Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-19 22:23:50 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #2210] [QUESTION]: Intermittent but frequent failures with the SQLConnector #1445

New Issue
Closed
opened 2026-02-22 18:24:50 -05:00 by yindo · 4 comments
yindo commented 2026-02-22 18:24:50 -05:00
Owner
Copy Link
Copy Source

Originally created by @KristianStarr on GitHub (Sep 2, 2024).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/2210

How are you running AnythingLLM?

Linux Desktop Client (local)

What happened?

I have a postgres connection configured with AnythingLLM. In this postgres database is a schema called RAG1 and a test table called MYTABLE. The AnythingLLM client is the linux version and is linked to ChatGPT 4o. All API tokens are working correctly and there is enough credit for 100x the amount of queries.

The table is correctly setup and can be queried using SQL directly using PSQL.
The table can also be queried successfully directly using AnythingLLM. However 50% of the time AnythingLLM seems to lose connection to this table for awhile, which then comes back.

example transcript to illustrate the issue is shown below

+-----------------------------------------------------------------------------------------------------------------------------------------------+
Q. what is the earliest date in mytable?

The earliest date recorded in the mytable table is 01-JAN-2020.
If you have any further questions or need assistance with anything else, feel free to let me know!

Q. remember to always interrogate the mytable table using the agent when asked a question about it, don’t try and answer using data in memory

@agent: I saved the content to long-term memory in this workspaces vector database.
I’ve noted that whenever questions are asked about the mytable table, I should interrogate the table using the agent instead of relying on data stored in memory.
If you have any further instructions or questions, feel free to let me know!

Q. how many rows are there in the mytable table
@agent: Im going to run a query on the RAGTest to get an answer.
Running SQL: SELECT COUNT(*) FROM "RAG1"."MYTABLE";
The mytable table contains a total of 12,446,543 rows.
If you have any more questions or need further assistance, feel free to ask!

Q. how many years are in the mytable table
It appears that I am currently facing persistent issues with establishing a connection to the RAG1 schema, which is preventing me from running any SQL queries.
To determine how many years are represented in the mytable table, you would typically use a query to extract distinct years from the date column. If there is a different way you would like to proceed or if you have additional instructions, please let me know!

Q. how many distinct years exist in the “RAG1”.“MYTABLE” table?
It seems that there is a persistent issue with connecting to the RAG1 schema, which is preventing me from executing any SQL queries to determine the number of distinct years in the MYTABLE table.
If there’s an alternative way to retrieve this information or if you have other questions, please let me know how I can assist you further!

I am currently unable to perform queries on the RAG1 schema to retrieve the sum of all sales for each year in the mytable table.
If you can provide an alternative method to access this information or if there’s anything else I can assist you with, please let me know!

Q. @agent repeat last question against the mytable table
Agent @agent invoked.
Swapping over to agent chat. Type /exit to exit agent execution loop early.
@agent: Checking what are the available databases.
@agent: Im going to run a query on the RAGTest to get an answer.
Running SQL: SELECT EXTRACT(YEAR FROM per_date) AS sales_year, SUM(ors_sale_net) AS total_sales FROM RAG1.MYTABLE GROUP BY sales_year ORDER BY sales_year;
Error: relation "rag1.MYTABLE" does not exist
@agent: Checking what are the available databases.
@agent: Im going to run a query on the RAGTest to get an answer.
Running SQL: SELECT EXTRACT(YEAR FROM per_date) AS year, SUM(ors_sale_net) AS total_sales FROM MYTABLE GROUP BY year ORDER BY year;
Error: relation "MYTABLE" does not exist
It appears there are ongoing connection and relation issues with the RAG1 schema and the MYTABLE table. The queries are resulting in errors indicating that the table does not exist.
If you have another schema, database, or method to access the information, or if you have any additional instructions, please let me know!

Are there known steps to reproduce?

repeat a successful query against the "RAG1"."MYTABLE" table, it may fail

Originally created by @KristianStarr on GitHub (Sep 2, 2024). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/2210 ### How are you running AnythingLLM? Linux Desktop Client (local) ### What happened? I have a postgres connection configured with AnythingLLM. In this postgres database is a schema called RAG1 and a test table called MYTABLE. The AnythingLLM client is the linux version and is linked to ChatGPT 4o. All API tokens are working correctly and there is enough credit for 100x the amount of queries. The table is correctly setup and can be queried using SQL directly using PSQL. The table can also be queried successfully directly using AnythingLLM. However 50% of the time AnythingLLM seems to lose connection to this table for awhile, which then comes back. example transcript to illustrate the issue is shown below +-----------------------------------------------------------------------------------------------------------------------------------------------+ Q. what is the earliest date in mytable? The earliest date recorded in the mytable table is 01-JAN-2020. If you have any further questions or need assistance with anything else, feel free to let me know! Q. remember to always interrogate the mytable table using the agent when asked a question about it, don’t try and answer using data in memory @agent: I saved the content to long-term memory in this workspaces vector database. I’ve noted that whenever questions are asked about the mytable table, I should interrogate the table using the agent instead of relying on data stored in memory. If you have any further instructions or questions, feel free to let me know! Q. how many rows are there in the mytable table @agent: Im going to run a query on the RAGTest to get an answer. Running SQL: SELECT COUNT(*) FROM "RAG1"."MYTABLE"; The mytable table contains a total of 12,446,543 rows. If you have any more questions or need further assistance, feel free to ask! Q. how many years are in the mytable table It appears that I am currently facing persistent issues with establishing a connection to the RAG1 schema, which is preventing me from running any SQL queries. To determine how many years are represented in the mytable table, you would typically use a query to extract distinct years from the date column. If there is a different way you would like to proceed or if you have additional instructions, please let me know! Q. how many distinct years exist in the “RAG1”.“MYTABLE” table? It seems that there is a persistent issue with connecting to the RAG1 schema, which is preventing me from executing any SQL queries to determine the number of distinct years in the MYTABLE table. If there’s an alternative way to retrieve this information or if you have other questions, please let me know how I can assist you further! I am currently unable to perform queries on the RAG1 schema to retrieve the sum of all sales for each year in the mytable table. If you can provide an alternative method to access this information or if there’s anything else I can assist you with, please let me know! Q. @agent repeat last question against the mytable table Agent @agent invoked. Swapping over to agent chat. Type /exit to exit agent execution loop early. @agent: Checking what are the available databases. @agent: Im going to run a query on the RAGTest to get an answer. Running SQL: SELECT EXTRACT(YEAR FROM per_date) AS sales_year, SUM(ors_sale_net) AS total_sales FROM RAG1.MYTABLE GROUP BY sales_year ORDER BY sales_year; Error: relation "rag1.MYTABLE" does not exist @agent: Checking what are the available databases. @agent: Im going to run a query on the RAGTest to get an answer. Running SQL: SELECT EXTRACT(YEAR FROM per_date) AS year, SUM(ors_sale_net) AS total_sales FROM MYTABLE GROUP BY year ORDER BY year; Error: relation "MYTABLE" does not exist It appears there are ongoing connection and relation issues with the RAG1 schema and the MYTABLE table. The queries are resulting in errors indicating that the table does not exist. If you have another schema, database, or method to access the information, or if you have any additional instructions, please let me know! ### Are there known steps to reproduce? repeat a successful query against the "RAG1"."MYTABLE" table, it may fail
yindo added the question label 2026-02-22 18:24:50 -05:00
yindo closed this issue 2026-02-22 18:24:50 -05:00
yindo commented 2026-02-22 18:24:50 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Sep 3, 2024):

