langchain-ai/langgraph
1
0
Fork 0
mirror of https://github.com/langchain-ai/langgraph.git synced 2026-07-15 12:35:43 -04:00
Code Issues 314 Packages Projects Releases 17 Wiki Activity

LangGraph: Checkpointer causes MySQL collation errors (1267/1366) + Hindi TTS loses Devanagari matras #937

New Issue
Closed
opened 2026-02-20 17:42:26 -05:00 by yindo · 2 comments
yindo commented 2026-02-20 17:42:26 -05:00
Owner
Copy Link
Copy Source

Originally created by @shubhammalik19 on GitHub (Aug 25, 2025).

Checked other resources

  • This is a bug, not a usage question. For questions, please use the LangChain Forum (https://forum.langchain.com/).
  • I added a clear and detailed title that summarizes the issue.
  • I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example).
  • I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue.

Example Code

# Minimal repro: create schema, then simulate the checkpointer using a different
# session collation (utf8mb4_general_ci) which causes MySQL 1267 on comparisons.
# Run with env vars: MYSQL_HOST, MYSQL_USER, MYSQL_PASSWORD, MYSQL_PORT

import os, pymysql, traceback
from langgraph.graph import StateGraph
from langgraph.checkpoint.mysql.pymysql import PyMySQLSaver

HOST = os.getenv("MYSQL_HOST", "127.0.0.1")
USER = os.getenv("MYSQL_USER", "root")
PWD  = os.getenv("MYSQL_PASSWORD", "")
PORT = int(os.getenv("MYSQL_PORT", "3306"))

def create_schema():
    conn = pymysql.connect(host=HOST, user=USER, password=PWD, port=PORT,
                           charset="utf8mb4", autocommit=True, cursorclass=pymysql.cursors.DictCursor)
    with conn.cursor() as cur:
        cur.execute("DROP DATABASE IF EXISTS langgraph_repro")
        cur.execute("CREATE DATABASE langgraph_repro CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci")
        cur.execute("USE langgraph_repro")
        # users uses unicode_ci
        cur.execute("CREATE TABLE users (id INT AUTO_INCREMENT PRIMARY KEY, username VARCHAR(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci)")
        # chat_sessions intentionally uses general_ci (mismatch)
        cur.execute("CREATE TABLE chat_sessions (id INT AUTO_INCREMENT PRIMARY KEY, session_title VARCHAR(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci, user_id INT)")
        cur.execute(\"INSERT INTO users (username) VALUES ('testuser')\")
        cur.execute(\"INSERT INTO chat_sessions (session_title, user_id) VALUES ('testuser', 1)\")
    conn.close()

def simulate_checkpointer_and_run_join():
    # This connection simulates the checkpointer: set its session collation to general_ci
    conn_cp = pymysql.connect(host=HOST, user=USER, password=PWD, port=PORT,
                              charset="utf8mb4", autocommit=True, cursorclass=pymysql.cursors.DictCursor)
    try:
        with conn_cp.cursor() as cur:
            cur.execute("USE langgraph_repro")
            # set session to a different collation (this simulates the problematic checkpointer)
            cur.execute("SET NAMES utf8mb4 COLLATE utf8mb4_general_ci")
            cur.execute("SET collation_connection = 'utf8mb4_general_ci'")
            # The next statement will raise 1267 if collations are mixed
            cur.execute("SELECT cs.id, cs.session_title, u.username FROM chat_sessions cs JOIN users u ON cs.session_title = u.username")
            rows = cur.fetchall()
            print("Unexpected: query returned rows:", rows)
    except Exception as e:
        # This is the reproduction: OperationalError / 1267 shown here
        print("Expected error from simulated checkpointer (repro):")
        traceback.print_exc()
    finally:
        conn_cp.close()

if __name__ == "__main__":
    create_schema()
    simulate_checkpointer_and_run_join()

# Optional short LangGraph snippet to show how a checkpointer would be attached:
#
# conn_cp = <PyMySQL connection as above>
# saver = PyMySQLSaver(conn_cp)
# g = StateGraph(dict)                       # minimal graph type
# # add nodes and compile with checkpointer -> invoking the graph will surface 1267
# graph = g.compile(checkpointer=saver)
# graph.invoke({...})

