[alexcastgonz (pcbeingused333)]

Motivation and Context

GoogleAITextEmbedding.generate_raw_embeddings sent every input text in a single embed_content request. The Gemini batchEmbedContents endpoint accepts at most 100 inputs per request (the same limit langchain-google-genai batches around), so calling generate_embeddings with more than 100 texts, e.g. when indexing a document set into a vector store, fails with a 400 instead of returning the embeddings.

Description

Split texts into chunks of at most 100 (MAX_EMBEDDING_BATCH_SIZE), call embed_content once per chunk and concatenate the embeddings in input order. The three client paths (user-supplied client, Vertex AI, API key) share the same helper. Requests with 100 or fewer texts are unchanged (a single call). The Vertex AI embedding service (vertex_ai_text_embedding.py) is left untouched: its limit was not verified.

Tests

New test_raw_embedding_batches_large_inputs: 250 texts produce calls of 100/100/50 and the results come back in order. It fails on main and passes with the fix; the rest of the file still passes.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines
• All unit tests pass, and I have added new tests where possible

This change was prepared with the assistance of an AI coding tool (Claude Code) and reviewed before submission.

[Copilot]

🟢 Approval recommended

There are no active review issues for the Google AI embedding-request batching change.

No open issues 🎉

What changed in this PR

This PR updates the Python Google AI embedding service to batch large text collections within Gemini’s 100-input request limit while preserving result order.

Changes:

• Adds a shared batching helper for custom-client, Vertex AI, and API-key paths.
• Adds a regression test verifying that 250 inputs produce batches of 100, 100, and 50 with embeddings returned in order.

File	Description
python/​tests/​unit/​connectors/​ai/​google/​google_ai/​services/​test_google_ai_text_embedding.py	Tests batch sizes, model selection, and output order for 250 inputs.
python/​semantic_kernel/​connectors/​ai/​google/​google_ai/​services/​google_ai_text_embedding.py	Limits embedding requests to 100 inputs and combines results across batches.

_{Review effort: Balanced}