Forum Replies Created
-
Posts
-
I can’t believe I am reading this. So you are telling me you can’t query this:
https://api.spoonacular.com/recipes/{recipeID}/information?apiKey={KEY}
Simply increase the recipeID integer in each query, and maybe to avoid getting banned by APi make some kind of random search from time to time. I am supposing is already an endpoint the plugin uses to get the recipes info, no idea.
The plugin is great but I ended up regenerating the title, the content, the featured image, the categories and the tags with AI, and it seems more practical to directly generate the recipe using only AI. Unfortunately I keep saying prompts are manipulated by Aiomatic in a way that is not helpful to me, and nobody gives a solution so I can’t use Aiomatic either.
-
-
Hi, I already saw the prompts there. That is not the point of this thread. It takes more time to find the prompts in the Activity Logs. The idea is to have them also located in the Usage Logs, where there is no other information (like the 16k url length the plugin generates for ValueSERP and always seems to fail, it doesn’t say what failed; or also lot of text about failed embeddings that give no real clue about what the error is)
-
Thank you for your answer.
Detailed logging already enabled.
Now, adding to the problems with the Excerpt prompt being forcefully wasting tokens, and the AI Content Completition generating 3-4 times more text then requested, and that I can’t activate Automatic Content Editing because import rule fails, is the fact that I setup the import rule to run every hour, but it runs only every 2 hours.
My server cron runs every 30 minutes.
At least I can run the Existing Content Editor and it does the job.
-
I just found where the Excerpt can be generated by AI, and like always there must always be a “BUT”. This is what instructions say for:
“Instructions to Send For the AI Editor (Content Editing):
Instruction for the AI editor, to edit post excerpt. Please specify your instruction without adding the %%post_excerpt%% shortcode, as the excerpt will be automatically added at processing time.”
This means plugin will waste my tokens by always sending %%post_excerpt%% even if I want the excerpt to be completely rewritten ignoring current excerpt. Interesting.
-
Interesting, I have tested and for about 15 seconds I was extremely happy. The feature you showed me could be extremely useful for me.
Your instructions work BUT:
1. There is no option to generate/replace the excerpt. Yummomatic generates the excerpt by simply copying first X characters, and that is ugly and not SEO friendly. (btw I marked the checkbox to “Disable Post Excerpt” in Yummomatic Settings but it still fills the excerpt)
2. Yummomatic returns a status error when running the rule, wich then executes this query:
It was running well until it reached offset 28. I look on all logs, even on php side, and there is no error related, but if I disable “Automatic Content Editing/Posts Automatic Editing”, in Aiomatic, it runs ok with no error. I turn it ON and status is error. Like with other errors, there is no useful information in logs.
I increased PHP execution time, memory and basically tried running it with little to no restriction and it stills fails.
The thing is I run “AI Content Editor” manually (when editing the post, and in bulk with “Existing Content Editor”) and it runs ok. It only fails when is generating a new post. So there must be some incompatibility betwen Aiomatic and Yummomatic.
On “Editing Templates and Options” I only activated:
– “AI Generated Featured Image” -> it works as expected
– “AI Content Completition” -> AI Model: gpt-4-preview -> Append to the end -> the problem is the prompt is modified, and my original prompt is not 100% followed. It would be a great improvement to show in logs ALL the prompts sent to OpenAI, exactly the way they are sent. I tested my prompt 100 times making manual requests to the API and response is always exactly as expected, but your plugin modifies the prompt. Basically the length is 3-4 times longer then specified in the prompt. It returns text of almost 3k characters length when I request it to be of 900 characters. And OpenAI tends to make it always shorter, not longer, and definitely not by so much difference.
-
Thank you very much, for your answer.
Unfortunately for now I had to move away from this plugin because I can’t find the reason of the high OpenAI cost it incurs into while using this. I run 4 times the “Single AI Post Creator” and 2 times the “Bulk AI Post Creator” set for 3 articles.
In total I ended up with just 2 articles created and it cost me more then 11€, because there were some errors like section headers showing only two dots “:”, and some articles directly didn’t apear.
The cost is simply to high. I created with same prompt over 40 articles and it cost me under 4€
I will not request a refund since I can see all the hard work and I trust that the plugin will get better in time.
I congratulate you for everything done and will keep in touch.
