature`	number	No	`0.7`	Controls randomness. Range 0-1 for Claude models, 0-2 for Nova models. Lower = more deterministic.
`max_tokens`	integer	No	`4096`	Maximum tokens to generate.
`system`	string	No	--	System prompt that defines the assistant's behavior and persona. Passed separately from messages. Supports multi-paragraph instructions.

Streaming

The premium endpoint returns a single complete response (no SSE streaming). For token-by-token streaming use POST /v1/ai/agent/stream — it serves the same premium models. The OpenAI-compatible endpoint does not stream either, regardless of stream: true.

Premium Models (Token-Based Pricing)

All models below are billed natively in USD, per token. The figures below are the base per-1M-token USD rates; the amount charged is computed from your actual input/output token counts plus the platform margin and returned as cost_usd. Because the calculation never leaves USD it does not move with the PLN exchange rate, and this endpoint's billing object carries no PLN field at all.

Claude

Model	ID	Input ($/1M)	Output ($/1M)	Features
Claude Sonnet 4.6	`claude-sonnet-4.6`	$3.00	$15.00	Vision, 200K ctx, tool-use (default)
Claude Sonnet 4.5	`claude-sonnet-4.5`	$3.00	$15.00	Vision, 200K ctx
Claude Sonnet 4	`claude-sonnet-4`	$3.00	$15.00	Vision, 200K ctx
Claude Haiku 4.5	`claude-haiku-4.5`	$0.80	$4.00	Vision, 200K ctx, fastest

Nova

Model	ID	Input ($/1M)	Output ($/1M)	Features
Nova Premier	`nova-premier`	$2.50	$10.00	Best Nova, complex tasks
Nova Pro	`nova-pro`	$0.80	$3.20	Vision, reasoning, balanced
Nova Lite	`nova-lite`	$0.06	$0.24	Vision, fast, budget
Nova 2 Lite	`nova-2-lite`	$0.04	$0.16	Vision, new generation
Nova Micro	`nova-micro`	$0.035	$0.14	Text-only, cheapest

Model IDs use dot notation

Premium model IDs use dots (claude-sonnet-4.6), unlike the credit-based endpoint which uses dashes (claude-sonnet-4-6). Use the exact ID from the table above.

Token-Based Billing

Premium models use token-based billing, meaning costs are calculated per-token after generation completes. The billing.cost_breakdown field in the response provides exact input/output token counts and the resulting USD charge in cost_usd. A minimum of 1 credit-equivalent is deducted per request.

usd_charged is what actually left your wallet: it is 0 while the request is covered by included credits (method: "credits") and equals the operation price once credits run out (method: "wallet"). cost_breakdown.cost_usd is the token cost either way — read that one to attribute spend per request.

Response

json

{
  "id": "chatcmpl-fh-xyz789",
  "object": "chat.completion",
  "created": 1719849600,
  "model": "claude-sonnet-4.6",
  "billing": {
    "method": "wallet",
    "usd_charged": 2,
    "usd_charged": 0,
    "cost_breakdown": {
      "input_tokens": 150,
      "output_tokens": 500,
      "total_tokens": 650,
      "cost_usd": 0.011925
    }
  },
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here is a detailed analysis of the architectural patterns used in your codebase..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "input_tokens": 150,
    "output_tokens": 500,
    "total_tokens": 650
  }
}

Streaming

The chat endpoints do not stream

Neither /v1/ai/chat/completions nor /v1/ai/chat/claude streams. Both return one complete JSON body. stream: true on the OpenAI-compatible endpoint is accepted and then ignored — you will receive an ordinary chat.completion response, not an SSE stream, and there is no chat.completion.chunk object anywhere in this API.

The only streaming endpoint is POST /v1/ai/agent/stream, documented below.

POST /v1/ai/agent/stream

The streaming twin of POST /v1/ai/agent. Same request contract, same billing, same single-turn semantics — the caller still drives each round of tool use. The difference is that the assistant turn arrives as SSE frames instead of one JSON body, so a CLI can render text as it is produced.

Models (a subset of the premium catalog — only models supporting tool-use): claude-sonnet-4.6 (default), claude-sonnet-4.5, claude-sonnet-4, claude-haiku-4.5. An unlisted ID returns 400.

