> ## Documentation Index
> Fetch the complete documentation index at: https://docs.neosantara.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Conversations & State

> Persist conversation history server-side using previous_response_id with OpenAI SDK.

The Responses API (`/v1/responses`) eliminates the need to re-transmit expanding transcripts of previous messages on every interaction turn. By setting `store=True` with the official [OpenAI](https://openai.com/?utm_source=neosantara-docs\&utm_medium=referral) SDK, Neosantara maintains conversation state on the gateway, enabling direct multi-turn interaction via `previous_response_id`.

## Multi-Turn Workflows with OpenAI SDK

Send your initial prompt with `store=True`. On subsequent turns, provide `previous_response_id` referencing the earlier response ID. The gateway reconstructs context automatically.

<CodeGroup>
  ```python Python (OpenAI SDK) icon="python" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  from openai import OpenAI
  import os

  client = OpenAI(
      base_url="https://api.neosantara.xyz/v1",
      api_key=os.environ["NEOSANTARA_API_KEY"]
  )

  # Turn 1: Store context on the gateway
  turn1 = client.responses.create(
      model="deepseek-v4.1-flash",
      input="Hello, my name is Alex and I am a network engineer based in Jakarta.",
      instructions="Respond in a helpful, professional tone.",
      store=True
  )

  print(f"Turn 1 ID: {turn1.id}")
  print(f"Turn 1 Output:\n{turn1.output_text}\n")

  # Turn 2: Continue conversation referencing Turn 1 ID
  turn2 = client.responses.create(
      model="deepseek-v4.1-flash",
      input="Based on my role and location, which certification should I pursue first?",
      previous_response_id=turn1.id,
      store=True
  )

  print(f"Turn 2 ID: {turn2.id}")
  print(f"Turn 2 Output (Context preserved):\n{turn2.output_text}\n")

  # Turn 3: Chain further from Turn 2
  turn3 = client.responses.create(
      model="deepseek-v4.1-flash",
      input="What is the estimated examination cost for that certification in Indonesia?",
      previous_response_id=turn2.id,
      store=True
  )

  print(f"Turn 3 Output:\n{turn3.output_text}")
  ```

  ```javascript TypeScript (Node.js) icon="js" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  import OpenAI from "openai";

  const client = new OpenAI({
    baseURL: "https://api.neosantara.xyz/v1",
    apiKey: process.env.NEOSANTARA_API_KEY,
  });

  // Turn 1: Persist context
  const turn1 = await client.responses.create({
    model: "deepseek-v4.1-flash",
    input: "Hello, my name is Alex and I am a network engineer based in Jakarta.",
    store: true,
  });

  console.log("Turn 1 Output:", turn1.output_text);

  // Turn 2: Forward previous_response_id
  const turn2 = await client.responses.create({
    model: "deepseek-v4.1-flash",
    input: "Based on my background, which certification should I pursue first?",
    previous_response_id: turn1.id,
    store: true,
  });

  console.log("Turn 2 Output:", turn2.output_text);

  // Turn 3: Chain further
  const turn3 = await client.responses.create({
    model: "deepseek-v4.1-flash",
    input: "What is the estimated examination cost for that certification in Indonesia?",
    previous_response_id: turn2.id,
    store: true,
  });

  console.log("Turn 3 Output:", turn3.output_text);
  ```
</CodeGroup>

## Inspecting Conversation State

Inspect saved response metadata and token consumption details at any time using `client.responses.retrieve`.

<CodeGroup>
  ```python Python (OpenAI SDK) icon="python" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  from openai import OpenAI
  import os

  client = OpenAI(
      base_url="https://api.neosantara.xyz/v1",
      api_key=os.environ["NEOSANTARA_API_KEY"]
  )

  # Inspect token breakdown and metadata from stored response
  saved_response = client.responses.retrieve("resp_01j7x8abc...")

  print(f"Model ID: {saved_response.model}")
  print(f"Status: {saved_response.status}")
  print(f"Total Tokens: {saved_response.usage.total_tokens}")
  ```

  ```javascript TypeScript (Node.js) icon="js" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  import OpenAI from "openai";

  const client = new OpenAI({
    baseURL: "https://api.neosantara.xyz/v1",
    apiKey: process.env.NEOSANTARA_API_KEY,
  });

  const savedResponse = await client.responses.retrieve("resp_01j7x8abc...");

  console.log(`Model ID: ${savedResponse.model}`);
  console.log(`Status: ${savedResponse.status}`);
  console.log(`Total Tokens: ${savedResponse.usage.total_tokens}`);
  ```
</CodeGroup>

## Benefits of Server-Side State

* **Bandwidth Efficiency:** Clients do not re-upload expanding transcripts of prior messages on every request.
* **Reliable Continuity:** State is persisted in high-speed, durable cache layers on the gateway.
* **Ideal for Thin Clients:** Well-suited for mobile applications, messaging bots, and IoT hardware with payload limitations.

## Next Steps

| Task                    | Guide                                                         |
| :---------------------- | :------------------------------------------------------------ |
| Background Execution    | [Background Tasks](/en/gateway/responses-api/background-jobs) |
| Server Tools & MCP      | [Tools & MCP](/en/gateway/responses-api/tools)                |
| Responses API Reference | [Responses API](/en/gateway/responses-api)                    |


## Related topics

- [OpenResponses API](/en/gateway/responses-api.md)
- [Background Tasks & Webhooks](/en/gateway/responses-api/background-jobs.md)
- [Chat Completions](/en/gateway/chat-completions.md)