update chatbot

assistant.py

@@ -0,0 +1,43 @@
+# chat_chain.py
+import os
+from typing import Optional
+
+from langchain_openai import ChatOpenAI
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.output_parsers import StrOutputParser
+
+
+# ---- Configuration (via env, with safe defaults) ----
+HF_TOKEN: Optional[str] = os.environ.get("HF_TOKEN")
+MODEL_ID: str = os.environ.get("MODEL_ID", "openai/gpt-oss-20b:nebius")
+BASE_URL: str = os.environ.get("BASE_URL", "https://router.huggingface.co/v1")
+TEMP: float = float(os.environ.get("TEMPERATURE", "0.2"))
+
+
+# ---- Build the chain once (module-level cache) ----
+if not HF_TOKEN:
+    raise RuntimeError("HF_TOKEN is not set. Add it to your environment or Spaces → Settings → Secrets.")
+
+_llm = ChatOpenAI(
+    model=MODEL_ID,
+    api_key=HF_TOKEN,
+    base_url=BASE_URL,
+    temperature=TEMP,
+)
+
+_prompt = ChatPromptTemplate.from_messages([
+    ("system", "You are a helpful, precise assistant. Reply in simple, neutral English."),
+    ("user", "{message}")
+])
+
+_chain = _prompt | _llm | StrOutputParser()
+
+
+def get_answer(message: str) -> str:
+    """
+    Generate a single reply for the given user message.
+    Keeps LangChain initialization separate from the web layer.
+    """
+    if not message or not message.strip():
+        raise ValueError("message cannot be empty.")
+    return _chain.invoke({"message": message.strip()})

pytrade.py

@@ -16,17 +16,7 @@ import os
 import time
 import requests
 from typing import List, Dict
-from chatbot import (
-        normalize_entities,
-        needs_live_context,
-        pick_is_news,
-        serpapi_search,
-        compose_live_user_prompt,
-        compose_general_user_prompt,
-        call_llm,
-        BASE_SYSTEM_PROMPT,
-        SEARCH_TOPK
-    )
+from assistant import get_answer
 
 app = Flask(__name__)
 
@@ -125,56 +115,18 @@ def analyze_all():
 
 @app.route("/chat", methods=["POST"])
 def chat():
-    """
-    Request JSON:
-      { "message": "your question" }
-      or
-      { "question": "your question" }
-
-    Response JSON:
-      {
-        "answer": "...",
-        "live": true/false,
-        "sources": [{title, link, source, snippet}]
-      }
-    """
-    data = request.get_json(force=True, silent=True) or {}
-    message = (data.get("message") or data.get("question") or "").strip()
-
-    if not message:
-        return jsonify({"error": "message or question is required"}), 400
-
-    # Normalize common aliases (e.g., TCS -> Tata Consultancy Services)
-    message = normalize_entities(message)
-
-    # Decide if this needs live context
-    live = needs_live_context(message)
-
-    hits: List[Dict[str, str]] = []
-    if live:
-        is_news = pick_is_news(message)
-        try:
-            hits = serpapi_search(message, is_news=is_news, num=SEARCH_TOPK)
-        except Exception:
-            hits = []
-            live = False
+    data = request.get_json(silent=True) or {}
+    user_message = (data.get("message") or "").strip()
+    if not user_message:
+        return jsonify({"error": "message is required"}), 400
 
     try:
-        if live and hits:
-            user_prompt = compose_live_user_prompt(message, hits)
-            answer = call_llm(BASE_SYSTEM_PROMPT, user_prompt)
-            return jsonify({"answer": answer, "live": True, "sources": hits})
-        else:
-            user_prompt = compose_general_user_prompt(message)
-            answer = call_llm(BASE_SYSTEM_PROMPT, user_prompt)
-            return jsonify({"answer": answer, "live": False, "sources": []})
+        reply = get_answer(user_message)
+        return jsonify({"answer": reply})
     except Exception as e:
-        return jsonify({
-            "error": "LLM call failed",
-            "details": str(e),
-            "live": live,
-            "sources": hits
-        }), 500
+        # Keep error neutral and simple for the client
+        return jsonify({"error": str(e)}), 500
+
 
 if __name__ == "__main__":
     # Default to 5000 locally; on Hugging Face Spaces the platform injects PORT.

requirements.txt

@@ -21,4 +21,6 @@ gunicorn
 torch
 python-dotenv
 openai>=1.0.0
+langchain==0.3.4
+langchain-openai==0.2.3