Unable to run flow with run_flow_from_json if Agent is present

[lucaseduoli]

Bug Description

If I create a flow with an Agent component, the flow cannot run with the Python function "run_flow_from_json". That is due to the Agent component receiving a Chat History, which tries to access the Database. Since "run_flow_from_json" does not run the Database, the execution fails.

Reproduction

Steps to reproduce the behavior:

1. Create a basic Agent flow
2. Add the API keys
3. Export the flow with the API keys
4. Run this code on a Python file:

from langflow.load import run_flow_from_json

flow = run_flow_from_json(flow="simple_agent.json", input_value="hello")

print(flow)

5. Receive the error: ValueError: Error running graph: Error building Component Agent: (sqlite3.OperationalError) no such table: message

Expected behavior

The flow should run without a chat history and display an output.

Who can help?

@italojohnny @ogabrielluiz @erichare

Operating System

macOS Sequoia

Langflow Version

1.3.2

Python Version

3.12

Screenshot

Flow File

simple_agent.json

[pybergonz]

Hi, I solved by doing this:

import sqlite3

# Connect to your SQLite database (or create it if it doesn't exist)
conn = sqlite3.connect('/home/vscode/.cache/langflow/langflow.db')

# Create a cursor object
cursor = conn.cursor()

# Create the 'message' table

cursor.execute('''DROP TABLE IF EXISTS message''')

cursor.execute('''
CREATE TABLE IF NOT EXISTS message (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    flow_id TEXT,
    sender TEXT,
    sender_name TEXT,
    session_id TEXT,
    text TEXT,
    files TEXT,
    timestamp DATETIME DEFAULT CURRENT_TIMESTAMP,
    error TEXT,
    content_blocks TEXT,
    category TEXT,
    properties TEXT,
    edit TEXT
)
''')

# Commit the changes and close the connection
conn.commit()
conn.close()

[Dowhanych]

@pybergonz Can you explain how exactly I should use it? Should I just paste this before the run_flow_from_json ?

[pybergonz]

@pybergonz Can you explain how exactly I should use it? Should I just paste this before the run_flow_from_json ?

create a new .py file called sql_lite_table.py.

copy paste this

`import sqlite3

Connect to your SQLite database (or create it if it doesn't exist)

conn = sqlite3.connect('/home/vscode/.cache/langflow/langflow.db')

Create a cursor object

cursor = conn.cursor()

Create the 'message' table

cursor.execute('''DROP TABLE IF EXISTS message''')

cursor.execute('''
CREATE TABLE IF NOT EXISTS message (
id TEXT,
flow_id TEXT,
sender TEXT,
sender_name TEXT,
session_id TEXT,
text TEXT,
files TEXT,
timestamp DATETIME DEFAULT CURRENT_TIMESTAMP,
error TEXT,
content_blocks TEXT,
category TEXT,
properties TEXT,
edit TEXT
)
''')

Commit the changes and close the connection

conn.commit()
conn.close()

print("OK!")`

run the following python sql_lite_table.py.

The script will create the missing table in /home/vscode/.cache/langflow/langflow.db (adjust to different path if needed).

[Dowhanych]

@pybergonz I have a several locations with that table. Not sure which one I need to change, but I tried for all of them, and it didn't work. The issue still exist..