Activity
Mon
Wed
Fri
Sun
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
What is this?
Less
More

Memberships

AI Automation Society

Public • 3.7k • Free

AI Automation Society Plus

Private • 128 • $59/m

1 contribution to AI Automation Society
Trying to structure data.
I keep running into entropy where my JSON getting different results. ``` {"output":{"output":{"isReply":true,"reasoning":"The email is from a hum...}}} ``` Mainly being that this double "output" object is being sent to my Structured Object Parser in n8n. The parser is expecting a single object, and the stack breaks when it gets nested data from the model. Has anyone run into this before? I can show my prompt is necessary. Thanks.
0
6
New comment 12d ago
0 likes • 12d
@Simon Coton Thanks Simon, this is what I have going on in this thing. ```prompt ... ... 7. **Timing and Relevance:** - If the email is older than 30 days, deem it likely unimportant. - In such cases, respond minimally or output `"FALSE"` if no reply is needed. **Final Output Requirements:** - Reply with a single JSON containing the following key values in one object (NO NESTING)): - `isReply`: `true` if replying, `false` if not. - `reasoning`: Explain why you chose to reply or not. - `to`: The recipient’s email. - `from`: `properties@fonserve.net` - `subject`: Same as the received email’s subject. - `message`: The composed reply or `"FALSE"` if not replying. **Date Context:** - Today’s date: `{{ $now.format('yyyy-MM-dd') }}` - Received email’s date: `{{ $json.emailRecievedDate }}` - Compare dates to determine if email is older than 30 days. ... ... ``` ```Structured Output Parser { "type": "object", "properties": { "isReply": { "type": "boolean" }, "reasoning": { "type": "string" }, "to": { "type": "string" }, "from": { "type": "string" }, "subject": { "type": "string" }, "message": { "type": "string" } } } ``` It works sometimes, but other times it wants to nest the output into two objects, not super consistent. Hopefully this gives you more detail.
0 likes • 12d
@James Walls Oh shoot, so telling model how to output then having the Output Parser is not good practice? How does the system understand the Parser, is there outside data being passed in after the prompt?
1-1 of 1
Bryce Fountain
1
5points to level up
@bryce-fountain-5485
Just a dude

Active 5d ago
Joined Dec 12, 2024
powered by