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 #1998] [BUG]: sql request sent 2 times #1302

New Issue
Closed
opened 2026-02-22 18:24:10 -05:00 by yindo · 1 comment
yindo commented 2026-02-22 18:24:10 -05:00
Owner
Copy Link
Copy Source

Originally created by @nicho2 on GitHub (Jul 30, 2024).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/1998

How are you running AnythingLLM?

Docker (local)

What happened?

I use lmstudio and i see at each time i request a sql command with agent skill, the request is sending two times to the llm and to the database.

I have two times the request with : content": "You are a program which picks the most optimal function and parameters to call....

To avoid that, i have modify in lmstudio.js the "complete" function

 async complete(messages, functions = null) {
    try {
      let completion;
      if (functions.length > 0) {
        const { toolCall, text } = await this.functionCall(
          messages,
          functions,
          this.#handleFunctionCallChat.bind(this)
        );

in

  async complete(messages, functions = null) {
    try {
      let completion;
      if (functions.length > 0 && messages.slice(-1)[0].role !== "function") {
        const { toolCall, text } = await this.functionCall(
          messages,
          functions,
          this.#handleFunctionCallChat.bind(this)
        );

to avoid to recall the functionCall if the last message is the answer of the database request

but i'm not sure of my modification , but it's works

Are there known steps to reproduce?

No response

Originally created by @nicho2 on GitHub (Jul 30, 2024). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/1998 ### How are you running AnythingLLM? Docker (local) ### What happened? I use lmstudio and i see at each time i request a sql command with agent skill, the request is sending two times to the llm and to the database. I have two times the request with : content": "You are a program which picks the most optimal function and parameters to call.... To avoid that, i have modify in lmstudio.js the "complete" function async complete(messages, functions = null) { try { let completion; if (functions.length > 0) { const { toolCall, text } = await this.functionCall( messages, functions, this.#handleFunctionCallChat.bind(this) ); in async complete(messages, functions = null) { try { let completion; if (functions.length > 0 && messages.slice(-1)[0].role !== "function") { const { toolCall, text } = await this.functionCall( messages, functions, this.#handleFunctionCallChat.bind(this) ); to avoid to recall the functionCall if the last message is the answer of the database request but i'm not sure of my modification , but it's works ### Are there known steps to reproduce? _No response_
yindo added the possible bug label 2026-02-22 18:24:10 -05:00
yindo closed this issue 2026-02-22 18:24:11 -05:00
yindo commented 2026-02-22 18:24:11 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Jul 30, 2024):

This will actually break a lot more function calls than you expect. For the OSS models we have function call throttling on every function that we think could be expensive, otherwise it's not a bug for an LLM to make the same tool call twice as some of them will do this for whatever reason to reach a text response.

I am not sure specific tool was being called, but if it is a low overhead one like sql-list-tables that is pretty inconsequential so its permissible especially since its not out of the realm of possibilities an agent needs to list the tables twice for a single action.

There is a call we have on mostly every skill called Deduplicator and that basically blocks OSS from infinite recursion of function calls via either one-time, a JSON signature that prevents functions with parameters from running more than once, and also a time-based cooldown. Given the variance between every OSS model and their ability to use tools and their outputs this is how we can prevent the duplication. An LLM calling a tool twice is not the end of the world, but calling it infinitely is.

Given the patch above, it would work in the case of general SQL queries but would fail on any call more than 2 steps long.

@timothycarambat commented on GitHub (Jul 30, 2024): This will actually break a lot more function calls than you expect. For the OSS models we have function call throttling on every function that we think could be expensive, otherwise it's not a bug for an LLM to make the same tool call twice as some of them will do this for whatever reason to reach a text response. I am not sure specific tool was being called, but if it is a low overhead one like _sql-list-tables_ that is pretty inconsequential so its permissible especially since its not out of the realm of possibilities an agent needs to list the tables twice for a single action. There is a call we have on mostly every skill called `Deduplicator` and that basically blocks OSS from infinite recursion of function calls via either one-time, a JSON signature that prevents functions with parameters from running more than once, and also a time-based cooldown. Given the variance between every OSS model and their ability to use tools and their outputs this is how we can prevent the duplication. An LLM calling a tool twice is not the end of the world, but calling it infinitely is. Given the patch above, it would work in the case of general SQL queries but would fail on any call more than 2 steps long.
yindo changed title from [BUG]: sql request sent 2 times to [GH-ISSUE #1998] [BUG]: sql request sent 2 times 2026-06-05 14:40:00 -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 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#1302
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?