Code Index has no limit on batch size

[A0nameless0man]

App Version

Version: 3.26.7 (1d70f95)

API Provider

OpenAI Compatible

Model Used

qwen3-embedding text-embedding-v4

Roo Code Task Links (Optional)

Qwne3-Embedding model has batch size limit,request input can have at most 10 string in it

https://bailian.console.aliyun.com/console?tab=doc#/doc/?type=model&url=2842587

🔁 Steps to Reproduce

just index some code base with more than 10 chunk with aliyun bailian

💥 Outcome Summary

got Indexing failed: 尝试 3 次后批次处理失败：尝试 3 次后创建嵌入失败：HTTP 400 - 400 <400> InternalError.Algo.InvalidParameter: Value error, batch size is invalid, it should not be larger than 10.: input.contents

📄 Relevant Logs or Errors (Optional)

[roomote]

I'm exploring a potential fix for the batch size limit issue with Qwen3-Embedding models. The error indicates that the embedding API has a maximum batch size of 10, but the code indexing is attempting to send larger batches. I'll work on implementing batch size limits for embedding requests.

[A0nameless0man]

I'm exploring a potential fix for the batch size limit issue with Qwen3-Embedding models. The error indicates that the embedding API has a maximum batch size of 10, but the code indexing is attempting to send larger batches. I'll work on implementing batch size limits for embedding requests.

@roomote-agent
I suggest make batch size limit and batch token size limit configable

Roo-Code/src/services/code-index/embedders/openai-compatible.ts

Line 4 in 55ea328

MAX_BATCH_TOKENS,

[roomote]

I've created a pull request to fix this issue: #7704

The fix implements configurable batch size limits for embedding models. Specifically for Aliyun Bailian's Qwen3-Embedding models, the batch size is now limited to 10 items per request as required by their API.

Key changes:

• Added maxBatchSize property to embedding model profiles
• Configured Qwen3-Embedding and text-embedding-v4 models with 10-item batch limit
• Updated OpenAICompatibleEmbedder to respect model-specific batch limits
• Added comprehensive tests to ensure batch size limiting works correctly

All CI checks have passed successfully. This should resolve the indexing failures you were experiencing.

[daniel-lxs]

Hey @A0nameless0man This is already a setting: