Roo code reindex my large codebase everytime I open my repo

[stathamos]

App Version

3.25.23

API Provider

LM Studio

Model Used

Qwen3-Coder-A3B-6bits

Roo Code Task Links (Optional)

No response

🔁 Steps to Reproduce

🔁 Steps to Reproduce

• Make sure a vector database is properly configured and running in Docker (confirmed working).
• Index a large codebase (takes about 3 hours).
• Shut down the computer after indexing is completed.
• Restart the computer the next day and reopen VSCode.

✅ Expected Behavior

• The already indexed codebase should be reused without requiring a full reindex.
• Only modified files should trigger partial reindexing.
• The existing vector database should store and reload the index across sessions.

❌ Actual Behavior

📦 Environment

• OS: macOS Sequoia
• Editor: VSCode
• Vector DB: running in Docker, properly configured and confirmed working
• Settings: Default

💥 Outcome Summary

• After restarting the computer, the extension starts a full reindex of the entire codebase, even if no changes were made.
• The configured vector database is up and running in Docker, but the extension does not seem to reuse the stored index.
• On large projects, this makes the feature almost unusable due to the very long indexing time.
• The issue only happens after a full reboot — not immediately after indexing.
• Suggests that the index/cache is not being persisted or retrieved correctly from the vector database.

📄 Relevant Logs or Errors (Optional)

[roomote]

I'm exploring a potential fix for the vector database index persistence issue. The problem appears to be that the index isn't being properly persisted or retrieved across system restarts, causing unnecessary full reindexing. I'll investigate the codebase and open a PR to share progress.

[roomote]

I've opened PR #7409 with a fix for this issue: #7409

The root cause was that after a system restart, initial connection attempts to the Qdrant vector database were failing, causing the system to incorrectly believe the collection didn't exist. This triggered the creation of a new collection and cleared the cache, resulting in a full reindex.

The fix adds:

• Connection retry logic with exponential backoff
• Better differentiation between connection failures and missing collections
• Cache preservation when reconnecting to existing collections
• Enhanced logging for debugging

This should eliminate the unnecessary reindexing after system restarts. The CI checks are currently running on the PR.

[daniel-lxs]

Closing as duplicate of #5941