Parameter	Type	Required	Description
`messages`	array	Yes	Conversation so far. Empty or missing returns `400`.
`model`	string	No	One of the four IDs above. Default `claude-sonnet-4.6`.
`tools`	array	No	Tool definitions. Must be a list if present, else `400`.

Authentication is checked before the stream opens, so an invalid key is a plain 401 rather than an error frame you have to parse.

Frame Format

Each frame is data: followed by one JSON object, separated by two newlines. The stream terminates with data: [DONE]. Note this is not the OpenAI chunk format — frames are discriminated by a type field, not by choices[].delta.

`type`	Payload
`text_delta`	`text` — the next fragment of assistant text
`tool_use`	`id`, `name`, `input` — emitted once the tool call is fully accumulated, never partially
`done`	`stop_reason`, `usage`, `billing` — always the last frame before `[DONE]`
`error`	`message` — terminal; `[DONE]` still follows

data: {"type":"text_delta","text":"Let me check"}

data: {"type":"text_delta","text":" the weather."}

data: {"type":"tool_use","id":"toolu_01A","name":"get_weather","input":{"city":"Warsaw"}}

data: {"type":"done","stop_reason":"tool_use","usage":{"input_tokens":150,"output_tokens":48,"total_tokens":198},"billing":{"method":"credits","usd_charged":2,"usd_charged":0}}

data: [DONE]

Disconnecting early does not avoid the charge

Billing runs exactly once, on completion, from the accumulated token usage. If your client disconnects mid-stream the turn is still billed for what the model produced — those tokens were generated and charged upstream either way. This is the same behaviour as the blocking route.

Streaming Example -- SSE Parsing

PythonTypeScriptGocURL

python

import requests
import json

url = "https://apis.fotohub.app/v1/ai/agent/stream"
headers = {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json"
}
payload = {
    "model": "claude-sonnet-4.6",
    "messages": [{"role": "user", "content": "Explain quantum computing"}],
}

response = requests.post(url, json=payload, headers=headers, stream=True)

full_content = ""
for line in response.iter_lines():
    if not line:
        continue
    line = line.decode("utf-8")
    if not line.startswith("data: "):
        continue
    data = line[6:]  # Remove "data: " prefix
    if data == "[DONE]":
        break

    frame = json.loads(data)
    if frame["type"] == "text_delta":
        full_content += frame["text"]
        print(frame["text"], end="", flush=True)
    elif frame["type"] == "tool_use":
        print(f"\n[tool] {frame['name']}({frame['input']})")
    elif frame["type"] == "done":
        print(f"\nstop_reason={frame['stop_reason']} "
              f"credits={frame['billing']['usd_charged']} "
              f"usd={frame['billing']['usd_charged']}")
    elif frame["type"] == "error":
        raise RuntimeError(frame["message"])

typescript

const response = await fetch("https://apis.fotohub.app/v1/ai/agent/stream", {
  method: "POST",
  headers: {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "claude-sonnet-4.6",
    messages: [{ role: "user", content: "Explain quantum computing" }],
  }),
});

const reader = response.body!.getReader();
const decoder = new TextDecoder();
let fullContent = "";
// Frames are split by a blank line, and a single read() can end mid-frame --
// buffer until a separator is seen rather than parsing each chunk directly.
let buffer = "";

outer: while (true) {
  const { done, value } = await reader.read();
  if (done) break;

  buffer += decoder.decode(value, { stream: true });
  const frames = buffer.split("\n\n");
  buffer = frames.pop() ?? "";

  for (const raw of frames) {
    if (!raw.startsWith("data: ")) continue;
    const data = raw.slice(6).trim();
    if (data === "[DONE]") break outer;

    const frame = JSON.parse(data);
    if (frame.type === "text_delta") {
      fullContent += frame.text;
      process.stdout.write(frame.text);
    } else if (frame.type === "tool_use") {
      console.log(`\n[tool] ${frame.name}`, frame.input);
    } else if (frame.type === "done") {
      console.log(`\nstop_reason=${frame.stop_reason}`, frame.billing);
    } else if (frame.type === "error") {
      throw new Error(frame.message);
    }
  }
}

package main

import (
	"bufio"
	"bytes"
	"encoding/json"
	"fmt"
	"net/http"
	"strings"
)

func main() {
	payload := map[string]interface{}{
		"model":    "claude-sonnet-4.6",
		"messages": []map[string]string{{"role": "user", "content": "Explain quantum computing"}},
	}
	body, _ := json.Marshal(payload)

