docs: Add guide for fixing model name issue

Author: anhmtk

docs/FIX_MODEL_NAME.md

@@ -0,0 +1,62 @@
+# Fix Model Name Issue
+
+## Problem
+
+StillMe is mentioning the incorrect embedding model name (`all-MiniLM-L6-v2`) instead of the correct one (`paraphrase-multilingual-MiniLM-L12-v2`).
+
+## Root Causes
+
+1. **Cached Responses**: LLM response cache contains old responses with incorrect model name
+2. **Foundational Knowledge**: Foundational knowledge in RAG database may have old model name
+
+## Solution
+
+### Step 1: Update Foundational Knowledge
+
+Run the script to update foundational knowledge in RAG database:
+
+```bash
+python scripts/update_foundational_knowledge_model_name.py
+```
+
+This script will:
+- Search for foundational knowledge documents with old model name
+- Replace `all-MiniLM-L6-v2` with `paraphrase-multilingual-MiniLM-L12-v2`
+- Update documents in ChromaDB
+
+### Step 2: Clear LLM Cache
+
+Clear the LLM response cache to force regeneration:
+
+```bash
+python scripts/clear_llm_cache.py
+```
+
+Or via API (if available):
+
+```bash
+curl -X POST http://localhost:8000/api/cache/clear
+```
+
+### Step 3: Verify
+
+Test StillMe response to verify correct model name:
+
+1. Ask StillMe: "Bạn đang sử dụng mô hình embedding nào?"
+2. Response should mention: `paraphrase-multilingual-MiniLM-L12-v2`
+3. Should NOT mention: `all-MiniLM-L6-v2`
+
+## Prevention
+
+To prevent this issue in the future:
+
+1. **Single Source of Truth**: Use `stillme_core/rag/model_info.py` for model information
+2. **Update Foundational Knowledge**: When model changes, update foundational knowledge script
+3. **Clear Cache**: After updating foundational knowledge, clear cache to force regeneration
+
+## Related Files
+
+- `stillme_core/rag/model_info.py` - Single source of truth for model information
+- `scripts/add_foundational_knowledge.py` - Foundational knowledge template
+- `backend/identity/prompt_builder.py` - Prompt instructions with model name
+