update backend for AWS

README.md

@@ -16,7 +16,7 @@ pinned: false
 | Service | Technology |
 |---------|-----------|
 | API Framework | FastAPI + Uvicorn |
-| LLM Inference | Ollama / Phi 3.5 Mini (local) |
+| LLM Inference | Ollama / Llama3 (local) |
 | Text Emotion | DistilBERT (`bhadresh-savani/distilbert-base-uncased-emotion`) |
 | Crisis Detection | Zero-shot NLI (MiniLM) |
 | Face Emotion | Keras CNN (custom trained, `emotion_model_trained.h5`) |
@@ -37,7 +37,7 @@ pinned: false
 ```bash
 # 1. Install Ollama + LLaMA 3 (one-time)
 winget install Ollama.Ollama
-ollama pull phi3.5:3.8b-mini-instruct-q4_0
+ollama pull llama3
 
 # 2. Install dependencies
 pip install -r requirements.txt

app/api/endpoints/therapist.py

@@ -5,7 +5,7 @@ Full inference pipeline:
   2. Text emotion classification (DistilBERT)
   3. Crisis evaluation (zero-shot NLI) — override if triggered
   4. Multimodal fusion (text + face)
-  5. Ollama/Phi-3.5 Mini structured report generation
+  5. Ollama/Llama3 structured report generation
   6. PsychReport JSON schema validation
   7. Streaming response option
 """

app/config.py

@@ -12,7 +12,7 @@ class Settings(BaseSettings):
     # Default is localhost (e.g. for development), but in production it should be like:
     # OLLAMA_BASE_URL: str = "http://123.45.67.89:11434"
     OLLAMA_BASE_URL: str = "http://127.0.0.1:11434"
-    OLLAMA_MODEL: str = "phi3.5:3.8b-mini-instruct-q4_0"
+    OLLAMA_MODEL: str = "llama3"
     OLLAMA_TIMEOUT_S: int = 90
     
     # Retry logic for external LLM API

app/main.py

@@ -108,7 +108,7 @@ def create_app() -> FastAPI:
         title="PsyPredict API",
         description=(
             "Production-grade multimodal mental health AI system. "
-            "Powered by Phi-3.5 Mini (Ollama) + DistilBERT + Keras CNN facial emotion model."
+            "Powered by Llama3 (Ollama) + DistilBERT + Keras CNN facial emotion model."
         ),
         version="2.0.0",
         lifespan=lifespan,

app/services/ollama_engine.py

@@ -1,5 +1,5 @@
 """
-ollama_engine.py — PsyPredict Local LLM Engine (Phi-3.5 Mini)
+ollama_engine.py — PsyPredict Local LLM Engine (Llama3)
 Async Ollama client with:
   - Structured JSON output enforced via schema-in-prompt + Ollama format param
   - Context window trimming
@@ -97,7 +97,7 @@ FACE_DISTRESS_MAP: dict[str, float] = {
 
 class OllamaEngine:
     """
-    Production async LLM engine backed by local Ollama/Phi-3.5 Mini.
+    Production async LLM engine backed by local Ollama/Llama3.
     """
 
     def __init__(self) -> None:

main.py

@@ -118,7 +118,7 @@ def create_app() -> FastAPI:
         title="PsyPredict API",
         description=(
             "Production-grade multimodal mental health AI system. "
-            "Powered by Phi-3.5 Mini (Ollama) + DistilBERT + Keras CNN facial emotion model."
+            "Powered by Llama3 (Ollama) + DistilBERT + Keras CNN facial emotion model."
         ),
         version="2.0.0",
         lifespan=lifespan,