llama 3

components/LLMs/LLama3.py

@@ -0,0 +1,43 @@
+import os
+import requests
+from typing import Optional
+
+# 🔐 Environment Variables
+LLAMA3_URL = "https://c5dk65n3sd14gjo1.us-east-1.aws.endpoints.huggingface.cloud"
+HF_TOKEN = os.environ.get("HF_TOKEN")
+
+# 📜 Headers
+HEADERS = {
+    "Authorization": f"Bearer {HF_TOKEN}",
+    "Content-Type": "application/json"
+}
+
+# 🧠 Prompt builder and caller for LLaMA 3 8B (QCA)
+def call_llama3_8b(base_prompt: str, tail_prompt: str) -> Optional[str]:
+    prompt = f"<s>[INST]{base_prompt}\n\n{tail_prompt}[/INST]</s>"
+
+    try:
+        response = requests.post(
+            LLAMA3_URL,
+            headers=HEADERS,
+            json={"inputs": prompt},
+            timeout=60
+        )
+        response.raise_for_status()
+        data = response.json()
+
+        # Parse generated output
+        if isinstance(data, list) and data:
+            raw_output = data[0].get("generated_text", "")
+        elif isinstance(data, dict):
+            raw_output = data.get("generated_text", "")
+        else:
+            return None
+
+        if "[/INST]</s>" in raw_output:
+            return raw_output.split("[/INST]</s>")[-1].strip()
+        return raw_output.strip()
+
+    except Exception as e:
+        print(f"⚠️ LLaMA 3.1 8B API call failed: {e}")
+        return None

components/generators/daily_feed.py

@@ -11,6 +11,7 @@ from llama_index.core.settings import Settings
 from components.LLMs.Mistral import call_mistral
 from components.LLMs.TinyLLama import call_tinyllama
 from components.LLMs.Bart import call_bart_summarizer
+from components.LLMs.LLama3 import call_llama3_8b
 
 # ✅ Disable implicit LLM usage
 Settings.llm = None
@@ -50,7 +51,7 @@ def summarize_topic(docs: List[str], topic: str) -> List[Dict]:
     for doc in docs[:5]:
         tail_prompt = f"Topic: {topic}\n\n{doc.strip()}"
         print(f"\n📤 Prompt tail for Mistral:\n{tail_prompt[:300]}...\n")
-        summary_block = call_bart_summarizer(base_prompt=BASE_PROMPT, tail_prompt=tail_prompt)
+        summary_block = call_llama3_8b(base_prompt=BASE_PROMPT, tail_prompt=tail_prompt)
 
         if summary_block:
             for line in summary_block.splitlines():