Assuming that there could be hallucinations from tools calls occurring, what kind of connector is this and do you have multiple DB connections set up at once? Every connection should be disconnected and closed after use so the connection is cleaned up.

Also logs from the app during these failures would be more verbose than what is reported in the frontend

@timothycarambat commented on GitHub (Sep 3, 2024): Assuming that there could be hallucinations from [tools calls](https://docs.anythingllm.com/agent-not-using-tools) occurring, what kind of connector is this and do you have multiple DB connections set up at once? Every connection should be disconnected and closed after use so the connection is cleaned up. Also logs from the app during these failures would be more verbose than what is reported in the frontend
yindo commented 2026-02-22 18:24:50 -05:00
Author
Owner
Copy Link
Copy Source

@KristianStarr commented on GitHub (Sep 5, 2024):

There are no multiple connections setup. It is a single connection to postgres using the in-built SQLConnector in AnythingLLM for Postgres.

Where would I find detailed logs?

It gets worse too - regardless of how many times I tell it to use the correct schema, it seems to quite often go into a loop trying to discover tables in the public schema - where in this case there are none.
e.g.

Agent @agent invoked.
Swapping over to agent chat. Type /exit to exit agent execution loop early.
@agent: Checking what are the available databases.
@agent: Checking what are the available tables in the RAGTest database.
Running SQL: SELECT * FROM pg_catalog.pg_tables WHERE schemaname = 'public'
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available tables in the RAGTest database.
Running SQL: SELECT * FROM pg_catalog.pg_tables WHERE schemaname = 'public'
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@agent: Checking what are the available databases.
@KristianStarr commented on GitHub (Sep 5, 2024): There are no multiple connections setup. It is a single connection to postgres using the in-built SQLConnector in AnythingLLM for Postgres. Where would I find detailed logs? It gets worse too - regardless of how many times I tell it to use the correct schema, it seems to quite often go into a loop trying to discover tables in the public schema - where in this case there are none. e.g. ``` Agent @agent invoked. Swapping over to agent chat. Type /exit to exit agent execution loop early. @agent: Checking what are the available databases. @agent: Checking what are the available tables in the RAGTest database. Running SQL: SELECT * FROM pg_catalog.pg_tables WHERE schemaname = 'public' @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available tables in the RAGTest database. Running SQL: SELECT * FROM pg_catalog.pg_tables WHERE schemaname = 'public' @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available databases. @agent: Checking what are the available databases. ```
yindo commented 2026-02-22 18:24:50 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Sep 5, 2024):

This for sure is the model acting bizzarly then, what model are you on? I would say cloud models do much better here, and on the whole they do - but I have seen looping like this with gpt-4o-mini, but not just gpt-4o. If running a local model, which would make even more sense - what quantization?

@timothycarambat commented on GitHub (Sep 5, 2024): This for sure is the model acting bizzarly then, what model are you on? I would say cloud models do much better here, and on the whole they do - but I have seen looping like this with `gpt-4o-mini`, but not just `gpt-4o`. If running a local model, which would make even more sense - what quantization?
yindo commented 2026-02-22 18:24:50 -05:00
Author
Owner
Copy Link
Copy Source

@KristianStarr commented on GitHub (Sep 6, 2024):

Not a local model, as mentioned it's ChatGpt - gpt-4o-mini. The only thing local is the postgres db.

The looping almost seems to me as if the agent is always defaulting to the public schema and it just confuses itself

@KristianStarr commented on GitHub (Sep 6, 2024): Not a local model, as mentioned it's ChatGpt - gpt-4o-mini. The only thing local is the postgres db. The looping almost seems to me as if the agent is always defaulting to the public schema and it just confuses itself
yindo changed title from [QUESTION]: Intermittent but frequent failures with the SQLConnector to [GH-ISSUE #2210] [QUESTION]: Intermittent but frequent failures with the SQLConnector 2026-06-05 14:40:48 -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 question
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#1445
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?