litellm - 💡(How to fix) Fix [Feature]: LiteLLM Cost Map Langfuse Sync Or enable Lang fuse ingesting [1 participants]

TL;DR

The user needs to find a way to ingest costs of models not present in the default Langfuse model table or sync the cost map with the Langfuse model table.

Guidance

• Review the LiteLLM and Langfuse documentation again to ensure no feature or API is missed that allows ingesting costs of non-default models.
• Investigate if the Langfuse Model Table can be manually updated or extended with custom models to enable cost inference.
• Check if there are any configuration options or parameters in the LiteLLM Python package (SDK) that could facilitate the ingestion of custom model costs.
• Consider reaching out to the LiteLLM or Langfuse community, or opening a discussion, as the solution might not be straightforward and could require collaboration or a feature request.

Notes

The issue seems to stem from a limitation in either the LiteLLM SDK or the Langfuse model table's ability to handle custom models. Without direct access to the documentation or the ability to test the SDK, providing a precise solution is challenging.

Recommendation

Apply workaround: Since the issue seems to involve missing functionality or an undocumented feature, applying a workaround such as manually syncing models or extending the model table, if possible, seems to be the most viable immediate solution.