	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/agent/stream", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	scanner := bufio.NewScanner(resp.Body)
	var fullContent string

	for scanner.Scan() {
		line := scanner.Text()
		if !strings.HasPrefix(line, "data: ") {
			continue
		}
		data := strings.TrimPrefix(line, "data: ")
		if data == "[DONE]" {
			break
		}

		var frame map[string]interface{}
		if err := json.Unmarshal([]byte(data), &frame); err != nil {
			continue
		}

		switch frame["type"] {
		case "text_delta":
			text, _ := frame["text"].(string)
			fullContent += text
			fmt.Print(text)
		case "tool_use":
			fmt.Printf("\n[tool] %v %v\n", frame["name"], frame["input"])
		case "done":
			fmt.Printf("\nstop_reason=%v billing=%v\n", frame["stop_reason"], frame["billing"])
		case "error":
			panic(frame["message"])
		}
	}
	fmt.Println()
	fmt.Printf("Total: %s\n", fullContent)
}

bash

curl -X POST "https://apis.fotohub.app/v1/ai/agent/stream" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -N \
  -d '{
    "model": "claude-sonnet-4.6",
    "messages": [
      {"role": "user", "content": "Explain quantum computing"}
    ]
  }'

Token Counting & Billing

Tokens are the fundamental unit of text processing for LLMs. A token is approximately 4 characters or 0.75 words in English. Both input (prompt) and output (completion) tokens are counted toward billing.

Token-Based USD Wallet Billing

Per-token pricing in USD (input and output rated separately at 1:1 provider cost)
Billed directly from your prepaid USD wallet balance
Billed after the completion, from the token counts the model reports, because they are not knowable before the call
A pre-flight check validates that your wallet has available funds before dispatching to the provider
Nothing is charged if the provider errors, so there is no refund to wait for
Exact cost and remaining balance are returned in the billing object

Estimating Costs

A typical conversational exchange (100-word prompt, 200-word response) uses approximately 75 input tokens + 150 output tokens = 225 total tokens. On gemini-flash that costs ~$0.00003; on claude-sonnet-4.6 it works out to roughly $0.0037.

Code Examples

Basic Chat Completion

PythonTypeScriptGocURL

python

import requests

url = "https://apis.fotohub.app/v1/ai/chat/completions"
headers = {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json"
}

response = requests.post(url, json={
    "model": "gemini-flash",
    "messages": [
        {"role": "user", "content": "What is the capital of France?"}
    ],
    "temperature": 0.3,
    "max_tokens": 256
}, headers=headers)

data = response.json()
print(data["choices"][0]["message"]["content"])
print(f"Tokens used: {data['usage']['total_tokens']}")
print(f"Credits used: {data['usd_charged']}")

typescript

const response = await fetch("https://apis.fotohub.app/v1/ai/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "gemini-flash",
    messages: [
      { role: "user", content: "What is the capital of France?" }
    ],
    temperature: 0.3,
    max_tokens: 256,
  }),
});

const data = await response.json();
console.log(data.choices[0].message.content);
console.log(`Tokens used: ${data.usage.total_tokens}`);
console.log(`Credits used: ${data.usd_charged}`);

package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
)

func main() {
	payload := map[string]interface{}{
		"model":       "gemini-flash",
		"messages":    []map[string]string{{"role": "user", "content": "What is the capital of France?"}},
		"temperature": 0.3,
		"max_tokens":  256,
	}
	body, _ := json.Marshal(payload)

