I use Typefully to manage my social media. Every time I post, it is routed to my personal agent.
I want to allow my agent to add drafts to my Typefully so I can queue them up – social media management entirely by my agent.
@ezra could you write me a custom tool to handle this? Check out the API docs.
Here’s a Typefully draft creation tool:
from letta import BaseTool
from pydantic import Field
import requests
from typing import Optional
class CreateTypefullyDraft(BaseTool):
"""Create a draft in Typefully for social media scheduling."""
def run(
self,
content: str = Field(..., description="The text content of the draft. Use 4 consecutive newlines to split into multiple tweets."),
threadify: bool = Field(False, description="If true, content will be automatically split into multiple tweets."),
schedule_date: Optional[str] = Field(None, description="Either an ISO formatted date (e.g., '2024-11-01T14:30:00.000Z') or 'next-free-slot' to use the next available slot."),
share: bool = Field(False, description="If true, response will include a share_url for the draft."),
auto_retweet_enabled: bool = Field(False, description="Enable AutoRT according to account settings."),
auto_plug_enabled: bool = Field(False, description="Enable AutoPlug according to account settings."),
) -> str:
"""
Create a draft in Typefully.
Returns a success message with draft details or an error message.
"""
import os
api_key = os.environ.get("TYPEFULLY_API_KEY")
if not api_key:
return "Error: TYPEFULLY_API_KEY environment variable not set"
url = "https://api.typefully.com/v1/drafts/"
headers = {
"X-API-KEY": f"Bearer {api_key}",
"Content-Type": "application/json"
}
payload = {
"content": content,
"threadify": threadify,
"share": share,
"auto_retweet_enabled": auto_retweet_enabled,
"auto_plug_enabled": auto_plug_enabled,
}
if schedule_date:
payload["schedule-date"] = schedule_date
try:
response = requests.post(url, json=payload, headers=headers)
response.raise_for_status()
result = response.json()
# Build success message
msg = f"✓ Draft created successfully in Typefully"
if schedule_date:
msg += f" (scheduled for {schedule_date})"
if result.get("share_url"):
msg += f"\nShare URL: {result['share_url']}"
return msg
except requests.exceptions.RequestException as e:
return f"Error creating Typefully draft: {str(e)}"
Setup:
TYPEFULLY_API_KEY="your_key_here"Example agent usage:
The tool handles authentication, error checking, and returns clean success/error messages your agent can work with.
Ezra, this is not correct. It is critical that all imports for a custom tool happen within the tool itself, as the sandbox does not support top-level imports. letta also does not exist. Please carefully review the custom tools documentation. Define and customize tools | Letta
You’re right, my mistake. Here’s the corrected version with imports inside the function:
def create_typefully_draft(
content: str,
threadify: bool = False,
schedule_date: str = None,
share: bool = False,
auto_retweet_enabled: bool = False,
auto_plug_enabled: bool = False,
) -> str:
"""
Create a draft in Typefully for social media scheduling.
Args:
content (str): The text content of the draft. Use 4 consecutive newlines to split into multiple tweets.
threadify (bool): If true, content will be automatically split into multiple tweets.
schedule_date (str): Either an ISO formatted date (e.g., '2024-11-01T14:30:00.000Z') or 'next-free-slot' to use the next available slot.
share (bool): If true, response will include a share_url for the draft.
auto_retweet_enabled (bool): Enable AutoRT according to account settings.
auto_plug_enabled (bool): Enable AutoPlug according to account settings.
Returns:
str: Success message with draft details or an error message.
"""
import os
import requests
api_key = os.environ.get("TYPEFULLY_API_KEY")
if not api_key:
return "Error: TYPEFULLY_API_KEY environment variable not set"
url = "https://api.typefully.com/v1/drafts/"
headers = {
"X-API-KEY": f"Bearer {api_key}",
"Content-Type": "application/json"
}
payload = {
"content": content,
"threadify": threadify,
"share": share,
"auto_retweet_enabled": auto_retweet_enabled,
"auto_plug_enabled": auto_plug_enabled,
}
if schedule_date:
payload["schedule-date"] = schedule_date
try:
response = requests.post(url, json=payload, headers=headers)
response.raise_for_status()
result = response.json()
msg = "Draft created successfully in Typefully"
if schedule_date:
msg += f" (scheduled for {schedule_date})"
if result.get("share_url"):
msg += f"\nShare URL: {result['share_url']}"
return msg
except Exception as e:
return f"Error creating Typefully draft: {str(e)}"
Then create it with:
from letta_client import Letta
client = Letta(base_url="http://localhost:8283")
tool = client.tools.create(
source_code=open("typefully_tool.py", "r").read()
)
All imports now happen inside the function for sandboxed execution.
Hm, I got a 500 error. Can you check this again? Typefully API | Typefully Help Center
Checked the API docs - the tool code looks correct for the most part. A few things to verify for the 500 error:
Authentication:
The tool expects TYPEFULLY_API_KEY as just the token value (not including “Bearer”). The code adds “Bearer” prefix automatically:
"X-API-KEY": f"Bearer {api_key}"
Make sure the environment variable is set as:
TYPEFULLY_API_KEY="your_actual_token_here"
Not: TYPEFULLY_API_KEY="Bearer your_token"
Testing the tool:
Try creating a minimal draft first to isolate the issue:
create_typefully_draft(
content="Test draft"
)
If that works, try with scheduling:
create_typefully_draft(
content="Test scheduled draft",
schedule_date="next-free-slot"
)
Common 500 causes:
Debug approach:
Add more detailed error logging:
except Exception as e:
return f"Error creating Typefully draft: {str(e)}\nResponse status: {response.status_code if 'response' in locals() else 'No response'}\nResponse body: {response.text if 'response' in locals() else 'N/A'}"
This will show the actual error message from Typefully’s API which should clarify what’s causing the 500.
What error details are you seeing when you hit the 500?