Update Space to ClarityGuard v2

README.md

@@ -23,6 +23,31 @@ models:
 
 ClarityGuard helps neurodivergent individuals decode ambiguous workplace and personal messages by analyzing message structure—not the user's ability to understand.
 
+## Active Model
+
+| Property | Value |
+|----------|-------|
+| Model repo | `CharlieBonito/clarity-guard-gemma4-7b` |
+| Active version | ClarityGuard v2 |
+| Training checkpoint | 750 |
+| Base model | Unsloth Gemma 4 E4B IT BNB 4-bit |
+| Architecture | Gemma 4 |
+| Parameters | 7.52B |
+| Quantization | GGUF / Q4_K_M |
+| Model context metadata | 131072 tokens |
+| Space deployed context | 12288 tokens |
+| Multimodal | Yes, via `mmproj-ClarityGuard-v2.gguf` |
+
+Active production files:
+
+- `ClarityGuard-v2.gguf` — main model
+- `mmproj-ClarityGuard-v2.gguf` — multimodal projector
+
+Deprecated checkpoint-375 files are not the active deployment artifacts:
+
+- `Checkpoint-375-Ollama-Clean-7.5B-Q4_K_M.gguf`
+- `mmproj-Checkpoint-375-Ollama-Clean-BF16.gguf`
+
 ## 🎯 Problem
 
 Neurodivergent people (autistic, ADHD, dyslexic) often struggle with:
@@ -54,24 +79,25 @@ The model then generates:
 ## 🏗️ Architecture
 
 ```
-User Message → Jina Embeddings (RAG) → Gemma 4 4B Finetuned → Structured Analysis
+User Message → Jina Embeddings (RAG) → ClarityGuard v2 / Gemma 4 E4B IT → Structured Analysis
                      ↓
            Knowledge Base (Chatty System)
 ```
 
 ### Components:
-- **Fine-tuned Gemma 4 E4B** (Unsloth) — 7.5B parameters, Q4_K_M quantization
+- **ClarityGuard v2 / Fine-tuned Gemma 4 E4B IT** (Unsloth) — 7.52B parameters, Q4_K_M quantization, checkpoint 750
 - **Jina Embeddings v3** — Semantic search over knowledge base
 - **RAG Documents** — Chatty 231051 framework + manipulation awareness content
-- **ZeroGPU H200** — Inference acceleration
+- **Hugging Face GPU Space** — CUDA-accelerated llama.cpp inference
 
 ## 🚀 Technical Details
 
 ### Model Training
-- Base: Gemma 4 E4B (4-bit quantized)
+- Base: Unsloth Gemma 4 E4B IT BNB 4-bit
 - Fine-tuning: Unsloth Studio
+- Active checkpoint: 750
 - Quantization: Q4_K_M for deployment
-- Multimodal support: mmproj for vision
+- Multimodal support: `mmproj-ClarityGuard-v2.gguf` for vision/audio projector support
 
 ### RAG System
 - Embeddings: Jina v3 (1024 dimensions)
@@ -135,4 +161,4 @@ Apache 2.0
 
 ---
 
-**Built with ❤️ for the neurodivergent community**
+**Built with ❤️ for the neurodivergent community**

SETUP.md

@@ -38,14 +38,14 @@ api = HfApi()
 api.create_repo("CharlieBonito/clarity-guard-gemma4-7b", repo_type="model", exist_ok=True)
 EOF
 
-# Upload model files (this will take ~30-60 minutes for 5GB)
+# Upload active ClarityGuard v2 model files (this will take ~30-60 minutes for 5GB)
 huggingface-cli upload CharlieBonito/clarity-guard-gemma4-7b \
-    /home/charlie/charlielinux/pasar/Checkpoint-375-Ollama-Clean-7.5B-Q4_K_M.gguf \
-    Checkpoint-375-Ollama-Clean-7.5B-Q4_K_M.gguf
+    /home/charlie/charlielinux/finalversion/models/ClarityGuard-v2.gguf \
+    ClarityGuard-v2.gguf
 
 huggingface-cli upload CharlieBonito/clarity-guard-gemma4-7b \
-    /home/charlie/charlielinux/pasar/mmproj-Checkpoint-375-Ollama-Clean-BF16.gguf \
-    mmproj-Checkpoint-375-Ollama-Clean-BF16.gguf
+    /home/charlie/charlielinux/finalversion/models/mmproj-ClarityGuard-v2.gguf \
+    mmproj-ClarityGuard-v2.gguf
 ```
 
 ### Step 4: Enable ZeroGPU
@@ -62,7 +62,7 @@ Gradio Interface
     ↓
 ClarityGuard Prompt + RAG Context
     ↓
-Gemma 4 4B Fine-tuned (ZeroGPU)
+ClarityGuard v2 / Gemma 4 E4B IT checkpoint 750 (GPU Space)
     ↓
 Structured Analysis (C.F.R.V.A.)
     ↓
@@ -74,4 +74,4 @@ User Response
 - Digital Equity & Inclusivity ($10,000)
 - Safety & Trust ($10,000)
 - Unsloth Track ($10,000)
-- llama.cpp Track ($10,000)
+- llama.cpp Track ($10,000)

app.py

@@ -46,7 +46,7 @@ RAG_DOCS = [
 ]
 
 CLARITYGUARD_SYSTEM_PROMPT = """CLARITYGUARD ASSISTANT — NEURO-INCLUSIVE EDITION v4.7
-Tuned for Gemma 4 4B Finetuned | Dify + Jina RAG
+Tuned for ClarityGuard v2 / Gemma 4 E4B IT checkpoint 750 | Dify + Jina RAG
 Based on C.F.R.V.A., created by Carlos Lengemann (2026) — CC BY 4.0
 
 Language policy (non-negotiable): These instructions are written in English for clarity for builders. Your replies to the user must always be in the same language the user uses in their current message (including step titles, examples, and suggested wording). If the user mixes languages, mirror the language of their question / framing (the part where they ask for help), not the quoted third-party text. Never concatenate words. Always write with correct spacing and normal punctuation.
@@ -178,7 +178,7 @@ Spacing Rule (all modes): Never concatenate words. Use correct spacing and stand
 
 ═══════════════════════════════════════════════════════
 Version: ClarityGuard v4.7 — Structural / Neuro-inclusive
-Stack: Gemma 4 4B Finetuned | Dify | Jina RAG
+Stack: ClarityGuard v2 / Gemma 4 E4B IT checkpoint 750 | Dify | Jina RAG
 Framework: C.F.R.V.A. (Lengemann, 2026) | Input Triage | Operational Pragmatics | Universality of Ambiguity
 Attribution: Based on C.F.R.V.A., created by Carlos Lengemann (2026). Licensed CC BY 4.0.
 https://creativecommons.org/licenses/by/4.0/deed.es"""