Error Message and Stack Trace (if applicable)

Checklist

Provide a concise 1267 (collation mix) traceback to paste — Done
Provide a concise 1366 (incorrect string value) traceback to paste — Done
Short note on where to get the full trace from your app and what else to include — Done
OperationalError: (1267, "Illegal mix of collations (utf8mb4_general_ci,IMPLICIT) and (utf8mb4_unicode_ci,IMPLICIT) for operation '='")
Traceback (most recent call last):
  File "/home/shubham/laravel/newversion/langgraph_issue/reproduce.py", line 78, in run_join_with_session_collation
    cur2.execute("SELECT cs.id, cs.session_title, u.username FROM chat_sessions cs JOIN users u ON cs.session_title = u.username")
  File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/pymysql/cursors.py", line 170, in execute
    result = self._query(query)
  File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/pymysql/cursors.py", line 323, in _query
    conn.query(q)
  File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/pymysql/connections.py", line 588, in query
    raise exc
pymysql.err.OperationalError: (1267, "Illegal mix of collations (utf8mb4_general_ci,IMPLICIT) and (utf8mb4_unicode_ci,IMPLICIT) for operation '='")

Pasteable Error Message & Stack Trace (incorrect string value — 1366)


ProgrammingError: (1366, "Incorrect string value: '\\xE0\\xA4\\x95\\xE0\\xA5\\x8D...' for column `newversion`.`chat_sessions`.`session_title` at row 1")
Traceback (most recent call last):
  File "/path/to/your/app/langgraph_agent.py", line 214, in some_save_path
    cursor.execute(\"INSERT INTO chat_sessions (session_title, user_id) VALUES (%s, %s)\", (session_title, user_id))
  File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/mysql/connector/cursor.py", line 565, in execute
    self._handle_result(self._connection.cmd_stmt_execute(stmt))
  File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/mysql/connector/connection.py", line 1160, in cmd_stmt_execute
    raise errors.get_mysql_exception(err_msg, errno)
mysql.connector.errors.ProgrammingError: 1366 (22007): Incorrect string value: '\\xE0\\xA4\\x95\\xE0\\xA5\\x8D...' for column `newversion`.`chat_sessions`.`session_title` at row 1


Suggested small app log excerpt (helpful to include)

2025-08-25 11:42:29 | ERROR | aiapp.ChatAgent | Graph execution failed: (1267, "Illegal mix of collations (utf8mb4_unicode_ci,IMPLICIT) and (utf8mb4_general_ci,IMPLICIT) for operation '='")
2025-08-25 11:42:29 | ERROR | aiapp.ChatAgent | Collation error detected in graph execution
2025-08-25 11:42:29 | ERROR | aiapp.ChatAgent | Chat session failed for user 2: (1267, "Illegal mix of collations ...")

Description

I’m seeing two related issues when running LangGraph with the MySQL checkpointer in my TMS app:

Summary: backend-generated TTS payloads strip Devanagari combining marks (matras) and the MySQL checkpointer triggers collation/encoding errors (1267 / 1366) when session titles or comparisons include non‑ASCII text.

Symptoms

TTS: voice.plain returned by the backend drops Unicode combining marks in Hindi (Devanagari). UI/display text is correct, but speech output becomes gibberish.
DB: Inserts/queries fail with:
1366: Incorrect string value when inserting Hindi into chat_sessions.session_title
1267: Illegal mix of collations (e.g. utf8mb4_unicode_ci vs utf8mb4_general_ci) when comparing text across tables or sessions
Why I think it’s the checkpointer

The errors occur during graph execution when the checkpointer (a separate DB client/connection) runs queries. I reproduced the problem by opening a second DB connection with a different session collation; that connection produces 1267 while the primary connection does not.
Example log excerpt:
Graph execution failed: (1267, "Illegal mix of collations (utf8mb4_unicode_ci,IMPLICIT) and (utf8mb4_general_ci,IMPLICIT) for operation '='")
How to reproduce (attachments included)

