[FEAT]: Support GPT4All-style LocalDocs RAG with one-click indexing of multiple folders and all nested directories. #3173

New issue

Open

opened 2026-02-28 06:32:33 -05:00 by deekerman · 0 comments

deekerman commented

2026-02-28 06:32:33 -05:00

Owner

Originally created by @TheAlex25 on GitHub (Feb 13, 2026).

What would you like to see?

Feature Request: LocalDocs-Style RAG with One-Click Multi-Folder Indexing in AnythingLLM

Title:
Support GPT4All-style LocalDocs RAG with one-click indexing of multiple folders and all nested directories.

Summary:
Add a LocalDocs-equivalent retrieval-augmented generation (RAG) feature to AnythingLLM that allows users to select one or more folders — including nested subfolders — from the operating system file explorer (Windows Explorer, macOS Finder, Ubuntu Files) and index them for semantic search and contextual retrieval in a single action. This would replace the current manual and clunky ingestion process.

Current Limitation:
AnythingLLM’s existing document ingestion for RAG requires manual uploads and repeated steps to add large collections or complex directory structures. There is no easy way to select entire folder trees at once for indexing.

Proposed Behavior:

One-Click Folder Selection:
Allow users to right-click one or multiple folders in the OS file explorer (Windows, macOS, Linux) and choose “Index with AnythingLLM LocalDocs” (or similar). This should recursively include all nested files and subfolders.
Automatic Embedding and Vector Storage:
Automatically process supported file formats (PDF, DOCX, TXT, MD, etc.), chunk content, generate embeddings, and store them in a local vector database without extra manual steps.
Integrated Per-Query Retrieval:
At runtime, automatically retrieve the most relevant document snippets from the indexed collection and include them in the prompt context.
Source Transparency:
Show clear source citations (file path and excerpt) for information used in responses.

Benefits:

Significantly simplifies ingestion of large document collections and complex directory structures.
Reduces friction for users with existing folder hierarchies.
Improves the RAG experience with transparent source attribution.

Success Criteria:

Users can select entire folder trees with one action in file explorer.
Indexing runs without repeated manual uploads.
Per-query retrieval is automatic and contextually relevant.
Responses include clear source references.

GPT4All's RAG Interface:

Originally created by @TheAlex25 on GitHub (Feb 13, 2026). ### What would you like to see? ## Feature Request: LocalDocs-Style RAG with One-Click Multi-Folder Indexing in AnythingLLM **Title:** Support GPT4All-style LocalDocs RAG with one-click indexing of multiple folders and all nested directories. **Summary:** Add a LocalDocs-equivalent retrieval-augmented generation (RAG) feature to AnythingLLM that allows users to select one or more folders — including nested subfolders — from the operating system file explorer (Windows Explorer, macOS Finder, Ubuntu Files) and index them for semantic search and contextual retrieval in a single action. This would replace the current manual and clunky ingestion process. **Current Limitation:** AnythingLLM’s existing document ingestion for RAG requires manual uploads and repeated steps to add large collections or complex directory structures. There is no easy way to select entire folder trees at once for indexing. **Proposed Behavior:** 1. **One-Click Folder Selection:** Allow users to right-click one or multiple folders in the OS file explorer (Windows, macOS, Linux) and choose “Index with AnythingLLM LocalDocs” (or similar). This should recursively include all nested files and subfolders. 2. **Automatic Embedding and Vector Storage:** Automatically process supported file formats (PDF, DOCX, TXT, MD, etc.), chunk content, generate embeddings, and store them in a local vector database without extra manual steps. 3. **Integrated Per-Query Retrieval:** At runtime, automatically retrieve the most relevant document snippets from the indexed collection and include them in the prompt context. 4. **Source Transparency:** Show clear source citations (file path and excerpt) for information used in responses. **Benefits:** * Significantly simplifies ingestion of large document collections and complex directory structures. * Reduces friction for users with existing folder hierarchies. * Improves the RAG experience with transparent source attribution. **Success Criteria:** * Users can select entire folder trees with one action in file explorer. * Indexing runs without repeated manual uploads. * Per-query retrieval is automatic and contextually relevant. * Responses include clear source references. **GPT4All's RAG Interface:** <img width="1281" height="1456" alt="Image" src="https://github.com/user-attachments/assets/5d4e4b46-36a3-44f9-a0d7-5aa29d1b7c14" />

deekerman added the

enhancement

feature request

labels

2026-02-28 06:32:33 -05:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starred/anything-llm#3173

No description provided.

Rows
Columns