	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/chat/completions", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	respBody, _ := io.ReadAll(resp.Body)
	var data map[string]interface{}
	json.Unmarshal(respBody, &data)

	choices := data["choices"].([]interface{})
	message := choices[0].(map[string]interface{})["message"].(map[string]interface{})
	fmt.Println(message["content"])

	usage := data["usage"].(map[string]interface{})
	fmt.Printf("Tokens used: %.0f\n", usage["total_tokens"])
	if b, ok := data["billing"].(map[string]interface{}); ok {
		fmt.Printf("Cost: $%.6f\n", b["cost_usd"])
	}
}

bash

curl -X POST "https://apis.fotohub.app/v1/ai/chat/completions" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-flash",
    "messages": [
      {"role": "user", "content": "What is the capital of France?"}
    ],
    "temperature": 0.3,
    "max_tokens": 256
  }'

Multi-Turn Conversation

Build multi-turn conversations by including the full message history in each request. The model uses previous messages as context to maintain coherent dialogue.

PythonTypeScriptGocURL

python

import requests

url = "https://apis.fotohub.app/v1/ai/chat/completions"
headers = {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json"
}

# Build conversation history
messages = [
    {"role": "system", "content": "You are a helpful coding assistant. Be concise."},
    {"role": "user", "content": "How do I read a file in Python?"},
    {"role": "assistant", "content": "Use open() with a context manager:\n\nwith open('file.txt', 'r') as f:\n    content = f.read()"},
    {"role": "user", "content": "What about reading it line by line?"}
]

response = requests.post(url, json={
    "model": "gemini-pro",
    "messages": messages,
    "temperature": 0.5,
    "max_tokens": 512
}, headers=headers)

data = response.json()
print(data["choices"][0]["message"]["content"])

# Append assistant response for next turn
messages.append(data["choices"][0]["message"])

# Continue the conversation
messages.append({"role": "user", "content": "Now show me how to write to a file"})
response = requests.post(url, json={
    "model": "gemini-pro",
    "messages": messages,
    "temperature": 0.5,
    "max_tokens": 512
}, headers=headers)

data = response.json()
print(data["choices"][0]["message"]["content"])

typescript

interface Message {
  role: "system" | "user" | "assistant";
  content: string;
}

const url = "https://apis.fotohub.app/v1/ai/chat/completions";
const headers = {
  "Authorization": "Bearer fh_live_your_api_key",
  "Content-Type": "application/json",
};

const messages: Message[] = [
  { role: "system", content: "You are a helpful coding assistant. Be concise." },
  { role: "user", content: "How do I read a file in Python?" },
  { role: "assistant", content: "Use open() with a context manager:\n\nwith open('file.txt', 'r') as f:\n    content = f.read()" },
  { role: "user", content: "What about reading it line by line?" },
];

let response = await fetch(url, {
  method: "POST",
  headers,
  body: JSON.stringify({
    model: "gemini-pro",
    messages,
    temperature: 0.5,
    max_tokens: 512,
  }),
});

let data = await response.json();
console.log(data.choices[0].message.content);

// Append for next turn
messages.push(data.choices[0].message);
messages.push({ role: "user", content: "Now show me how to write to a file" });

response = await fetch(url, {
  method: "POST",
  headers,
  body: JSON.stringify({
    model: "gemini-pro",
    messages,
    temperature: 0.5,
    max_tokens: 512,
  }),
});

data = await response.json();
console.log(data.choices[0].message.content);

package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
)

type Message struct {
	Role    string `json:"role"`
	Content string `json:"content"`
}

type ChatRequest struct {
	Model       string    `json:"model"`
	Messages    []Message `json:"messages"`
	Temperature float64   `json:"temperature"`
	MaxTokens   int       `json:"max_tokens"`
}

func chat(messages []Message) (string, error) {
	payload := ChatRequest{
		Model:       "gemini-pro",
		Messages:    messages,
		Temperature: 0.5,
		MaxTokens:   512,
	}
	body, _ := json.Marshal(payload)

	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/chat/completions", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		return "", err
	}
	defer resp.Body.Close()

	respBody, _ := io.ReadAll(resp.Body)
	var data map[string]interface{}
	json.Unmarshal(respBody, &data)

	choices := data["choices"].([]interface{})
	message := choices[0].(map[string]interface{})["message"].(map[string]interface{})
	return message["content"].(string), nil
}

func main() {
	messages := []Message{
		{Role: "system", Content: "You are a helpful coding assistant. Be concise."},
		{Role: "user", Content: "How do I read a file in Python?"},
		{Role: "assistant", Content: "Use open() with a context manager:\n\nwith open('file.txt', 'r') as f:\n    content = f.read()"},
		{Role: "user", Content: "What about reading it line by line?"},
	}

	reply, _ := chat(messages)
	fmt.Println(reply)