I attached a minimal bundle (langgraph_issue.zip) that contains:
reproduce.sql — creates a DB with mixed collations
reproduce.py — runs the schema and simulates a checkpointer connection that sets utf8mb4_general_ci, reproducing the 1267 error
fix_collation.sql — suggested migration to convert table/column to utf8mb4_unicode_ci
logs.txt and report_email.txt
Run: MYSQL_HOST=... MYSQL_USER=... MYSQL_PASSWORD=... MYSQL_PORT=... python reproduce.py (script prints the 1267 OperationalError when the simulated checkpointer runs a JOIN)
Expected behavior

voice.plain should preserve all Unicode combining marks (Devanagari matras) so TTS produces correct Hindi pronunciation.
The checkpointer should use the same charset/collation as the DB (or explicitly set SET NAMES utf8mb4 COLLATE utf8mb4_unicode_ci) so inserts and text comparisons do not raise 1267/1366.
Notes & environment

MySQL/MariaDB: (attach your exact version)
Connectors tried: mysql-connector-python and PyMySQL (both tested)
App: Python 3.x, LangGraph used as a checkpointer in the graph runtime
I can provide a small respond() snippet showing where voice.plain is built, or add a Docker Compose repro if helpful.

System Info

Plan: I’ll give (A) a short list of commands to run on your machine, (B) a ready-to-paste system-info block with the values I can fill from your workspace and placeholders for the rest, and (C) a short note about where to find LangGraph/LangChain versions and DB details.

Checklist

Provide commands to collect Python, env, package and MySQL versions — Done
Provide a ready-to-paste System Info block with known values filled and placeholders for you to complete — Done
Note about sensitive values and how to attach environment info securely — Done
Commands to run (copy/paste into your shell)

Python runtime

python3 -V

Virtualenv python (use your venv path if different)

/home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -V

pip freeze (in the same env used by the app)

/home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -m pip freeze

Show specific packages (if installed)

/home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -m pip show PyMySQL mysql-connector-python langchain langchain-core langgraph

MySQL / MariaDB server version (run on the DB server)

mysql --version

If langchain_core.sys_info exists:

/home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -m langchain_core.sys_info

Ready-to-paste “System Info *” block (fill placeholders if needed)

OS: Linux (distribution: <fill e.g. Ubuntu 22.04 or Debian 11>)
Kernel: $(uname -r) # run uname -r and paste value

Python:

  • System python: python3 -> <run python3 -V and paste>
  • Virtualenv python: /home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -> 3.13.6 (detected)

Virtualenv path:

  • /home/shubham/laravel/newversion/aiapp/ai_chat_env

Installed packages (selected):

  • PyMySQL: 1.1.0
  • mysql-connector-python: <paste version if installed, else 'not installed'>
  • langchain / langchain-core / langgraph:
  • Other notable packages:

MySQL / MariaDB:

  • Server: MariaDB 10.11.11 (from SQL dump) # run mysql --version to confirm exact
  • Connector(s) used by app: PyMySQL and/or mysql-connector-python (both tested) — specify which is active in your runtime

Database settings / collation overview (paste if available):

  • Database default charset/collation: utf8mb4 / utf8mb4_unicode_ci
  • Affected table(s): chat_sessions (session_title), messages, checkpoints (if used)
  • Note: in my repro the error showed a mix of utf8mb4_general_ci and utf8mb4_unicode_ci — please include result of:
    SHOW CREATE TABLE chat_sessions;
    SHOW TABLE STATUS WHERE Name='chat_sessions';

App details:

  • LangGraph usage: LangGraph Python checkpointer using PyMySQL / PyMySQLSaver (attached in reproduction)
  • App repo / commit:
  • How run: local dev (virtualenv) on Linux; reproduce script ran with:
    MYSQL_HOST=127.0.0.1 MYSQL_USER=root MYSQL_PASSWORD=12345 MYSQL_PORT=3306 python langgraph_issue/reproduce.py

Repro output summary (paste exact):

