AI/ML●●●Banger
Mate – Emotional layer on top of LLMs
Pure math kernel produced emergent behavior — one instance started dreaming unprompted.
Zero to OneRabbit HoleWizardry
smugglerFlynn
432mo ago
Back to browse
A Structured CBT Orchestration Engine Built on Top of LLMs
by sucharithan·Feb 18, 2026·1 point·0 comments
AI Analysis
●●SolidBig BrainNiche GemSlick
The Take
This forces LLMs to play inside a deterministic CBT pipeline — it extracts distortion signals, calibrates emotional intensity, applies rule-based risk tiers, then generates tone-locked replies with word caps. The split between deterministic detection and constrained drafting is smart and makes outputs more auditable; Reflect vs Assist modes show sensible product framing. Promising concept for safety-minded builders, but the real value hinges on the model tuning and risk-handling under real conversations, not the attractive landing UI.
Category
Target Audience
Coaches and therapists, empathy-focused product builders, developers of constrained LLM workflows, and anyone who wants structured help replying to emotional messages
Post Description
I built a structured CBT engine that sits on top of LLMs and enforces cognitive workflow logic before generating responses.
Most AI tools in this space are purely conversational. This system instead:
Extracts cognitive distortion signals
Calibrates emotional intensity
Applies rule-based risk-tier logic
Separates deterministic detection from generative drafting
Enforces tone presets and word caps to avoid generic output
It runs in two modes:
Reflect → structured self-guided reframing Assist → structured signal extraction + constrained response drafting for coaches/therapists
The goal wasn’t to build another chatbot, but to explore whether LLMs can be constrained inside a deterministic cognitive architecture.
Would love feedback from people building structured AI systems or workflow-constrained LLM tools.
Similar Projects
Security●●Solid
LLM-Audit – Semgrep Rules for OWASP LLM Top in TypeScript
Fills the TypeScript gap that Semgrep's official AI best practices pack misses.
Niche Gem
Javierlozo
101mo ago
Health●●Solid
NightClear – AI anxiety clearing for bedtime
CBT-grounded anxiety container, but meditation app category already flooded.
Solve My ProblemCozyDark Horse
AliThink
103mo ago
Developer Tools●●Solid
HN Bot Detector - Detects LLM-Generated Comments on Hacker News
Clever n-gram TF-IDF detection of LLM paraphrases catches smart evasion; solves real HN problem but narrow use case.
Big BrainDark Horse
umairnadeem123
733mo ago
AI/ML●●Solid
Regression tests for detecting cross-domain hallucinations in LLMs
Regression tests catch cross-domain hallucinations, but prompt-based approach won't scale.
Big BrainNiche Gem
Ginsabo
123mo ago
AI/ML●●Solid
SIB-ENGINE Pre-emptive hallucination detection via geometric structure
Monitors internal latent collapse before tokens are sampled, not output semantics.
Big BrainWizardry
yubainu
113mo ago