feat(adk): add native Bedrock embedding support for agent memory

[go-faustino]

Summary

Adds native AWS Bedrock embedding support to KagentMemoryService, so agent memory works out of the box with Bedrock embedding models (e.g. amazon.titan-embed-text-v1, amazon.titan-embed-text-v2:0).

Currently, when provider="bedrock" is set on an embedding ModelConfig, the memory service falls back to an OpenAI-compatible client which fails because no OPENAI_API_KEY is available:

WARNING - Unknown embedding provider 'bedrock'; attempting OpenAI-compatible call.
ERROR   - The api_key client option must be set either by passing api_key to the client
          or by setting the OPENAI_API_KEY environment variable

Changes

• _memory_service.py: Add _embed_bedrock() method and register "bedrock" in _call_embedding_provider dispatch
• test_embedding.py: Add 3 unit tests (basic embedding, region from env, error handling)

Implementation details

• Uses boto3.client("bedrock-runtime").invoke_model(), consistent with how KAgentBedrockLlm handles LLM calls in models/_bedrock.py
• Region resolved from AWS_DEFAULT_REGION / AWS_REGION env vars (same credential chain as the Bedrock LLM provider — IRSA, instance profile, etc.)
• Each text is embedded individually since the Titan Embedding API accepts a single inputText per invocation; asyncio.gather parallelises the calls
• The existing truncation-to-768 + L2-normalization logic in _generate_embedding_async handles dimension differences across Titan v1 (1536d) and v2 (1024d) models, so no model-specific dimensions parameter is needed
• boto3 is already a declared dependency of kagent-adk

Testing

cd python
uv run pytest packages/kagent-adk/tests/unittests/test_embedding.py -v
# 13 passed (10 existing + 3 new)

New tests:

• test_bedrock_embed — verifies embedding generation and truncation to 768 dimensions
• test_bedrock_embed_uses_region_from_env — verifies AWS_REGION env var is passed to boto3 client
• test_bedrock_embed_error_returns_empty — verifies graceful error handling returns []

[Copilot]

Pull request overview

This PR adds first-class AWS Bedrock (Titan) embedding support to KagentMemoryService, allowing agent memory embeddings to work when EmbeddingConfig.provider="bedrock" without falling back to an OpenAI-compatible client.

Changes:

• Add a Bedrock embedding implementation (_embed_bedrock) and wire it into the embedding-provider dispatcher.
• Use boto3.client("bedrock-runtime").invoke_model() per text (parallelized via asyncio.gather + to_thread).
• Add unit tests covering Bedrock embedding, region selection via env vars, and error handling.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
python/packages/kagent-adk/src/kagent/adk/_memory_service.py	Adds Bedrock embedding provider dispatch + _embed_bedrock() implementation using boto3 Bedrock Runtime.
python/packages/kagent-adk/tests/unittests/test_embedding.py	Adds unit tests validating Bedrock embedding behavior, env-region handling, and graceful failure.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[supreme-gg-gg]

lgtm, thanks!