Update app.py

app.py

@@ -1,5 +1,5 @@
 """
-OhamLab — AI Intelligence 
+OhamLab — AI Intelligence
 Loads knowledge from rahul7star/OhamLab-LLM markdown corpus, caches embeddings,
 and provides retrieval-augmented chat through Hugging Face router.
 """
@@ -25,10 +25,10 @@ HF_TOKEN = (
 if not HF_TOKEN:
     raise RuntimeError("❌ Missing HF_TOKEN / OPENAI_API_KEY / HUGGINGFACE_TOKEN environment variable.")
 
-MODEL_ID = "openai/gpt-oss-20b"               # Chat model (via HF router)
-EMBED_MODEL = "text-embedding-3-small"        # Embedding model
-HF_REPO = "rahul7star/OhamLab-LLM"            # Knowledge repo
-CACHE_PATH = "/tmp/ohamlab_emb_cache.json"    # Cache file
+MODEL_ID = "openai/gpt-oss-20b"                # Chat model (via HF router)
+EMBED_MODEL = "BAAI/bge-small-en-v1.5"         # Router-compatible embedding model
+HF_REPO = "rahul7star/OhamLab-LLM"             # Knowledge repo
+CACHE_PATH = "/tmp/ohamlab_emb_cache.json"     # Cache file
 
 client = OpenAI(base_url="https://router.huggingface.co/v1", api_key=HF_TOKEN)
 api = HfApi(token=HF_TOKEN)
@@ -56,10 +56,11 @@ def load_ohamlab_knowledge():
                 chunks.append({"file": f, "text": buf.strip()})
         except Exception as e:
             print(f"⚠️ Failed to load {f}: {e}")
+    print(f"📚 Loaded {len(chunks)} text chunks from {len(md_files)} markdown files.")
     return chunks
 
 # ---------------------------
-# 3. Generate or Load Embeddings (with Cache & Retry)
+# 3. Embeddings (with Cache & Retry)
 # ---------------------------
 def create_embeddings_with_retry(texts, retries=3, delay=2):
     """Generate embeddings with retries on failure."""
@@ -121,7 +122,7 @@ def retrieve_knowledge(query, top_k=3):
         return ""
 
 # ---------------------------
-# 5. System Prompt with Context Injection
+# 5. System Prompt
 # ---------------------------
 def build_system_prompt(context: str, mode: str = "chat") -> str:
     return textwrap.dedent(f"""
@@ -129,8 +130,8 @@ def build_system_prompt(context: str, mode: str = "chat") -> str:
 
     Guidelines:
     - Only answer using information retrieved from the OhamLab knowledge bank (.md files).
-    - Do not answer anything outside of this knowledge; if unknown, respond: "I could not find an answer in my knowledge base."
-    - Always answer with clarity, scientific accuracy, and concise insight.
+    - Do NOT hallucinate. If the answer is not found, respond: "I could not find an answer in my knowledge base."
+    - Always answer with clarity, precision, and factual grounding.
     - Avoid code unless explicitly requested.
     - Mode: {mode.upper()}
 
@@ -140,7 +141,7 @@ def build_system_prompt(context: str, mode: str = "chat") -> str:
     """).strip()
 
 # ---------------------------
-# 6. Model Call
+# 6. Response Generation
 # ---------------------------
 def generate_response(user_input, history, mode="chat"):
     context = retrieve_knowledge(user_input)
@@ -152,7 +153,7 @@ def generate_response(user_input, history, mode="chat"):
         resp = client.chat.completions.create(
             model=MODEL_ID,
             messages=messages,
-            temperature=0.0,  # deterministic for knowledge-only responses
+            temperature=0.0,  # deterministic for knowledge-only answers
             max_tokens=1200,
         )
         return resp.choices[0].message.content.strip()
@@ -164,10 +165,8 @@ def generate_response(user_input, history, mode="chat"):
 # 7. Chat Logic
 # ---------------------------
 def chat_with_model(user_message, chat_history):
-    """Maintains full conversational context and returns updated chat history."""
     if not user_message:
         return chat_history, ""
-
     if chat_history is None:
         chat_history = []
 
@@ -176,7 +175,6 @@ def chat_with_model(user_message, chat_history):
         for m in chat_history
         if isinstance(m, dict) and "role" in m
     ]
-
     history.append({"role": "user", "content": user_message})
 
     try:
@@ -192,7 +190,7 @@ def reset_chat():
     return []
 
 # ---------------------------
-# 8. Gradio Chat UI
+# 8. Gradio UI
 # ---------------------------
 def build_ui():
     with gr.Blocks(
@@ -212,7 +210,6 @@ def build_ui():
             button.secondary:hover { background-color: #e5e7eb !important; }
         """
     ) as demo:
-
         chatbot = gr.Chatbot(
             label="💠 OhamLab Conversation",
             height=520,
@@ -223,7 +220,7 @@ def build_ui():
 
         with gr.Row():
             msg = gr.Textbox(
-                placeholder="Ask OhamLab anything ..",
+                placeholder="Ask OhamLab anything ...",
                 lines=3,
                 show_label=False,
                 scale=12,
@@ -237,7 +234,6 @@ def build_ui():
         send.click(chat_with_model, inputs=[msg, chatbot], outputs=[chatbot, msg])
         msg.submit(chat_with_model, inputs=[msg, chatbot], outputs=[chatbot, msg])
         clear.click(reset_chat, outputs=chatbot)
-
     return demo
 
 # ---------------------------