LearnNewsExamplesServices
Frontmatter
id11402
title[Memory Core] Support lightweight lms server start in TextEmbeddingService
stateClosed
labels
enhancementai
assigneesneo-gpt
createdAtMay 15, 2026, 10:26 AM
updatedAtMay 19, 2026, 8:36 AM
githubUrlhttps://github.com/neomjs/neo/issues/11402
authorneo-gemini-3-1-pro
commentsCount0
parentIssuenull
subIssues[]
subIssuesCompleted0
subIssuesTotal0
blockedBy[]
blocking[]
closedAtMay 19, 2026, 8:36 AM

[Memory Core] Support lightweight lms server start in TextEmbeddingService

Closedenhancementai
neo-gemini-3-1-pro
neo-gemini-3-1-pro commented on May 15, 2026, 10:26 AM

Bug Description

N/A - enhancement

Steps to Reproduce

N/A - enhancement

Expected Behavior

TextEmbeddingService currently detects LM Studio when running as a full desktop program (Shape A and Shape B retries). However, LM Studio also offers a lightweight headless mode via lms server start. This headless mode provides lower overhead for testing and deployment. We need to support detecting and handling lms server start alongside the full desktop program.

Actual Behavior

Currently, TextEmbeddingService retry heuristics are tailored for the full desktop program.

Proposed Solution

  1. Identify if the connection/retry heuristics need adjustment when connecting to lms server start.
  2. Ensure both full desktop program and lightweight lms server modes are fully supported.
tobiu referenced in commit 014a66f - "test(memory-core): cover lms server start embeddings (#11402) (#11616) on May 19, 2026, 8:36 AM
tobiu closed this issue on May 19, 2026, 8:36 AM