generateContent

Compatible with the Gemini API’s generateContent; you can use the Google Gen AI SDKs directly.

POST https://noviahub.com/v1beta/models/{model}:generateContent
POST https://noviahub.com/v1beta/models/{model}:streamGenerateContent?alt=sse

Replace {model} with a model ID, such as gemini-3-flash. For models whose endpoint labels include Gemini (check on Models & Pricing).

Authentication

Use any one of these:

Form	Example
Header x-goog-api-key	x-goog-api-key: <your API key>
Query parameter key	...:generateContent?key=<your API key>
Header Authorization	Authorization: Bearer <your API key>

Also send Content-Type: application/json.

Request parameters

Write the body as defined by Gemini. The table lists the fields NoviaHub recognises and forwards; whether a parameter takes effect depends on the model.

Parameter	Type	Required	Description
contents	array	Yes	The conversation. Each item has a role (user or model) and parts.
systemInstruction	object	No	System instruction, shaped like one item of contents. The snake_case system_instruction is also accepted.
generationConfig	object	No	Generation settings; fields below.
tools	array	No	Tools (function declarations, search and so on), forwarded as is.
toolConfig	object	No	Tool calling settings.
safetySettings	array	No	Safety settings.
cachedContent	string	No	Use a context cache you created earlier.

Each item in parts can be text, inlineData (mimeType plus Base64 data, for images and other files), fileData, functionCall, functionResponse and so on.

Fields accepted in generationConfig: temperature, topP, topK, maxOutputTokens, candidateCount, stopSequences, responseMimeType, responseSchema, responseJsonSchema, presencePenalty, frequencyPenalty, seed, responseLogprobs, logprobs, responseModalities, mediaResolution, thinkingConfig (includeThoughts, thinkingBudget, thinkingLevel), speechConfig, imageConfig. The snake_case forms (such as max_output_tokens) are also accepted.

Note

Fields not on this list are dropped.

Sample code

cURL

curl "https://noviahub.com/v1beta/models/gemini-3-flash:generateContent" \
  -H "x-goog-api-key: $NOVIAHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"parts": [{"text": "Explain what an API is in one sentence."}]}]
  }'

Python

import os

from google import genai
from google.genai import types

client = genai.Client(
    api_key=os.environ["NOVIAHUB_API_KEY"],
    http_options=types.HttpOptions(base_url="https://noviahub.com"),
)

response = client.models.generate_content(
    model="gemini-3-flash",
    contents="Explain what an API is in one sentence.",
)

print(response.text)

TypeScript

import { GoogleGenAI } from '@google/genai'

const ai = new GoogleGenAI({
  apiKey: process.env.NOVIAHUB_API_KEY,
  httpOptions: { baseUrl: 'https://noviahub.com' },
})

const response = await ai.models.generateContent({
  model: 'gemini-3-flash',
  contents: 'Explain what an API is in one sentence.',
})

console.log(response.text)

The Google Gen AI SDK’s base URL is just https://noviahub.com; don’t include /v1beta.

Response

Field	Description
candidates[].content.parts[]	The reply. Text is in text; images from image models are in inlineData (mimeType plus Base64 data).
candidates[].finishReason	Why generation stopped, such as STOP (finished) or MAX_TOKENS (hit the limit).
usageMetadata.promptTokenCount	Input tokens.
usageMetadata.candidatesTokenCount	Output tokens.
usageMetadata.totalTokenCount	Total.
usageMetadata.thoughtsTokenCount	Tokens spent on thinking (when the model reports it).
usageMetadata.cachedContentTokenCount	Tokens served from cache (when the model reports it).
modelVersion	Model version.
responseId	ID of this reply.

Example response (test instance; content from a simulated upstream)

{
  "candidates": [
    {
      "content": { "role": "model", "parts": [{ "text": "Hello from the mock upstream." }] },
      "finishReason": "STOP",
      "index": 0
    }
  ],
  "usageMetadata": { "promptTokenCount": 12, "candidatesTokenCount": 7, "totalTokenCount": 19 },
  "modelVersion": "gemini-3-flash",
  "responseId": "mock-resp-123"
}

When you call a non-Gemini model here (through protocol conversion), the response also contains safetyRatings (an empty array) and some internal fields in usageMetadata; you can ignore them.

Streaming

Use streamGenerateContent and add ?alt=sse to the URL. The response arrives as SSE; each data: piece has the same structure as above, with text arriving piece by piece in candidates[0].content.parts[0].text. The last piece carries finishReason and the full usageMetadata.

In our tests NoviaHub answered with SSE even without ?alt=sse. Add it anyway so every client behaves the same.

Example stream (test instance)

data: {"candidates":[{"content":{"role":"model","parts":[{"text":"Hello"}]},"index":0}],"usageMetadata":{"promptTokenCount":12,"totalTokenCount":12},"modelVersion":"gemini-3-flash","responseId":"mock-resp-123"}

data: {"candidates":[{"content":{"role":"model","parts":[{"text":" from the mock upstream."}]},"finishReason":"STOP","index":0}],"usageMetadata":{"promptTokenCount":12,"candidatesTokenCount":7,"totalTokenCount":19},"modelVersion":"gemini-3-flash","responseId":"mock-resp-123"}

cURL

curl "https://noviahub.com/v1beta/models/gemini-3-flash:streamGenerateContent?alt=sse" \
  -H "x-goog-api-key: $NOVIAHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -N \
  -d '{"contents": [{"parts": [{"text": "Write a four-line poem."}]}]}'

Python

import os

from google import genai
from google.genai import types

client = genai.Client(
    api_key=os.environ["NOVIAHUB_API_KEY"],
    http_options=types.HttpOptions(base_url="https://noviahub.com"),
)

for chunk in client.models.generate_content_stream(
    model="gemini-3-flash",
    contents="Write a four-line poem.",
):
    print(chunk.text or "", end="", flush=True)

TypeScript

import { GoogleGenAI } from '@google/genai'

const ai = new GoogleGenAI({
  apiKey: process.env.NOVIAHUB_API_KEY,
  httpOptions: { baseUrl: 'https://noviahub.com' },
})

const stream = await ai.models.generateContentStream({
  model: 'gemini-3-flash',
  contents: 'Write a four-line poem.',
})

for await (const chunk of stream) {
  process.stdout.write(chunk.text ?? '')
}

Generating images

Gemini image models (such as gemini-3.1-flash-image) use this endpoint too. Put the image model’s ID in {model}; the generated image is in inlineData in the response. Decode data from Base64 and save it as a file of the given mimeType. These models can’t be called through /v1/images/generations; see Image generation.

Not supported

• :countTokens (token counting) is not available and returns 404.

Common errors

HTTP status	Cause
400	The body isn’t valid JSON.
401	The key is invalid, disabled, expired or out of quota, or it was sent in the wrong place.
403	The key isn’t allowed to use this model, the IP isn’t on the allow list, or the account balance is too low.
503	Wrong model ID, or no channel is currently available for the model.

Errors use the OpenAI shape ({"error": {...}}); see API overview · Error format. See Errors and troubleshooting for details.