Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-21 01:25:28 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #4204] [FEAT]: Add a "Deep Research" mode with web search #2676

New Issue
Open
opened 2026-02-22 18:30:44 -05:00 by yindo · 5 comments
yindo commented 2026-02-22 18:30:44 -05:00
Owner
Copy Link
Copy Source

Originally created by @nikhilprasanth on GitHub (Jul 26, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4204

What would you like to see?

It would be great to have a research mode that can search the web.
I could ask a complex question, and AnythingLLM would find the best sources online and write a summary with links. This would save me the huge step of manually finding, saving, and uploading articles.
Ideally, it could even add the web sources it found directly into the workspace, so I can keep chatting with them. This would make it a complete, all-in-one research tool.

Originally created by @nikhilprasanth on GitHub (Jul 26, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4204 ### What would you like to see? It would be great to have a research mode that can search the web. I could ask a complex question, and AnythingLLM would find the best sources online and write a summary with links. This would save me the huge step of manually finding, saving, and uploading articles. Ideally, it could even add the web sources it found directly into the workspace, so I can keep chatting with them. This would make it a complete, all-in-one research tool.
yindo added the enhancementfeature request labels 2026-02-22 18:30:44 -05:00
yindo commented 2026-02-22 18:30:44 -05:00
Author
Owner
Copy Link
Copy Source

@Saswatsusmoy commented on GitHub (Jul 29, 2025):

Just outlining the Research Mode workflow I'm thinking. Would like to have some feedbacks before I start implementing parts of it and testing the implementations

-> Web Search: Hit up existing web search to find the best online sources.
-> Scrape & Process: Use the Collector API to grab content from the top 'N' results, then run it through current standard document pipeline.
-> Workspace Update: Automatically add and embed these new sources into the user's current workspace.
-> Synthesize Answer: Finally, generate an answer using these newly added documents as core context.

On the backend, I'll be chaining together web search -> scrape -> embed -> chat. And for the frontend, I'll implement a "Research Mode" selector, a settings panel for user preferences (like how many sources to fetch), and real-time progress updates for the user ("Searching...", "Analyzing sources...", etc.).

@Saswatsusmoy commented on GitHub (Jul 29, 2025): Just outlining the Research Mode workflow I'm thinking. Would like to have some feedbacks before I start implementing parts of it and testing the implementations -> Web Search: Hit up existing web search to find the best online sources. -> Scrape & Process: Use the Collector API to grab content from the top 'N' results, then run it through current standard document pipeline. -> Workspace Update: Automatically add and embed these new sources into the user's current workspace. -> Synthesize Answer: Finally, generate an answer using these newly added documents as core context. On the backend, I'll be chaining together web search -> scrape -> embed -> chat. And for the frontend, I'll implement a "Research Mode" selector, a settings panel for user preferences (like how many sources to fetch), and real-time progress updates for the user ("Searching...", "Analyzing sources...", etc.).
yindo commented 2026-02-22 18:30:45 -05:00
Author
Owner
Copy Link
Copy Source

@nikhilprasanth commented on GitHub (Jul 29, 2025):

This looks like a really solid plan. I like how everything flows — from searching to scraping to embedding, and finally generating the answer. The idea of plugging new sources straight into the user’s workspace is especially nice; keeps things feeling seamless.

For the settings, maybe also let users pick between “fast” and “deep” modes? Like, quick summary vs. more thorough digging.

@nikhilprasanth commented on GitHub (Jul 29, 2025): This looks like a really solid plan. I like how everything flows — from searching to scraping to embedding, and finally generating the answer. The idea of plugging new sources straight into the user’s workspace is especially nice; keeps things feeling seamless. For the settings, maybe also let users pick between “fast” and “deep” modes? Like, quick summary vs. more thorough digging.
yindo commented 2026-02-22 18:30:45 -05:00
Author
Owner
Copy Link
Copy Source

@Saswatsusmoy commented on GitHub (Jul 29, 2025):

Instead of giving users option of "fast" and "deep" we can just provide them with an option "Deep Research". This shall be self explanatory for users that "Deep Research" mode does thorough digging and the only chat features does the "fast" quick summary of the prompts and search. As web search is already available in the codebase through plugins and other implementations, we can easily implement both of them natively to the code without changing much of the existing logic

@Saswatsusmoy commented on GitHub (Jul 29, 2025): Instead of giving users option of "fast" and "deep" we can just provide them with an option "Deep Research". This shall be self explanatory for users that "Deep Research" mode does thorough digging and the only chat features does the "fast" quick summary of the prompts and search. As web search is already available in the codebase through plugins and other implementations, we can easily implement both of them natively to the code without changing much of the existing logic
yindo commented 2026-02-22 18:30:45 -05:00
Author
Owner
Copy Link
Copy Source

@nikhilprasanth commented on GitHub (Jul 30, 2025):

Yeah, totally agree with that. “Deep Research” sounds clear and user-friendly, it tells you exactly what to expect without needing extra explanation. And keeping the fast summary as the default chat experience works perfectly. Since web search is already in place, we can easily set this up without touching too much of the existing logic.

@nikhilprasanth commented on GitHub (Jul 30, 2025): Yeah, totally agree with that. “Deep Research” sounds clear and user-friendly, it tells you exactly what to expect without needing extra explanation. And keeping the fast summary as the default chat experience works perfectly. Since web search is already in place, we can easily set this up without touching too much of the existing logic.
yindo commented 2026-02-22 18:30:45 -05:00
Author
Owner
Copy Link
Copy Source

@pratik-narain commented on GitHub (Oct 3, 2025):

Any progress on this? I have been able to perform sort of a research using a slash command with a detailed prompt, but having a perplexity alternative would be awesome.

@pratik-narain commented on GitHub (Oct 3, 2025): Any progress on this? I have been able to perform sort of a research using a slash command with a detailed prompt, but having a perplexity alternative would be awesome.
yindo changed title from [FEAT]: Add a "Deep Research" mode with web search to [GH-ISSUE #4204] [FEAT]: Add a "Deep Research" mode with web search 2026-06-05 14:47:54 -04:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Desktop
Docker
Integration Request
Integration Request
OS: Linux
OS: Mobile
OS: Windows
UI/UX
blocked
bug
bug
core-team-only
documentation
duplicate
embed-widget
enhancement
feature request
github_actions
good first issue
investigating
needs info / can't replicate
possible bug
pull-request
Mirrored from GitHub Pull Request
 question
stage: specifications
wontfix
No Label enhancement feature request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2676
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?