  • Observed error: OperationalError: (1267, "Illegal mix of collations (utf8mb4_general_ci,IMPLICIT) and (utf8mb4_unicode_ci,IMPLICIT) for operation '='")
  • Observed error (insertion): ProgrammingError: (1366, "Incorrect string value: '\xE0\xA4...' for column newversion.chat_sessions.session_title at row 1")

Additional useful info (optional):

  • Full output of python -m langchain_core.sys_info (if available)
  • DB user/privileges used by the checkpointer (do not paste passwords; paste roles/privs only)
    Short guidance on what to paste into the issue

Paste the filled System Info block above into the GitHub "System Info *" field.
Attach the langgraph_issue.zip bundle and the reproduce.py run output (the console output you saw).
Do not paste passwords or secrets; paste only versions, config, and the exact error text.

langgraph_issue.zip

Originally created by @shubhammalik19 on GitHub (Aug 25, 2025). ### Checked other resources - [x] This is a bug, not a usage question. For questions, please use the LangChain Forum (https://forum.langchain.com/). - [x] I added a clear and detailed title that summarizes the issue. - [x] I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example). - [x] I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue. ### Example Code ```python # Minimal repro: create schema, then simulate the checkpointer using a different # session collation (utf8mb4_general_ci) which causes MySQL 1267 on comparisons. # Run with env vars: MYSQL_HOST, MYSQL_USER, MYSQL_PASSWORD, MYSQL_PORT import os, pymysql, traceback from langgraph.graph import StateGraph from langgraph.checkpoint.mysql.pymysql import PyMySQLSaver HOST = os.getenv("MYSQL_HOST", "127.0.0.1") USER = os.getenv("MYSQL_USER", "root") PWD = os.getenv("MYSQL_PASSWORD", "") PORT = int(os.getenv("MYSQL_PORT", "3306")) def create_schema(): conn = pymysql.connect(host=HOST, user=USER, password=PWD, port=PORT, charset="utf8mb4", autocommit=True, cursorclass=pymysql.cursors.DictCursor) with conn.cursor() as cur: cur.execute("DROP DATABASE IF EXISTS langgraph_repro") cur.execute("CREATE DATABASE langgraph_repro CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci") cur.execute("USE langgraph_repro") # users uses unicode_ci cur.execute("CREATE TABLE users (id INT AUTO_INCREMENT PRIMARY KEY, username VARCHAR(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci)") # chat_sessions intentionally uses general_ci (mismatch) cur.execute("CREATE TABLE chat_sessions (id INT AUTO_INCREMENT PRIMARY KEY, session_title VARCHAR(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci, user_id INT)") cur.execute(\"INSERT INTO users (username) VALUES ('testuser')\") cur.execute(\"INSERT INTO chat_sessions (session_title, user_id) VALUES ('testuser', 1)\") conn.close() def simulate_checkpointer_and_run_join(): # This connection simulates the checkpointer: set its session collation to general_ci conn_cp = pymysql.connect(host=HOST, user=USER, password=PWD, port=PORT, charset="utf8mb4", autocommit=True, cursorclass=pymysql.cursors.DictCursor) try: with conn_cp.cursor() as cur: cur.execute("USE langgraph_repro") # set session to a different collation (this simulates the problematic checkpointer) cur.execute("SET NAMES utf8mb4 COLLATE utf8mb4_general_ci") cur.execute("SET collation_connection = 'utf8mb4_general_ci'") # The next statement will raise 1267 if collations are mixed cur.execute("SELECT cs.id, cs.session_title, u.username FROM chat_sessions cs JOIN users u ON cs.session_title = u.username") rows = cur.fetchall() print("Unexpected: query returned rows:", rows) except Exception as e: # This is the reproduction: OperationalError / 1267 shown here print("Expected error from simulated checkpointer (repro):") traceback.print_exc() finally: conn_cp.close() if __name__ == "__main__": create_schema() simulate_checkpointer_and_run_join() # Optional short LangGraph snippet to show how a checkpointer would be attached: # # conn_cp = <PyMySQL connection as above> # saver = PyMySQLSaver(conn_cp) # g = StateGraph(dict) # minimal graph type # # add nodes and compile with checkpointer -> invoking the graph will surface 1267 # graph = g.compile(checkpointer=saver) # graph.invoke({...