voideditor
[Feature] Request: Codebase Indexing and Custom Docs Integration
Description:
It would be extremely helpful to have native support for codebase indexing with embeddings and the ability to add custom documentation within the editor environment.
Proposed features:
-
Codebase indexing: Automatically parse and index all files in a repo or project directory (with an indicator of progress and file count).
-
Embeddings-based search and navigation: Enable smarter code navigation and contextual answers based on embeddings (stored locally, metadata optionally in the cloud).
-
Resync/Delete controls for maintaining or updating the index.
-
Custom documentation section: Ability to add markdown or plain text documents that can be referenced during development or AI interactions (e.g. with @Add or similar syntax).
-
UI example attached to illustrate how this might be visually structured.
Use case:
-
Improved developer efficiency with context-aware suggestions.
-
Enhanced AI assistant capabilities by leveraging the full context of the codebase and supporting documents.
-
Easier documentation referencing without switching tabs or external viewers.
I would like to be able to pick my tools and services for these... imo needs to have interfaces so we can extend with provider implementations
@osvfelices
- Vector store
- Embeddings model
- indexing logic
- searching logic
- additional workspace context logic
... what else might we need? And what is the relationship between searching logic and other tool calls?
@zpg6 Good breakdown, I’d add:
Document parser (for Markdown, code comments, OpenAPI, etc.)
Context prioritization logic (to manage what gets injected into prompts when context space is limited)
Snapshot/versioning support (to handle re-indexing and track changes intelligently)
As for the searching logic, I see it as a layer that queries the vector store and optionally enriches tool calls (like ask, autocomplete, or AI chat/agents), but it should remain decoupled from the tool execution logic itself.
Does that align with what you had in mind?
When I saw videos like "Void AI is the new Cursor killer," I was hoping that features such as codebase indexing and the ability to add documentation were already available in Void AI. I don't see how you can work with an AI agent without being able to provide him with documentation.
I'm really looking forward to adding this feature!
@NANAMINER, not sure if the dev teams see this as a priority. @andrewpareles ?
#225 is relevant.
Generally indexing is not a huge priority for us because of agents. Humans don’t use vector databases and we worry about implementing something potentially not useful in the future as a small team.
That said, we’d welcome contributions around this, and I’m sure a lot of people would appreciate it. I think it’d make sense to talk first, want to shoot me a DM on discord?
In response to some other comments - parsing online docs is a higher priority than codebase indexing, but MCP servers might be a good enough solution
In response to some other comments - parsing online docs is a higher priority than codebase indexing, but MCP servers might be a good enough solution
Inside cursor when ask something it would generate a question and use vector search to find the accurate area to change instead of using the whole context to know where the code is.
I might be wrong but I think vector indexing and searching based on it will be huge
