Update app.py

Replace large model with small model for initial tasks.

app.py

@@ -405,6 +405,38 @@ def writing_task(prompt: str) -> str:
     
     return content
 
+def smol_writing_task(prompt: str) -> str:
+
+    API_TOKEN_SMOLLM = getenv("HF_TOKEN")
+
+    client = InferenceClient(
+        provider="hf-inference",
+        api_key=API_TOKEN_SMOLLM,
+    )
+
+    completion = client.chat.completions.create(
+        model="HuggingFaceTB/SmolLM3-3B",
+        messages=[
+            {
+                "role": "user",
+                "content": prompt
+            }
+        ],
+    )
+
+    raw_content = completion.choices[0].message['content']
+    print(f"Raw content: {raw_content}")
+
+    content_split = raw_content.split("</think>")
+
+    if len(content_split) > 1:
+        think = content_split[0]
+        content = "".join(content_split[1:])
+    else:
+        think = content_split[0]
+        content = "No data found."
+    
+    return content
 
 def smol_lm_jd_process(job_description, system_prompt, max_new_tokens=512):
     prompt = f"""<|im_start|>system
@@ -419,7 +451,8 @@ def smol_lm_jd_process(job_description, system_prompt, max_new_tokens=512):
     # start_idx = response.find("<|im_start|>assistant")
     # end_idx = response.find("<|im_end|>", start_idx)
     # response = response[start_idx + len("<|im_start|>assistant\n"):end_idx].strip()
-    response = writing_task(prompt)
+    # response = writing_task(prompt)
+    response = smol_writing_task(prompt)
     return response
 
 def process_job_description(company_name, company_url, job_description, resume):