	// Continue conversation
	messages = append(messages, Message{Role: "assistant", Content: reply})
	messages = append(messages, Message{Role: "user", Content: "Now show me how to write to a file"})

	reply, _ = chat(messages)
	fmt.Println(reply)
}

bash

curl -X POST "https://apis.fotohub.app/v1/ai/chat/completions" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-pro",
    "messages": [
      {"role": "system", "content": "You are a helpful coding assistant."},
      {"role": "user", "content": "How do I read a file in Python?"},
      {"role": "assistant", "content": "Use open() with a context manager."},
      {"role": "user", "content": "What about reading it line by line?"}
    ],
    "temperature": 0.5,
    "max_tokens": 512
  }'

Premium Usage -- Claude with System Prompt

Use the premium endpoint for token-based billing with precise cost control. The system parameter is passed separately from messages.

PythonTypeScriptGocURL

python

import requests

url = "https://apis.fotohub.app/v1/ai/chat/claude"
headers = {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json"
}

response = requests.post(url, json={
    "model": "claude-sonnet-4.6",
    "system": """You are a senior software architect. Provide detailed,
production-ready advice. Include code examples when relevant.
Consider scalability, maintainability, and security in all recommendations.""",
    "messages": [
        {"role": "user", "content": "Design a rate limiting system for a REST API that handles 10K req/s"}
    ],
    "temperature": 0.4,
    "max_tokens": 4096
}, headers=headers)

data = response.json()
content = data["choices"][0]["message"]["content"]
billing = data["billing"]

print(content)
print(f"\n--- Billing ---")
print(f"Method: {billing['method']}")
print(f"Input tokens: {billing['cost_breakdown']['input_tokens']}")
print(f"Output tokens: {billing['cost_breakdown']['output_tokens']}")
print(f"Token cost: {billing['cost_breakdown']['cost_usd']} USD")
print(f"Wallet charge: {billing['usd_charged']} USD")  # 0 while covered by credits

typescript

const response = await fetch("https://apis.fotohub.app/v1/ai/chat/claude", {
  method: "POST",
  headers: {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "claude-sonnet-4.6",
    system: `You are a senior software architect. Provide detailed,
production-ready advice. Include code examples when relevant.
Consider scalability, maintainability, and security.`,
    messages: [
      { role: "user", content: "Design a rate limiting system for a REST API that handles 10K req/s" }
    ],
    temperature: 0.4,
    max_tokens: 4096,
  }),
});

const data = await response.json();
const content = data.choices[0].message.content;
const billing = data.billing;

console.log(content);
console.log("\n--- Billing ---");
console.log(`Method: ${billing.method}`);
console.log(`Input tokens: ${billing.cost_breakdown.input_tokens}`);
console.log(`Output tokens: ${billing.cost_breakdown.output_tokens}`);
console.log(`Token cost: ${billing.cost_breakdown.cost_usd} USD`);
console.log(`Wallet charge: ${billing.usd_charged} USD`); // 0 while covered by credits

package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
)

func main() {
	payload := map[string]interface{}{
		"model":  "claude-sonnet-4.6",
		"system": "You are a senior software architect. Provide detailed, production-ready advice. Include code examples when relevant. Consider scalability, maintainability, and security.",
		"messages": []map[string]string{
			{"role": "user", "content": "Design a rate limiting system for a REST API that handles 10K req/s"},
		},
		"temperature": 0.4,
		"max_tokens":  4096,
	}
	body, _ := json.Marshal(payload)

