Neosantara

Overview

Property	Details
Description	Neosantara provides an OpenAI-compatible API for chat completions and the Responses API.
Provider Route on LiteLLM	neosantara/
Link to Provider Doc	Neosantara ↗
Base URL	https://api.neosantara.xyz/v1
Supported Operations	/chat/completions, /responses, function calling

We support Neosantara models through the neosantara/ prefix in LiteLLM.

Required Variables

Environment Variables

os.environ["NEOSANTARA_API_KEY"] = ""  # your Neosantara API key

A free-tier Neosantara account can call gemini-3.1-flash-lite, which the examples on this page use. Some models, such as gemini-3.5-flash, answer 403 insufficient_tier until the account has deposited credits.

You can override the default base URL with:

Optional Base URL Override

os.environ["NEOSANTARA_API_BASE"] = "https://api.neosantara.xyz/v1"

Usage - LiteLLM Python SDK

Non-streaming

Neosantara Non-streaming Completion

import os
from litellm import completion

os.environ["NEOSANTARA_API_KEY"] = ""

response = completion(
    model="neosantara/gemini-3.1-flash-lite",
    messages=[{"role": "user", "content": "Hello from LiteLLM"}],
    max_tokens=64,
)

print(response.choices[0].message.content)

Streaming

Neosantara Streaming Completion

import os
from litellm import completion

os.environ["NEOSANTARA_API_KEY"] = ""

response = completion(
    model="neosantara/gemini-3.1-flash-lite",
    messages=[{"role": "user", "content": "Write a one-line poem about Jakarta"}],
    max_tokens=64,
    stream=True,
)

for chunk in response:
    if chunk.choices[0].delta.content is not None:
        print(chunk.choices[0].delta.content, end="")

Function Calling

Chat Completions

Function calling works on Neosantara through completion(). Depending on the model, LiteLLM may require explicitly allowing the OpenAI params on this provider route.

Neosantara Function Calling via completion()

import os
from litellm import completion

os.environ["NEOSANTARA_API_KEY"] = ""

response = completion(
    model="neosantara/gemini-3.1-flash-lite",
    messages=[{"role": "user", "content": "What is the weather in Jakarta? Use the tool."}],
    tools=[
        {
            "type": "function",
            "function": {
                "name": "get_weather",
                "description": "Get weather by city name",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "city": {"type": "string"}
                    },
                    "required": ["city"]
                }
            }
        }
    ],
    tool_choice="auto",
    allowed_openai_params=["tools", "tool_choice"],
    max_tokens=128,
)

print(response.choices[0].message.tool_calls)

Responses API

Neosantara also supports LiteLLM's responses() interface.

Neosantara Responses API

import os
from litellm import responses

os.environ["NEOSANTARA_API_KEY"] = ""

response = responses(
    model="neosantara/gemini-3.1-flash-lite",
    input="Reply with exactly: pong",
    max_output_tokens=8,
)

print(response.output_text)

Responses API Tool Calling

Function calling also works through responses():

Neosantara Function Calling via responses()

import os
from litellm import responses

os.environ["NEOSANTARA_API_KEY"] = ""

response = responses(
    model="neosantara/gemini-3.1-flash-lite",
    input="What is the weather in Jakarta? Use the tool.",
    tools=[
        {
            "type": "function",
            "name": "get_weather",
            "description": "Get weather by city name",
            "parameters": {
                "type": "object",
                "properties": {
                    "city": {"type": "string"}
                },
                "required": ["city"]
            }
        }
    ],
    max_output_tokens=128,
)

print(response.output)

Parameter Compatibility

LiteLLM maps max_completion_tokens to max_tokens for Neosantara automatically.

Neosantara max_completion_tokens Compatibility

import os
from litellm import completion

os.environ["NEOSANTARA_API_KEY"] = ""

response = completion(
    model="neosantara/gemini-3.1-flash-lite",
    messages=[{"role": "user", "content": "Summarize LiteLLM in one sentence"}],
    max_completion_tokens=64,
)

print(response.choices[0].message.content)

Reasoning Effort

LiteLLM can pass reasoning_effort through to Neosantara, but like tool calling on completion(), you should explicitly allow the parameter on this provider route.

