Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-19 14:13:35 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #5361] [BUG]:Desktop Assistant fails with Error: No embedding base path was set. even when main app chat and embedder config are valid #5042

New Issue
Closed
opened 2026-06-05 14:51:41 -04:00 by yindo · 6 comments
yindo commented 2026-06-05 14:51:41 -04:00
Owner
Copy Link
Copy Source

Originally created by @dennis-linux on GitHub (Apr 6, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5361

How are you running AnythingLLM?

AnythingLLM desktop app

What happened?

Description

On macOS with AnythingLLM Desktop v1.12.0, the main app works normally, but Desktop Assistant fails consistently with:

Error: No embedding base path was set.

This appears to be specific to the Desktop Assistant path, not the general LLM or embedding configuration.

Environment

  • AnythingLLM Desktop: 1.12.0
  • OS: macOS
  • Workspace chat in main app: works
  • Desktop Assistant: fails

What works

Inside the main desktop app, normal chat works correctly in my workspace.

Example:

  • Workspace: DennisLinux
  • Model: deepseek-v3-0324
  • Prompt/response works normally

What fails

Desktop Assistant fails immediately with:

Error: No embedding base path was set.

What I tested

I tested multiple embedding providers:

  1. Native embedder
  • EMBEDDING_ENGINE='native'
  • EMBEDDING_MODEL_PREF='Xenova/all-MiniLM-L6-v2'
  1. LM Studio
  1. OpenAI
  • EMBEDDING_ENGINE='openai'
  • EMBEDDING_MODEL_PREF='text-embedding-3-small'

Desktop Assistant failed in all cases.

Additional debugging

I verified that:

  • the embedder settings are being saved to the desktop app .env
  • main app chat still works
  • clearing Electron UI state did not fix it

I cleared only:

  • ~/Library/Application Support/anythingllm-desktop/Local Storage
  • ~/Library/Application Support/anythingllm-desktop/Session Storage

No change.

Relevant log clues

The backend initializes normal embedders correctly in the main app path, but Desktop Assistant still fails.

I also saw log entries like:

Assistant workspace not found - exiting.

The packaged app code appears to throw "No embedding base path was set." only from API-style embedder classes, not from the native embedder, which
suggests Desktop Assistant may be resolving the wrong embedder path or loading stale/incorrect assistant-specific config.

Expected behavior

Desktop Assistant should either:

  • use the configured system embedder correctly, or
  • show a provider-specific configuration error that matches the actual failing provider

Actual behavior

Desktop Assistant throws:

Error: No embedding base path was set.

even when the configured embedder is valid and the main app works.

Are there known steps to reproduce?

No response

Originally created by @dennis-linux on GitHub (Apr 6, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5361 ### How are you running AnythingLLM? AnythingLLM desktop app ### What happened? ### Description On macOS with AnythingLLM Desktop v1.12.0, the main app works normally, but Desktop Assistant fails consistently with: Error: No embedding base path was set. This appears to be specific to the Desktop Assistant path, not the general LLM or embedding configuration. ### Environment - AnythingLLM Desktop: 1.12.0 - OS: macOS - Workspace chat in main app: works - Desktop Assistant: fails ### What works Inside the main desktop app, normal chat works correctly in my workspace. Example: - Workspace: DennisLinux - Model: deepseek-v3-0324 - Prompt/response works normally ### What fails Desktop Assistant fails immediately with: Error: No embedding base path was set. ### What I tested I tested multiple embedding providers: 1. Native embedder - EMBEDDING_ENGINE='native' - EMBEDDING_MODEL_PREF='Xenova/all-MiniLM-L6-v2' 2. LM Studio - EMBEDDING_ENGINE='lmstudio' - EMBEDDING_BASE_PATH='http://127.0.0.1:1234/v1' 3. OpenAI - EMBEDDING_ENGINE='openai' - EMBEDDING_MODEL_PREF='text-embedding-3-small' Desktop Assistant failed in all cases. ### Additional debugging I verified that: - the embedder settings are being saved to the desktop app .env - main app chat still works - clearing Electron UI state did not fix it I cleared only: - ~/Library/Application Support/anythingllm-desktop/Local Storage - ~/Library/Application Support/anythingllm-desktop/Session Storage No change. ### Relevant log clues The backend initializes normal embedders correctly in the main app path, but Desktop Assistant still fails. I also saw log entries like: Assistant workspace not found - exiting. The packaged app code appears to throw "No embedding base path was set." only from API-style embedder classes, not from the native embedder, which suggests Desktop Assistant may be resolving the wrong embedder path or loading stale/incorrect assistant-specific config. ### Expected behavior Desktop Assistant should either: - use the configured system embedder correctly, or - show a provider-specific configuration error that matches the actual failing provider ### Actual behavior Desktop Assistant throws: Error: No embedding base path was set. even when the configured embedder is valid and the main app works. ### Are there known steps to reproduce? _No response_
yindo added the possible bugneeds info / can't replicate labels 2026-06-05 14:51:41 -04:00
yindo closed this issue 2026-06-05 14:51:41 -04:00
yindo commented 2026-06-05 14:51:42 -04:00
Author
Owner
Copy Link
Copy Source

