Is there a way to use bedrock anthropic in the playground?

Yes, Phoenix supports using Anthropic models via AWS Bedrock in the playground. You can instrument Bedrock calls using the `BedrockInstrumentor` and send traces to Phoenix, and the playground UI has added support for Anthropic tool calls and prompt conversion for Bedrock Anthropic models. However, you must ensure the correct model parameters and authentication are set up, and some features (like thinking blocks) are still being enhanced for full parity with direct Anthropic API usage. For setup, instrument `boto3` with `BedrockInstrumentor`, use the Bedrock client to call Anthropic models, and run the Phoenix server to observe traces. See the <https://github.com/Arize-ai/phoenix/blob/main/docs/tracing/integrations-tracing/bedrock.md|Bedrock integration guide> and <https://github.com/Arize-ai/phoenix/blob/main/CHANGELOG.md|recent playground enhancements> for details. Would you like more detail or a code example? Sources: - <https://github.com/Arize-ai/phoenix/blob/main/CHANGELOG.md|Phoenix CHANGELOG (playground anthropic support)> - <https://github.com/Arize-ai/openinference/blob/main/python/instrumentation/openinference-instrumentation-bedrock/README.md|OpenInference Bedrock Instrumentation> - <https://github.com/Arize-ai/phoenix/blob/main/docs/tracing/integrations-tracing/bedrock.md|Bedrock integration guide>

It sounds like you received an irrelevant answer from a model (possibly via Phoenix or Bedrock Anthropic). If you want to analyze or improve retrieval relevance, Phoenix provides tools to evaluate and trace such cases, letting you inspect why irrelevant context was retrieved and how often it happens using built-in relevance evaluators and trace inspection features (<https://github.com/Arize-ai/phoenix/blob/main/docs/retrieval/concepts-retrieval/troubleshooting-llm-retrieval-with-vector-stores.md|troubleshooting LLM retrieval>). Would you like guidance on how to debug or improve retrieval relevance for your application? Sources: - <https://github.com/Arize-ai/phoenix/blob/main/docs/retrieval/concepts-retrieval/troubleshooting-llm-retrieval-with-vector-stores.md|Troubleshooting LLM Retrieval>