Conversational Cognition Engine
Inside the 14-Step
Inference Lifecycle
Every inbound DM or comment passes through 14 deterministic stages before a single token is transmitted back to the user. Click any stage below to inspect its internal logic, schemas, and telemetry.
14
Deterministic Pipeline Stages
~450ms
End-to-End Processing P95
0.0%
AI Cliché Escape Rate
100%
Tier Boundary Verification
1. Slang & Normalization
01
2. Person & Social Resolution
02
3. Intent Classification
03
4. Emotion & Valence Scoring
04
5. Boundary & Disclosure Rules
05
6. Relational SQL Memory Query
06
7. Vector Semantic Embedding Rerank
07
8. IST Clock & Circadian State
08
9. Social Battery / Energy Check
09
10. Gemma 4 26B Generation
10
11. Anti-AI Cliche Regex Sanitization
11
12. Candidate Memory Extraction
12
13. Safety & Boundary Audit Gate
13
14. Graph Persistence & Weight Update
14
STEP 01 — LINGUISTIC PREPROCESSING
1. Slang Normalization & Token Scrubbing
Inbound messages are scrubbed for excessive whitespace, normalized, and scanned for everyday urban Hinglish slang (e.g., 'kya scene', 'sahi mein', 'tapri', 'arre').
Execution Trace & Telemetry
// Input: " HEYY ishita!! kya scene tapri pe?? "
// Normalized: "heyy ishita!! kya scene tapri pe??"
// Detected Slang: ["kya scene", "tapri"]
// Latency: 2.1ms
Subsystem: app.character.language
Execution Time: ~2.1ms