}) ``` ### Error Message and Stack Trace (if applicable) ```shell Checklist Provide a concise 1267 (collation mix) traceback to paste — Done Provide a concise 1366 (incorrect string value) traceback to paste — Done Short note on where to get the full trace from your app and what else to include — Done OperationalError: (1267, "Illegal mix of collations (utf8mb4_general_ci,IMPLICIT) and (utf8mb4_unicode_ci,IMPLICIT) for operation '='") Traceback (most recent call last): File "/home/shubham/laravel/newversion/langgraph_issue/reproduce.py", line 78, in run_join_with_session_collation cur2.execute("SELECT cs.id, cs.session_title, u.username FROM chat_sessions cs JOIN users u ON cs.session_title = u.username") File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/pymysql/cursors.py", line 170, in execute result = self._query(query) File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/pymysql/cursors.py", line 323, in _query conn.query(q) File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/pymysql/connections.py", line 588, in query raise exc pymysql.err.OperationalError: (1267, "Illegal mix of collations (utf8mb4_general_ci,IMPLICIT) and (utf8mb4_unicode_ci,IMPLICIT) for operation '='") Pasteable Error Message & Stack Trace (incorrect string value — 1366) ProgrammingError: (1366, "Incorrect string value: '\\xE0\\xA4\\x95\\xE0\\xA5\\x8D...' for column `newversion`.`chat_sessions`.`session_title` at row 1") Traceback (most recent call last): File "/path/to/your/app/langgraph_agent.py", line 214, in some_save_path cursor.execute(\"INSERT INTO chat_sessions (session_title, user_id) VALUES (%s, %s)\", (session_title, user_id)) File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/mysql/connector/cursor.py", line 565, in execute self._handle_result(self._connection.cmd_stmt_execute(stmt)) File "/home/shubham/laravel/newversion/aiapp/ai_chat_env/lib/python3.13/site-packages/mysql/connector/connection.py", line 1160, in cmd_stmt_execute raise errors.get_mysql_exception(err_msg, errno) mysql.connector.errors.ProgrammingError: 1366 (22007): Incorrect string value: '\\xE0\\xA4\\x95\\xE0\\xA5\\x8D...' for column `newversion`.`chat_sessions`.`session_title` at row 1 Suggested small app log excerpt (helpful to include) 2025-08-25 11:42:29 | ERROR | aiapp.ChatAgent | Graph execution failed: (1267, "Illegal mix of collations (utf8mb4_unicode_ci,IMPLICIT) and (utf8mb4_general_ci,IMPLICIT) for operation '='") 2025-08-25 11:42:29 | ERROR | aiapp.ChatAgent | Collation error detected in graph execution 2025-08-25 11:42:29 | ERROR | aiapp.ChatAgent | Chat session failed for user 2: (1267, "Illegal mix of collations ...") ``` ### Description I’m seeing two related issues when running LangGraph with the MySQL checkpointer in my TMS app: Summary: backend-generated TTS payloads strip Devanagari combining marks (matras) and the MySQL checkpointer triggers collation/encoding errors (1267 / 1366) when session titles or comparisons include non‑ASCII text. Symptoms TTS: voice.plain returned by the backend drops Unicode combining marks in Hindi (Devanagari). UI/display text is correct, but speech output becomes gibberish. DB: Inserts/queries fail with: 1366: Incorrect string value when inserting Hindi into chat_sessions.session_title 1267: Illegal mix of collations (e.g. utf8mb4_unicode_ci vs utf8mb4_general_ci) when comparing text across tables or sessions Why I think it’s the checkpointer The errors occur during graph execution when the checkpointer (a separate DB client/connection) runs queries. I reproduced the problem by opening a second DB connection with a different session collation; that connection produces 1267 while the primary connection does not. Example log excerpt: Graph execution failed: (1267, "Illegal mix of collations (utf8mb4_unicode_ci,IMPLICIT) and (utf8mb4_general_ci,IMPLICIT) for operation '='") How to reproduce (attachments included) I attached a minimal bundle (langgraph_issue.zip) that contains: [reproduce.sql](vscode-file://vscode-app/usr/share/code/resources/app/out/vs/code/electron-browser/workbench/workbench.html) — creates a DB with mixed collations [reproduce.py](vscode-file://vscode-app/usr/share/code/resources/app/out/vs/code/electron-browser/workbench/workbench.html) — runs the schema and simulates a checkpointer connection that sets utf8mb4_general_ci, reproducing the 1267 error fix_collation.sql — suggested migration to convert table/column to utf8mb4_unicode_ci logs.txt and report_email.txt Run: MYSQL_HOST=... MYSQL_USER=... MYSQL_PASSWORD=... MYSQL_PORT=... python reproduce.py (script prints the 1267 OperationalError when the simulated checkpointer runs a JOIN) Expected behavior voice.plain should preserve all Unicode combining marks (Devanagari matras) so TTS produces correct Hindi pronunciation. The checkpointer should use the same charset/collation as the DB (or explicitly set SET NAMES utf8mb4 COLLATE utf8mb4_unicode_ci) so