	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/chat/claude", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	respBody, _ := io.ReadAll(resp.Body)
	var data map[string]interface{}
	json.Unmarshal(respBody, &data)

	choices := data["choices"].([]interface{})
	message := choices[0].(map[string]interface{})["message"].(map[string]interface{})
	fmt.Println(message["content"])

	billing := data["billing"].(map[string]interface{})
	fmt.Println("\n--- Billing ---")
	fmt.Printf("Method: %s\n", billing["method"])

	breakdown := billing["cost_breakdown"].(map[string]interface{})
	fmt.Printf("Input tokens: %.0f\n", breakdown["input_tokens"])
	fmt.Printf("Output tokens: %.0f\n", breakdown["output_tokens"])
	fmt.Printf("Token cost: %v USD\n", breakdown["cost_usd"])
	fmt.Printf("Wallet charge: %v USD\n", billing["usd_charged"]) // 0 while covered by credits
}

bash

curl -X POST "https://apis.fotohub.app/v1/ai/chat/claude" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4.6",
    "system": "You are a senior software architect. Provide production-ready advice.",
    "messages": [
      {"role": "user", "content": "Design a rate limiting system for a REST API"}
    ],
    "temperature": 0.4,
    "max_tokens": 4096
  }'

Vision -- Analyzing Images

Send images to vision-capable models by passing a content array with both text and image URL objects. Supported on Claude, GPT, Nova Lite/Pro, and Gemini models.

PythonTypeScriptGocURL

python

import requests

url = "https://apis.fotohub.app/v1/ai/chat/completions"
headers = {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json"
}

response = requests.post(url, json={
    "model": "gpt-4o",
    "messages": [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Describe this image in detail. What objects are visible?"},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": "https://example.com/photo.jpg"
                    }
                }
            ]
        }
    ],
    "max_tokens": 1024
}, headers=headers)

data = response.json()
print(data["choices"][0]["message"]["content"])
print(f"Credits used: {data['usd_charged']}")

typescript

const response = await fetch("https://apis.fotohub.app/v1/ai/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "gpt-4o",
    messages: [
      {
        role: "user",
        content: [
          { type: "text", text: "Describe this image in detail. What objects are visible?" },
          {
            type: "image_url",
            image_url: {
              url: "https://example.com/photo.jpg",
            },
          },
        ],
      },
    ],
    max_tokens: 1024,
  }),
});

const data = await response.json();
console.log(data.choices[0].message.content);
console.log(`Credits used: ${data.usd_charged}`);

package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
)

func main() {
	payload := map[string]interface{}{
		"model": "gpt-4o",
		"messages": []map[string]interface{}{
			{
				"role": "user",
				"content": []map[string]interface{}{
					{"type": "text", "text": "Describe this image in detail. What objects are visible?"},
					{
						"type": "image_url",
						"image_url": map[string]string{
							"url": "https://example.com/photo.jpg",
						},
					},
				},
			},
		},
		"max_tokens": 1024,
	}
	body, _ := json.Marshal(payload)

	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/chat/completions", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	respBody, _ := io.ReadAll(resp.Body)
	var data map[string]interface{}
	json.Unmarshal(respBody, &data)

	choices := data["choices"].([]interface{})
	message := choices[0].(map[string]interface{})["message"].(map[string]interface{})
	fmt.Println(message["content"])
	fmt.Printf("Credits used: %.0f\n", data["usd_charged"])
}

bash

curl -X POST "https://apis.fotohub.app/v1/ai/chat/completions" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type": "text", "text": "Describe this image in detail. What objects are visible?"},
          {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}}
        ]
      }
    ],
    "max_tokens": 1024
  }'

Vision-Capable Models

The following models support image input. On /v1/ai/chat/completions: gpt-4o, claude-sonnet, gemini-pro, gemini-flash. On the token-billed /v1/ai/chat/claude endpoint: claude-sonnet-4.6, claude-sonnet-4.5, claude-sonnet-4, claude-haiku-4.5, nova-pro, nova-lite, nova-2-lite.