Neosantara reasoning_effort

import os
from litellm import completion

os.environ["NEOSANTARA_API_KEY"] = ""

response = completion(
    model="neosantara/gemini-3.1-flash-lite",
    messages=[{"role": "user", "content": "Reply with exactly: reasoned"}],
    reasoning_effort="high",
    allowed_openai_params=["reasoning_effort"],
    max_tokens=8,
)

print(response.choices[0].message.content)

Usage - LiteLLM Proxy

Add the following to your LiteLLM Proxy configuration file:

config.yaml

model_list:
  - model_name: neosantara-gemini-flash-lite
    litellm_params:
      model: neosantara/gemini-3.1-flash-lite
      api_key: os.environ/NEOSANTARA_API_KEY

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY

Start your LiteLLM Proxy server:

Start LiteLLM Proxy

export NEOSANTARA_API_KEY="your-neosantara-api-key"
export LITELLM_MASTER_KEY="${LITELLM_MASTER_KEY:-sk-$(openssl rand -hex 16)}"
printf 'Local proxy key: %s\n' "$LITELLM_MASTER_KEY"
litellm --config config.yaml

# RUNNING on http://0.0.0.0:4000

In another terminal, set LITELLM_MASTER_KEY to the key printed above and call the proxy:

OpenAI SDK

Neosantara via Proxy - OpenAI SDK

import os
from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:4000",
    api_key=os.environ["LITELLM_MASTER_KEY"],
)

response = client.chat.completions.create(
    model="neosantara-gemini-flash-lite",
    messages=[{"role": "user", "content": "Say hello from the proxy"}],
)

print(response.choices[0].message.content)

LiteLLM SDK

Neosantara via Proxy - LiteLLM SDK

import os
import litellm

response = litellm.completion(
    model="litellm_proxy/neosantara-gemini-flash-lite",
    messages=[{"role": "user", "content": "Say hello from the proxy"}],
    api_base="http://localhost:4000",
    api_key=os.environ["LITELLM_MASTER_KEY"],
)

print(response.choices[0].message.content)

cURL

Neosantara via Proxy - cURL

curl http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -d '{
    "model": "neosantara-gemini-flash-lite",
    "messages": [{"role": "user", "content": "Say hello from the proxy"}]
  }'

Neosantara via Proxy - Responses API

curl http://localhost:4000/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -d '{
    "model": "neosantara-gemini-flash-lite",
    "input": "Reply with exactly: pong"
  }'

Tools and reasoning_effort on the Proxy

Depending on the model, the proxy can answer 400 UnsupportedParamsError to tools, tool_choice, parallel_tool_calls, or reasoning_effort, as it does for glm-4.7-flash. Allow the params on the deployment, the same way allowed_openai_params works on completion():

config.yaml

model_list:
  - model_name: neosantara-glm-flash
    litellm_params:
      model: neosantara/glm-4.7-flash
      api_key: os.environ/NEOSANTARA_API_KEY
      allowed_openai_params: ["tools", "tool_choice", "parallel_tool_calls", "reasoning_effort"]

Spend Tracking on the Proxy

Neosantara models are not in LiteLLM's model cost map, so the proxy records $0 spend for them and key or team budgets never move. To track spend, set input_cost_per_token and output_cost_per_token in litellm_params, as shown in Custom Pricing. Both are in USD per token, so divide a Neosantara price per million tokens by 1,000,000.

Notes

• Use the neosantara/ model prefix when calling Neosantara through LiteLLM.
• The default auth env var is NEOSANTARA_API_KEY.
• NEOSANTARA_API_BASE can be used to point LiteLLM at a custom Neosantara-compatible base URL.
• max_completion_tokens is mapped to max_tokens automatically.
• On completion() and on the proxy, tools, tool_choice, parallel_tool_calls, and reasoning_effort may require allowed_openai_params so LiteLLM will pass them through.