Add custom base URL support for Gemini embedding provider in Codebase Indexing

[therainisme]

What specific problem does this solve?

In Roo Code's Codebase Indexing configuration, when selecting Gemini as the Embedder Provider, there's no "Use custom base URL" option available, while other providers include this feature.

This issue primarily affects users who need proxy servers or mirror services. When users select Gemini as their embedding model, they can currently only use the hardcoded https://generativelanguage.googleapis.com/v1beta/openai/ endpoint, but expect to have a "Use custom base URL" option available.

This prevents users from using self-hosted or proxy services, making Gemini embeddings unusable in network-restricted environments.

Additional context (optional)

No response

Roo Code Task Links (Optional)

No response

Request checklist

• I've searched existing Issues and Discussions for duplicates
• This describes a specific problem with clear impact and context

Interested in implementing this?

• Yes, I'd like to help implement this feature

Implementation requirements

• I understand this needs approval before implementation begins

How should this be solved? (REQUIRED if contributing, optional otherwise)

No response

How will we know it works? (Acceptance Criteria - REQUIRED if contributing, optional otherwise)

No response

Technical considerations (REQUIRED if contributing, optional otherwise)

No response

Trade-offs and risks (REQUIRED if contributing, optional otherwise)

No response

[roomote]

I am working on implementing custom base URL support for the Gemini embedding provider in Codebase Indexing. This will add the missing "Use custom base URL" option to match the functionality available in other providers, enabling users to use proxy servers or mirror services with Gemini embeddings.

[roomote]

✅ Implementation Complete

I have successfully implemented custom base URL support for the Gemini embedding provider as requested in this issue.

🔗 Pull Request

PR #5745: #5745

📋 Implementation Summary

Backend Changes:

• ✅ Added codebaseIndexGeminiBaseUrl field to type definitions in packages/types/src/codebase-index.ts
• ✅ Updated CodeIndexConfigManager to handle Gemini base URL configuration and restart detection
• ✅ Modified CodeIndexServiceFactory to pass base URL parameter to GeminiEmbedder
• ✅ Verified GeminiEmbedder already supports optional baseUrl parameter

Frontend Changes:

• ✅ Added Gemini base URL input field in CodeIndexPopover.tsx (positioned between API key and model selection)
• ✅ Implemented proper validation schema with URL validation
• ✅ Updated TypeScript interfaces and configuration handling

Testing:

• ✅ Added comprehensive backend tests for configuration management (4 new test cases)
• ✅ Updated service factory tests with base URL parameter validation (2 new test cases)
• ✅ Added frontend validation tests (5 new test cases)
• ✅ All tests passing: Backend (76/76), Service Factory (39/39)
• ✅ Type checking and linting passed across all packages

🎯 Key Features

• Backward Compatible: Existing configurations continue to work without changes
• Optional Field: Users can leave base URL empty to use default Gemini endpoint
• Proper Validation: URL validation ensures only valid URLs are accepted
• Restart Detection: Configuration changes properly trigger service restarts
• Consistent UX: Follows same pattern as OpenAI Compatible and Ollama providers

🚀 Use Cases Enabled

This implementation allows users to:

• Use proxy servers for Gemini API access in corporate environments
• Use mirror services or alternative endpoints
• Work around network restrictions blocking generativelanguage.googleapis.com
• Maintain consistency with other embedding providers

The PR is currently going through CI checks. Once approved and merged, users will be able to configure custom base URLs for Gemini embeddings in their Codebase Indexing settings.

[shariqriazz]

Can't you just use OpenAI compatible endpoint for this?

[SatoshiReport]

ok but then it is the same issue. It tries to index 12,000+ blocks and times out at 6,600 and then won't run. We are hitting the api limit of the embedding endpoint.

[therainisme]

Can't you just use OpenAI compatible endpoint for this?

Using OpenAI-compatible endpoints would make all features and PRs developed specifically for Gemini unusable.

[SatoshiReport]

I tried OpenAI compatible endpoints using ollama and hugging face - none worked. I used Gemini but now get "Failed during initial scan: fetch failed". Tried ollama directly as well. nothing works. Thank you for Indexing but it does not seem to be production ready at all at least for a code base of any size which is when you would actually need it.