Budget-Friendly with Nova

PythonTypeScriptGocURL

python

import requests

url = "https://apis.fotohub.app/v1/ai/chat/claude"
headers = {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json"
}

# Nova Micro — ultra-budget option for simple tasks
# With margin: $0.0525 per 1M input tokens, $0.21 per 1M output tokens
response = requests.post(url, json={
    "model": "nova-micro",
    "messages": [
        {"role": "user", "content": "Classify this review as positive/negative: 'Great product, fast shipping!'"}
    ],
    "temperature": 0.1,  # Low temp for classification
    "max_tokens": 10
}, headers=headers)

data = response.json()
print(data["choices"][0]["message"]["content"])  # "positive"
print(f"Cost: {data['billing']['cost_breakdown']['cost_usd']} USD")  # ~$0.000002

typescript

// Nova Micro — ultra-budget: $0.0525 per 1M input tokens, $0.21 per 1M output (incl. margin)
const response = await fetch("https://apis.fotohub.app/v1/ai/chat/claude", {
  method: "POST",
  headers: {
    "Authorization": "Bearer fh_live_your_api_key",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "nova-micro",
    messages: [
      { role: "user", content: "Classify this review as positive/negative: 'Great product, fast shipping!'" }
    ],
    temperature: 0.1,
    max_tokens: 10,
  }),
});

const data = await response.json();
console.log(data.choices[0].message.content); // "positive"
console.log(`Cost: ${data.billing.cost_breakdown.cost_usd} USD`);

package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
)

func main() {
	// Nova Micro — ultra-budget for classification, extraction, simple tasks
	payload := map[string]interface{}{
		"model": "nova-micro",
		"messages": []map[string]string{
			{"role": "user", "content": "Classify this review as positive/negative: 'Great product, fast shipping!'"},
		},
		"temperature": 0.1,
		"max_tokens":  10,
	}
	body, _ := json.Marshal(payload)

	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/chat/claude", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	respBody, _ := io.ReadAll(resp.Body)
	var data map[string]interface{}
	json.Unmarshal(respBody, &data)

	choices := data["choices"].([]interface{})
	message := choices[0].(map[string]interface{})["message"].(map[string]interface{})
	fmt.Println(message["content"]) // "positive"

	billing := data["billing"].(map[string]interface{})
	breakdown := billing["cost_breakdown"].(map[string]interface{})
	fmt.Printf("Cost: %v USD\n", breakdown["cost_usd"])
}

bash

curl -X POST "https://apis.fotohub.app/v1/ai/chat/claude" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nova-micro",
    "messages": [
      {"role": "user", "content": "Classify: positive or negative? Great product!"}
    ],
    "temperature": 0.1,
    "max_tokens": 10
  }'

OpenAI SDK Compatibility

The /v1/ai/chat/completions endpoint is a drop-in replacement for the OpenAI API. Use the official OpenAI SDK with your FOTOhub API key and base URL -- no code changes required beyond configuration.

PythonTypeScriptGocURL

python

from openai import OpenAI

# Point the OpenAI client at FOTOhub
client = OpenAI(
    api_key="fh_live_your_api_key",
    base_url="https://apis.fotohub.app/v1/ai"
)

# Use exactly like the OpenAI API
response = client.chat.completions.create(
    model="gemini-flash",          # Any FOTOhub model ID
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain REST APIs in 3 sentences."}
    ],
    temperature=0.7,
    max_tokens=256,
    stream=False
)

print(response.choices[0].message.content)
print(f"Model: {response.model}")
print(f"Tokens: {response.usage.total_tokens}")

typescript

import OpenAI from "openai";

// Point the OpenAI client at FOTOhub
const client = new OpenAI({
  apiKey: "fh_live_your_api_key",
  baseURL: "https://apis.fotohub.app/v1/ai",
});

// Use exactly like the OpenAI API
const response = await client.chat.completions.create({
  model: "gemini-flash",          // Any FOTOhub model ID
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "Explain REST APIs in 3 sentences." },
  ],
  temperature: 0.7,
  max_tokens: 256,
  stream: false,
});

console.log(response.choices[0].message.content);
console.log(`Model: ${response.model}`);
console.log(`Tokens: ${response.usage?.total_tokens}`);

package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
)

// FOTOhub is OpenAI-compatible — use any HTTP client with the same payload format.
// For Go, use github.com/sashabaranov/go-openai or plain HTTP:
func main() {
	payload := map[string]interface{}{
		"model": "gemini-flash",
		"messages": []map[string]string{
			{"role": "system", "content": "You are a helpful assistant."},
			{"role": "user", "content": "Explain REST APIs in 3 sentences."},
		},
		"temperature": 0.7,
		"max_tokens":  256,
	}
	body, _ := json.Marshal(payload)

