Update app.py

app.py

@@ -1,7 +1,8 @@
+# app.py
 import os, traceback, regex as re2
 import gradio as gr
 import pandas as pd
-from typing import List, Tuple, Dict
+from typing import List, Tuple, Dict, Any
 
 from settings import HEALTHCARE_SETTINGS, GENERAL_CONVERSATION_PROMPT, USE_SCENARIO_ENGINE
 from audit_log import log_event
@@ -12,20 +13,23 @@ from healthcare_analysis import HealthcareAnalyzer
 from scenario_planner import parse_to_plan
 from scenario_engine import ScenarioEngine
 from rag import RAGIndex
-from llm_router import generate_narrative, cohere_chat
+from llm_router import generate_narrative, cohere_chat, open_fallback_chat
 
 def _sanitize_text(s: str) -> str:
     if not isinstance(s, str): return s
+    # strip control chars (keep newlines/tabs)
     return re2.sub(r'[\p{C}--[\n\t]]+', '', s)
 
-def _dataset_catalog(results: Dict[str, any]) -> Dict[str, List[str]]:
-    cat = {}
+def _dataset_catalog(results: Dict[str, Any]) -> Dict[str, List[str]]:
+    """Expose available columns per dataset to the planner."""
+    cat: Dict[str, List[str]] = {}
     for k, v in results.items():
         if isinstance(v, pd.DataFrame):
             cat[k] = v.columns.tolist()
     return cat
 
 def is_healthcare_scenario(text: str, has_files: bool) -> bool:
+    """Heuristic: scenario mode when user provided files + scenario-ish text."""
     t = (text or "").lower()
     kws = HEALTHCARE_SETTINGS["healthcare_keywords"]
     structured = any(s in t for s in ["background", "situation", "tasks", "deliverables"])
@@ -34,43 +38,47 @@ def is_healthcare_scenario(text: str, has_files: bool) -> bool:
 def handle(user_msg: str, history: list, files: list) -> Tuple[list, str]:
     try:
         safe_in, blocked_in, reason_in = safety_filter(user_msg, mode="input")
-        if blocked_in: return history + [(user_msg, refusal_reply(reason_in))], ""
+        if blocked_in:
+            return history + [(user_msg, refusal_reply(reason_in))], ""
 
-        # Normalize files -> paths
+        # Normalize files -> paths (safe when files is None)
         file_paths = [getattr(f, "name", None) or f for f in (files or [])]
 
-        # Register CSVs
+        # Register CSVs into the registry
         registry = DataRegistry()
         for p in file_paths:
-            try: registry.add_path(p)
-            except Exception as e: log_event("ingest_error", None, {"file": p, "err": str(e)})
+            try:
+                registry.add_path(p)
+            except Exception as e:
+                log_event("ingest_error", None, {"file": p, "err": str(e)})
 
-        # RAG ingest (safe on empty)
+        # RAG ingest (best-effort, text only; safe on empty)
         rag = RAGIndex()
         ing = extract_text_from_files(file_paths)
         rag.add(ing.get("chunks", []))
 
+        # Scenario mode: plan -> deterministic execution -> narrative
         if is_healthcare_scenario(safe_in, bool(file_paths)) and USE_SCENARIO_ENGINE:
             analyzer = HealthcareAnalyzer(registry)
-            datasets = analyzer.comprehensive_analysis(safe_in)
+            datasets = analyzer.comprehensive_analysis(safe_in)  # expose dataframes by filename
             catalog = _dataset_catalog(datasets)
 
-            # LLM → plan (no hardcoding)
+            # 1) LLM parses scenario into a plan (scenario-agnostic, no hardcoding)
             plan = parse_to_plan(safe_in, catalog)
 
-            # Deterministic execution
+            # 2) Deterministic execution of the plan (pandas-based)
             structured_md = ScenarioEngine.execute_plan(plan, datasets)
 
-            # Narrative with Canadian grounding
+            # 3) Canadian grounding + narrative (Cohere primary, open-model fallback)
             rag_hits = [txt for txt, _ in rag.retrieve(safe_in, k=6)]
             narrative = generate_narrative(safe_in, structured_md, rag_hits)
 
             final = f"{structured_md}\n\n# Narrative & Recommendations\n\n{narrative}"
             return history + [(user_msg, _sanitize_text(final))], ""
 
-        # General conversation (Cohere primary, open-model fallback inside cohere_chat if needed)
+        # General conversation mode (no scenario/files required)
         prompt = f"{GENERAL_CONVERSATION_PROMPT}\n\nUser: {safe_in}\nAssistant:"
-        ans = cohere_chat(prompt) or "How can I help further?"
+        ans = cohere_chat(prompt) or open_fallback_chat(prompt) or "How can I help further?"
         return history + [(user_msg, _sanitize_text(ans))], ""
 
     except Exception as e:
@@ -78,10 +86,16 @@ def handle(user_msg: str, history: list, files: list) -> Tuple[list, str]:
         log_event("app_error", None, {"err": str(e), "tb": tb})
         return history + [(user_msg, f"Error: {e}\n\n{tb}")], ""
 
+# -------- UI --------
 with gr.Blocks(analytics_enabled=False) as demo:
     gr.Markdown("## Canadian Healthcare AI • Scenario-Agnostic (Cohere primary • Deterministic analytics)")
-    chat = gr.Chatbot(type="tuple", height=520)  # tuple mode (matches how we store history)
-    files = gr.Files(file_count="multiple", type="filepath", file_types=HEALTHCARE_SETTINGS["supported_file_types"])
+    # tuple mode matches how we store history as (user, assistant)
+    chat = gr.Chatbot(type="tuple", height=520)
+    files = gr.Files(
+        file_count="multiple",
+        type="filepath",
+        file_types=HEALTHCARE_SETTINGS["supported_file_types"]
+    )
     msg = gr.Textbox(placeholder="Paste any scenario (Background / Situation / Tasks / Deliverables) or just chat.")
     send = gr.Button("Send")
     clear = gr.Button("Clear")
@@ -95,5 +109,5 @@ with gr.Blocks(analytics_enabled=False) as demo:
     clear.click(lambda: ([], ""), outputs=[chat, msg])
 
 if __name__ == "__main__":
+    # You can set share=True if you want a public link in dev.
     demo.launch(server_name="0.0.0.0", server_port=int(os.getenv("PORT", "7860")))
-