@dennis-linux commented on GitHub (Apr 6, 2026):

Additional debugging:

I got the MCP server fully connected in Desktop and confirmed AnythingLLM is loading the tool schemas successfully.

The MCP server shows as:

  • Knowledgebase File Server -> On

The backend logs show the server is connected and listTools() returns all tools correctly, including:

  • list_directory
  • read_file
  • write_file

However, actual @agent execution fails before the tool call runs.

Repro:

  1. Open workspace thread
  2. Set model to qwen2-57b-a14b-instruct
  3. Prompt:
    @agent use knowledgebase-file-server:list_directory with path "." and return raw JSON

Result:

  • Agent starts
  • Then fails with:
    Cannot read properties of null (reading 'workspace')

UI output:

  • Agent @agent invoked. Swapping over to agent chat.
  • Could not respond to message.
  • Cannot read properties of null (reading 'workspace')

Important detail:

  • This happens even though the workspace exists locally and normal workspace chat works.
  • MCP discovery works; failure appears to be in the agent/workspace binding before tool execution.
<!-- gh-comment-id:4194659054 --> @dennis-linux commented on GitHub (Apr 6, 2026): Additional debugging: I got the MCP server fully connected in Desktop and confirmed AnythingLLM is loading the tool schemas successfully. The MCP server shows as: - `Knowledgebase File Server` -> `On` The backend logs show the server is connected and `listTools()` returns all tools correctly, including: - `list_directory` - `read_file` - `write_file` However, actual `@agent` execution fails before the tool call runs. Repro: 1. Open workspace thread 2. Set model to `qwen2-57b-a14b-instruct` 3. Prompt: `@agent use knowledgebase-file-server:list_directory with path "." and return raw JSON` Result: - Agent starts - Then fails with: `Cannot read properties of null (reading 'workspace')` UI output: - `Agent @agent invoked. Swapping over to agent chat.` - `Could not respond to message.` - `Cannot read properties of null (reading 'workspace')` Important detail: - This happens even though the workspace exists locally and normal workspace chat works. - MCP discovery works; failure appears to be in the agent/workspace binding before tool execution.
yindo commented 2026-06-05 14:51:42 -04:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Apr 6, 2026):

Can you show a video of this and also what provider are you running deepseek-v3-0324 with? The model is not important in this particular issue.

Additional requests:

  • Can you show the Embedder preference page?
  • For the chats in the assistant panel, are you calling agents or just normal QA chat?
  • Do you have any Documents that you drag/drop into the Assistant Panel before chat or that you have embedded in the Assistant Workspace (if it exists) prior to send a chat?
<!-- gh-comment-id:4194768391 --> @timothycarambat commented on GitHub (Apr 6, 2026): Can you show a video of this and also what **provider** are you running `deepseek-v3-0324` with? The model is not important in this particular issue. Additional requests: - Can you show the `Embedder` preference page? - For the chats in the assistant panel, are you calling agents or just normal QA chat? - Do you have any Documents that you drag/drop into the Assistant Panel before chat or that you have embedded in the Assistant Workspace (if it exists) prior to send a chat?
yindo commented 2026-06-05 14:51:42 -04:00
Author
Owner
Copy Link
Copy Source

@dennis-linux commented on GitHub (Apr 6, 2026):

deepseek-v3-0324 is running local (I have LM Studio and Ollama)
RESPONSE from Codex running local on my Mac Studio:

Answers below.

  1. Video

I can record and upload a short repro video if needed. The repro is:

  • Open Desktop Assistant

  • Type a normal prompt

  • Submit

  • Assistant fails with:

    Error: No embedding base path was set.

  1. Provider for deepseek-v3-0324

That model is running through LM Studio in my setup.

Main app/provider details:

  • LLM provider: lmstudio

  • LM Studio base path: http://127.0.0.1:1234/v1

  • Main app chat works normally with deepseek-v3-0324

  1. Embedder preference page

I tested all of these and the Desktop Assistant still failed:

  • Native

    • EMBEDDING_ENGINE='native'

    • EMBEDDING_MODEL_PREF='Xenova/all-MiniLM-L6-v2'

  • LM Studio

    • EMBEDDING_ENGINE='lmstudio'

    • EMBEDDING_BASE_PATH='http://127.0.0.1:1234/v1'

  • OpenAI

    • EMBEDDING_ENGINE='openai'

    • EMBEDDING_MODEL_PREF='text-embedding-3-small'

The main app persisted those settings correctly, but Desktop Assistant
still failed.