inserts and text comparisons do not raise 1267/1366. Notes & environment MySQL/MariaDB: (attach your exact version) Connectors tried: mysql-connector-python and PyMySQL (both tested) App: Python 3.x, LangGraph used as a checkpointer in the graph runtime I can provide a small respond() snippet showing where voice.plain is built, or add a Docker Compose repro if helpful. ### System Info Plan: I’ll give (A) a short list of commands to run on your machine, (B) a ready-to-paste system-info block with the values I can fill from your workspace and placeholders for the rest, and (C) a short note about where to find LangGraph/LangChain versions and DB details. Checklist Provide commands to collect Python, env, package and MySQL versions — Done Provide a ready-to-paste System Info block with known values filled and placeholders for you to complete — Done Note about sensitive values and how to attach environment info securely — Done Commands to run (copy/paste into your shell) # Python runtime python3 -V # Virtualenv python (use your venv path if different) /home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -V # pip freeze (in the same env used by the app) /home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -m pip freeze # Show specific packages (if installed) /home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -m pip show PyMySQL mysql-connector-python langchain langchain-core langgraph # MySQL / MariaDB server version (run on the DB server) mysql --version # If langchain_core.sys_info exists: # /home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -m langchain_core.sys_info Ready-to-paste “System Info *” block (fill placeholders if needed) OS: Linux (distribution: <fill e.g. Ubuntu 22.04 or Debian 11>) Kernel: $(uname -r) # run `uname -r` and paste value Python: - System python: python3 -> <run `python3 -V` and paste> - Virtualenv python: /home/shubham/laravel/newversion/aiapp/ai_chat_env/bin/python -> 3.13.6 (detected) Virtualenv path: - /home/shubham/laravel/newversion/aiapp/ai_chat_env Installed packages (selected): - PyMySQL: 1.1.0 - mysql-connector-python: <paste version if installed, else 'not installed'> - langchain / langchain-core / langgraph: <paste exact versions from pip show or freeze> - Other notable packages: <paste output of pip freeze or list> MySQL / MariaDB: - Server: MariaDB 10.11.11 (from SQL dump) # run `mysql --version` to confirm exact - Connector(s) used by app: PyMySQL and/or mysql-connector-python (both tested) — specify which is active in your runtime Database settings / collation overview (paste if available): - Database default charset/collation: utf8mb4 / utf8mb4_unicode_ci - Affected table(s): `chat_sessions` (session_title), `messages`, `checkpoints` (if used) - Note: in my repro the error showed a mix of `utf8mb4_general_ci` and `utf8mb4_unicode_ci` — please include result of: SHOW CREATE TABLE chat_sessions; SHOW TABLE STATUS WHERE Name='chat_sessions'; App details: - LangGraph usage: LangGraph Python checkpointer using PyMySQL / PyMySQLSaver (attached in reproduction) - App repo / commit: <paste commit hash or branch if relevant> - How run: local dev (virtualenv) on Linux; reproduce script ran with: MYSQL_HOST=127.0.0.1 MYSQL_USER=root MYSQL_PASSWORD=12345 MYSQL_PORT=3306 python langgraph_issue/reproduce.py Repro output summary (paste exact): - Observed error: OperationalError: (1267, "Illegal mix of collations (utf8mb4_general_ci,IMPLICIT) and (utf8mb4_unicode_ci,IMPLICIT) for operation '='") - Observed error (insertion): ProgrammingError: (1366, "Incorrect string value: '\\xE0\\xA4...' for column `newversion`.`chat_sessions`.`session_title` at row 1") Additional useful info (optional): - Full output of `python -m langchain_core.sys_info` (if available) - DB user/privileges used by the checkpointer (do not paste passwords; paste roles/privs only) Short guidance on what to paste into the issue Paste the filled System Info block above into the GitHub "System Info *" field. Attach the langgraph_issue.zip bundle and the [reproduce.py](vscode-file://vscode-app/usr/share/code/resources/app/out/vs/code/electron-browser/workbench/workbench.html) run output (the console output you saw). Do not paste passwords or secrets; paste only versions, config, and the exact error text. [langgraph_issue.zip](https://github.com/user-attachments/files/21962694/langgraph_issue.zip)
yindo added the bugpending labels 2026-02-20 17:42:26 -05:00
yindo closed this issue 2026-02-20 17:42:26 -05:00
yindo commented 2026-02-20 17:42:27 -05:00
Author
Owner
Copy Link
Copy Source

