Can session IDs be customized or managed manually in DeepSeek R1 and Together.ai

DeepSeek R1

DeepSeek R1 is an open-source reasoning model that excels in math, code, and reasoning tasks. While there is no specific documentation on customizing session IDs directly within DeepSeek R1, the model can be fine-tuned using custom datasets and integrated with various platforms that support session management. For instance, when using DeepSeek R1 with platforms like Langfuse or Arize, you can manage sessions by adding a session ID to traces or spans, which helps group interactions and analyze performance[2][6].

However, DeepSeek R1 itself does not provide native support for managing session IDs. Any customization would likely involve integrating it with external tools or services that handle session management.

Together.ai

Together.ai offers a platform for deploying and managing AI models, including DeepSeek R1. While Together.ai provides tools for training and deploying models, there is no specific documentation on manually customizing session IDs directly within their platform. However, when integrating Together.ai with other tools or services like Arize, you can use session IDs to group traces and analyze model performance[7].

Similar to DeepSeek R1, Together.ai does not seem to offer native support for customizing session IDs. Instead, users would need to rely on external integrations or services that support session management.

Customization Approach

To customize or manage session IDs with these platforms, you would typically need to integrate them with external tools or services designed for session management, such as Langfuse or Arize. These tools allow you to add session IDs to traces or spans, enabling better analysis and grouping of interactions.

For example, with Langfuse, you can create a session ID when tracing features in your application:

javascript
const langfuse = new Langfuse();
const trace = langfuse.trace({
  sessionId: "your-session-id",
});

Similarly, with Arize, you can use session IDs to group traces and analyze performance:

python
# Using OpenInference instrumentation
from openinference_instrumentation_openai import using_session

with using_session(session_id="your-session-id"):
    # Your LLM calls here
    pass

In summary, while DeepSeek R1 and Together.ai do not provide direct support for customizing session IDs, you can achieve this functionality by integrating them with external tools designed for session management.

Citations:
[1] https://www.firecrawl.dev/blog/fine-tuning-deepseek
[2] https://langfuse.com/docs/tracing-features/sessions
[3] https://www.together.ai/products
[4] https://www.together.ai/models/deepseek-r1
[5] https://www.reddit.com/r/JanitorAI_Official/comments/1ij1d3w/how_to_hopefully_use_deepseek/
[6] https://docs.arize.com/arize/llm-tracing/how-to-tracing-manual/sessions-and-users
[7] https://docs.arize.com/arize/llm-tracing/tracing-integrations-auto/together-ai
[8] https://docs.together.ai/docs/deepseek-r1
[9] https://aws.amazon.com/blogs/machine-learning/customize-deepseek-r1-distilled-models-using-amazon-sagemaker-hyperpod-recipes-part-1/