  1. Assistant panel usage

This was normal QA chat, not agents.

  1. Documents in Assistant panel / Assistant workspace

No documents were dragged into the Assistant panel before chat.

I also do not believe there was an Assistant Workspace successfully
created/populated before the error. My local logs included messages like:

  • Assistant workspace not found - exiting.

Additional note:

I also tested MCP separately afterward and confirmed MCP/tooling works in
another client, so the issue still appears specific to the Desktop

Assistant / agent-workspace path in Desktop rather than the underlying
model or embedder configuration.

Yours,
Dennis A. Hunter
NICE Working Group: Cybersecurity Training and Certifications
CySA+, Security+, CompTIA Cloud, & AWS Certified

Mobile: +1-919-633-7090
eMail: @.*** @.***>

Get 30% Off an ITProTV (IT Education) subscription with promo code
DENNISHUNTER2 at http://ssqt.co/m5fu6UO

View my Public Profile at:
https://www.linkedin.com/in/dennislinux

View my certs at:
https://www.youracclaim.com/badges/67006b4f-0e04-4d86-ac7f-5a0bdbcc7704/public_url

Things to think about:

  1. Any sufficiently advanced technology is indistinguishable from magic
  2. Never attribute to malice something adequately explained by a
    misunderstanding.
  3. If you THINK it's simple, you truly do not understand. But if you can
    EXPLAIN it simply, truly you understand...

On Mon, Apr 6, 2026 at 4:20 PM Timothy Carambat @.***>
wrote:

timothycarambat left a comment (Mintplex-Labs/anything-llm#5361)
https://github.com/Mintplex-Labs/anything-llm/issues/5361#issuecomment-4194768391

Can you show a video of this and also what provider are you running
deepseek-v3-0324 with? The model is not important in this particular
issue.

Additional requests:

  • Can you show the Embedder preference page?
  • For the chats in the assistant panel, are you calling agents or just
    normal QA chat?
  • Do you have any Documents that you drag/drop into the Assistant
    Panel before chat or that you have embedded in the Assistant Workspace (if
    it exists) prior to send a chat?

—
Reply to this email directly, view it on GitHub
https://github.com/Mintplex-Labs/anything-llm/issues/5361#issuecomment-4194768391,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AD32M6IKQKG6UD7YDZETXCT4UQGRTAVCNFSM6AAAAACXOJB5Q6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCOJUG43DQMZZGE
.
You are receiving this because you authored the thread.Message ID:
@.***>