	// Same endpoint as OpenAI — just change the base URL
	req, _ := http.NewRequest("POST", "https://apis.fotohub.app/v1/ai/chat/completions", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer fh_live_your_api_key")
	req.Header.Set("Content-Type", "application/json")

	resp, _ := http.DefaultClient.Do(req)
	defer resp.Body.Close()

	respBody, _ := io.ReadAll(resp.Body)
	var data map[string]interface{}
	json.Unmarshal(respBody, &data)

	choices := data["choices"].([]interface{})
	message := choices[0].(map[string]interface{})["message"].(map[string]interface{})
	fmt.Println(message["content"])
}

bash

# Same format as OpenAI — just change the URL and API key
curl -X POST "https://apis.fotohub.app/v1/ai/chat/completions" \
  -H "Authorization: Bearer fh_live_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-flash",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Explain REST APIs in 3 sentences."}
    ],
    "temperature": 0.7,
    "max_tokens": 256
  }'

Migration from OpenAI

To migrate existing OpenAI integrations to FOTOhub:

Change base_url / baseURL to https://apis.fotohub.app/v1/ai
Replace your OpenAI API key with your FOTOhub key (fh_live_...)
Optionally change the model parameter to a FOTOhub model ID (or keep gpt-4o -- it works)

All standard OpenAI parameters are accepted — your existing request bodies will not error. Only model and messages change the result, however: temperature, max_tokens, stream, top_p, frequency_penalty and presence_penalty are ignored. If your integration depends on sampling parameters or on streaming, migrate to /v1/ai/chat/claude or /v1/ai/agent/stream rather than the compatibility shim.

Error Responses

400 -- Invalid Request

json

{
  "error": {
    "type": "invalid_request",
    "message": "messages array is required and must contain at least one message",
    "code": "missing_messages"
  }
}

402 -- Insufficient Funds

json

{
  "error": {
    "type": "billing_error",
    "message": "Insufficient credits. Required: 2, available: 0",
    "code": "insufficient_funds",
    "credits_required": 2,
    "credits_available": 0
  }
}

429 -- Rate Limited

json

{
  "error": {
    "type": "rate_limit",
    "message": "Rate limit exceeded. Please retry after 2 seconds.",
    "code": "rate_limited",
    "retry_after": 2
  }
}

Rate Limits

Rate Limits by Tier

Free tier: 10 requests/minute, 100 requests/day
Pro tier: 60 requests/minute, 5,000 requests/day
Enterprise: Custom limits -- contact support
Streaming requests count as 1 request regardless of duration

Chat / LLM ​

OpenAI-Compatible Chat Completions ​

Parameters ​

Pricing Table ​

Available Models ​

Response ​

Premium Chat (Token-Based Billing) ​

Parameters ​

Premium Models (Token-Based Pricing) ​

Claude ​

Nova ​

Response ​

Streaming ​

POST /v1/ai/agent/stream ​

Frame Format ​

Streaming Example -- SSE Parsing ​

Token Counting & Billing ​

Token-Based USD Wallet Billing ​

Code Examples ​

Basic Chat Completion ​

Multi-Turn Conversation ​

Premium Usage -- Claude with System Prompt ​

Vision -- Analyzing Images ​

Budget-Friendly with Nova ​

OpenAI SDK Compatibility ​

Error Responses ​

400 -- Invalid Request ​

402 -- Insufficient Funds ​

429 -- Rate Limited ​

Rate Limits ​

Chat / LLM

OpenAI-Compatible Chat Completions

Parameters

Pricing Table

Available Models

Response

Premium Chat (Token-Based Billing)

Parameters

Premium Models (Token-Based Pricing)

Claude

Nova

Response

Streaming

POST /v1/ai/agent/stream

Frame Format

Streaming Example -- SSE Parsing

Token Counting & Billing

Token-Based USD Wallet Billing

Code Examples

Basic Chat Completion

Multi-Turn Conversation

Premium Usage -- Claude with System Prompt

Vision -- Analyzing Images

Budget-Friendly with Nova

OpenAI SDK Compatibility

Error Responses

400 -- Invalid Request

402 -- Insufficient Funds

429 -- Rate Limited

Rate Limits