@impptg commented on GitHub (Sep 18, 2025):

I meet the same problem, I tried to create the table by myself, and it works.

Temporary Solution：
Use these commands to view the content of migrations：

use this command to view the content of each migration.

Migration 1:
CREATE TABLE IF NOT EXISTS checkpoints (
thread_id VARCHAR(150) NOT NULL,
checkpoint_ns VARCHAR(150) NOT NULL DEFAULT '',
checkpoint_id VARCHAR(150) NOT NULL,
parent_checkpoint_id VARCHAR(150),
type VARCHAR(150),
checkpoint JSON NOT NULL,
metadata JSON NOT NULL DEFAULT ('{}'),
PRIMARY KEY (thread_id, checkpoint_ns, checkpoint_id)
);
....

Then manually specify COLLATE at the end

CREATE TABLE checkpoints (
......
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci

@impptg commented on GitHub (Sep 18, 2025): I meet the same problem, I tried to create the table by myself, and it works. Temporary Solution： Use these commands to view the content of migrations： use this command to view the content of each migration. Migration 1: CREATE TABLE IF NOT EXISTS checkpoints ( thread_id VARCHAR(150) NOT NULL, checkpoint_ns VARCHAR(150) NOT NULL DEFAULT '', checkpoint_id VARCHAR(150) NOT NULL, parent_checkpoint_id VARCHAR(150), type VARCHAR(150), checkpoint JSON NOT NULL, metadata JSON NOT NULL DEFAULT ('{}'), PRIMARY KEY (thread_id, checkpoint_ns, checkpoint_id) ); .... Then manually specify COLLATE at the end CREATE TABLE checkpoints ( ...... ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci
yindo commented 2026-02-20 17:42:27 -05:00
Author
Owner
Copy Link
Copy Source

@eyurtsev commented on GitHub (Nov 7, 2025):

PyMySQLSaver is not maintain by the langhcain team. I'm closing this issue, and would suggest raising it here: https://github.com/tjni/langgraph-checkpoint-mysql

@eyurtsev commented on GitHub (Nov 7, 2025): `PyMySQLSaver` is not maintain by the langhcain team. I'm closing this issue, and would suggest raising it here: https://github.com/tjni/langgraph-checkpoint-mysql
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
agents
awaiting author revision
bug
change
chat models
checkpointer
ci
code quality
code quality
dependencies
documentation
drawing
duplicate
enhancement
good first issue
help wanted
invalid
maintainer
meta
missing information
not maintained by langchain
open-swe-auto
open-swe-max
open-swe-max-auto
pending
performance
pull-request
Mirrored from GitHub Pull Request
 question
requires design
sdk
server
stale
studio
tests
type checking
usage question
v0.5
v1.1
No Label bug pending
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: langchain-ai/langgraph#937
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?