<!-- gh-comment-id:4195501184 --> @dennis-linux commented on GitHub (Apr 6, 2026): deepseek-v3-0324 is running local (I have LM Studio and Ollama) RESPONSE from Codex running local on my Mac Studio: Answers below. 1. Video I can record and upload a short repro video if needed. The repro is: - Open Desktop Assistant - Type a normal prompt - Submit - Assistant fails with: `Error: No embedding base path was set.` 2. Provider for `deepseek-v3-0324` That model is running through **LM Studio** in my setup. Main app/provider details: - LLM provider: `lmstudio` - LM Studio base path: `http://127.0.0.1:1234/v1` - Main app chat works normally with `deepseek-v3-0324` 3. Embedder preference page I tested all of these and the Desktop Assistant still failed: - **Native** - `EMBEDDING_ENGINE='native'` - `EMBEDDING_MODEL_PREF='Xenova/all-MiniLM-L6-v2'` - **LM Studio** - `EMBEDDING_ENGINE='lmstudio'` - `EMBEDDING_BASE_PATH='http://127.0.0.1:1234/v1'` - **OpenAI** - `EMBEDDING_ENGINE='openai'` - `EMBEDDING_MODEL_PREF='text-embedding-3-small'` The main app persisted those settings correctly, but Desktop Assistant still failed. 4. Assistant panel usage This was **normal QA chat**, not agents. 5. Documents in Assistant panel / Assistant workspace No documents were dragged into the Assistant panel before chat. I also do not believe there was an Assistant Workspace successfully created/populated before the error. My local logs included messages like: - `Assistant workspace not found - exiting.` Additional note: I also tested MCP separately afterward and confirmed MCP/tooling works in another client, so the issue still appears specific to the Desktop Assistant / agent-workspace path in Desktop rather than the underlying model or embedder configuration. Yours, Dennis A. Hunter NICE Working Group: Cybersecurity Training and Certifications CySA+, Security+, CompTIA Cloud, & AWS Certified Mobile: *+1-919-633-7090* eMail: ***@***.*** ***@***.***> Get 30% Off an ITProTV (IT Education) subscription with promo code *DENNISHUNTER2* at http://ssqt.co/m5fu6UO *View my Public Profile at:* https://www.linkedin.com/in/dennislinux *View my certs at:* https://www.youracclaim.com/badges/67006b4f-0e04-4d86-ac7f-5a0bdbcc7704/public_url Things to think about: 1. Any sufficiently advanced technology is indistinguishable from magic 2. Never attribute to malice something adequately explained by a misunderstanding. 3. If you THINK it's simple, you truly do not understand. But if you can EXPLAIN it simply, truly you understand... On Mon, Apr 6, 2026 at 4:20 PM Timothy Carambat ***@***.***> wrote: > *timothycarambat* left a comment (Mintplex-Labs/anything-llm#5361) > <https://github.com/Mintplex-Labs/anything-llm/issues/5361#issuecomment-4194768391> > > Can you show a video of this and also what *provider* are you running > deepseek-v3-0324 with? The model is not important in this particular > issue. > > Additional requests: > > - Can you show the Embedder preference page? > - For the chats in the assistant panel, are you calling agents or just > normal QA chat? > - Do you have any Documents that you drag/drop into the Assistant > Panel before chat or that you have embedded in the Assistant Workspace (if > it exists) prior to send a chat? > > — > Reply to this email directly, view it on GitHub > <https://github.com/Mintplex-Labs/anything-llm/issues/5361#issuecomment-4194768391>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AD32M6IKQKG6UD7YDZETXCT4UQGRTAVCNFSM6AAAAACXOJB5Q6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCOJUG43DQMZZGE> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> >
yindo commented 2026-06-05 14:51:42 -04:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Apr 10, 2026):

Somewhere in your ENV that you are editing is likely two or more EMBEDDING_ENGINE keys and something is going on where it is loading the "last seen" value and then failing for LMStudio validation as that is the only provider for embedding you listed that could throw such an error.

Valid LMStudio embedding settings in ENV

# EMBEDDING_ENGINE='lmstudio'
# EMBEDDING_BASE_PATH='https://host.docker.internal:1234/v1'
# EMBEDDING_MODEL_PREF='nomic-ai/nomic-embed-text-v1.5-GGUF/nomic-embed-text-v1.5.Q4_0.gguf'
# EMBEDDING_MODEL_MAX_CHUNK_LENGTH=8192
<!-- gh-comment-id:4226682126 --> @timothycarambat commented on GitHub (Apr 10, 2026): Somewhere in your ENV that you are editing is likely **two or more** `EMBEDDING_ENGINE` keys and something is going on where it is loading the "last seen" value and then failing for LMStudio validation as that is the only provider for embedding you listed that could throw such an error. Valid LMStudio embedding settings in ENV ``` # EMBEDDING_ENGINE='lmstudio' # EMBEDDING_BASE_PATH='https://host.docker.internal:1234/v1' # EMBEDDING_MODEL_PREF='nomic-ai/nomic-embed-text-v1.5-GGUF/nomic-embed-text-v1.5.Q4_0.gguf' # EMBEDDING_MODEL_MAX_CHUNK_LENGTH=8192 ```
yindo commented 2026-06-05 14:51:42 -04:00
Author
Owner
Copy Link
Copy Source

@dennis-linux commented on GitHub (Apr 10, 2026):

Issue is RESOLVED. Thank you for your prompt responses. I am bragging about your product to my 12K followers on LinkedIn

<!-- gh-comment-id:4226703115 --> @dennis-linux commented on GitHub (Apr 10, 2026): Issue is RESOLVED. Thank you for your prompt responses. I am bragging about your product to my 12K followers on LinkedIn
yindo commented 2026-06-05 14:51:43 -04:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Apr 10, 2026):

@dennis-linux !! Thank you for the quick response. You have no idea how much time I have been spending coming back to this issue to try and figure out what in the world could possibly be wrong with your setup. The detailed comment helped a ton.

Glad it was an easy fix! Thank you for the shoutouts!

<!-- gh-comment-id:4228080928 --> @timothycarambat commented on GitHub (Apr 10, 2026): @dennis-linux !! Thank you for the quick response. You have no idea how much time I have been spending coming back to this issue to try and figure out what in the world could possibly be wrong with your setup. The detailed comment helped a ton. Glad it was an easy fix! Thank you for the shoutouts!
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 needs info / can't replicate possible bug
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#5042
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?