The recent fix (PR #626) got the simulation running, but it exposed a completely unacceptable issue for a paid product: The "Generate Report" output is 100% hardcoded text. It does not analyze the prompt at all.
I suspected the LLM wasn't actually processing the simulation data, so I ran a test with a completely absurd and non-serious prompt to prove it.
My Input (Reality Seed & Prompt):
I inputted a scenario about a local coffee shop selling a "Spicy Garlic Matcha Latte" just to clear out garlic inventory, and asked if teenagers on TikTok would buy it as a joke. I explicitly instructed: "Do not talk about institutions or politics."
The Output (Generate Report):
The system completely ignored my prompt and outputted the exact same static placeholder text I get for every single simulation. It literally said:
"The first decisive shift happens when institutional framing collides with emotionally legible counter-narratives."
"Round 1-2 establish the moral frame of the scenario."
"Entity graph pressure is concentrated around institutional response..."
There is no "institutional framing" or "moral frame" for a garlic latte.
This proves that the report generation is just a static frontend template and is not connected to any actual LLM analysis. Why is placeholder text being served in a production environment to paid users?
Please clarify immediately if this feature is genuinely broken or if the core analysis function has not been implemented yet.
The recent fix (PR #626) got the simulation running, but it exposed a completely unacceptable issue for a paid product: The "Generate Report" output is 100% hardcoded text. It does not analyze the prompt at all.
I suspected the LLM wasn't actually processing the simulation data, so I ran a test with a completely absurd and non-serious prompt to prove it.
My Input (Reality Seed & Prompt):
I inputted a scenario about a local coffee shop selling a "Spicy Garlic Matcha Latte" just to clear out garlic inventory, and asked if teenagers on TikTok would buy it as a joke. I explicitly instructed: "Do not talk about institutions or politics."
The Output (Generate Report):
The system completely ignored my prompt and outputted the exact same static placeholder text I get for every single simulation. It literally said:
There is no "institutional framing" or "moral frame" for a garlic latte.
This proves that the report generation is just a static frontend template and is not connected to any actual LLM analysis. Why is placeholder text being served in a production environment to paid users?
Please clarify immediately if this feature is genuinely broken or if the core analysis function has not been implemented yet.