Updated for Game

app.py

@@ -7,6 +7,7 @@ from fastapi import FastAPI, HTTPException, BackgroundTasks, Request
 from fastapi.responses import StreamingResponse, Response
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
+from typing import Optional, Union, Dict, Any
 from llama_cpp import Llama
 from huggingface_hub import login, hf_hub_download
 import logging
@@ -188,16 +189,46 @@ except Exception as e:
     logger.error(f"Error loading cv_text.txt: {str(e)}")
     raise
 
-async def stream_response(query, history):
+async def stream_response(query, history, game_context=None, mode: Optional[str] = None):
     """Main streaming response function that routes to either Gemini or local model"""
     if USE_GEMINI:
-        async for chunk in stream_response_gemini(query, history):
+        async for chunk in stream_response_gemini(query, history, game_context, mode):
             yield chunk
     else:
-        async for chunk in stream_response_local(query, history):
+        async for chunk in stream_response_local(query, history, game_context, mode):
             yield chunk
 
-async def stream_response_gemini(query, history):
+def _format_game_context_for_prompt(game_context: Optional[Union[str, Dict[str, Any]]]) -> str:
+    """Return a concise text snippet to inject into the system prompt from game context."""
+    if not game_context:
+        return ""
+    try:
+        if isinstance(game_context, str):
+            return f"\nGAME CONTEXT: The player is currently at a station about {game_context}."
+        if isinstance(game_context, dict):
+            current = game_context.get('current_station') or game_context.get('station') or ''
+            visited = game_context.get('visited_stations') or []
+            context = game_context.get('context') or game_context.get('current_context') or ''
+            parts = ["\nGAME CONTEXT:"]
+            if current:
+                parts.append(f"Current station: {current}.")
+            if context:
+                parts.append(f"Station details: {context}.")
+            if visited:
+                try:
+                    uniq = []
+                    for v in visited:
+                        if v and v not in uniq:
+                            uniq.append(v)
+                    if uniq:
+                        parts.append(f"Visited stations so far: {', '.join(uniq)}.")
+                except Exception:
+                    pass
+            return " ".join(parts)
+    except Exception:
+        return ""
+
+async def stream_response_gemini(query, history, game_context=None, mode: Optional[str] = None):
     """Stream response using Gemini API with a proper system_instruction."""
     logger.info(f"Processing query with Gemini: {query}")
     start_time = time.time()
@@ -205,14 +236,26 @@ async def stream_response_gemini(query, history):
 
     # 1) Build your system prompt once
     current_date = datetime.now().strftime("%Y-%m-%d")
-    system_prompt = (
-        "You are Tim Luka Horstmann, a Computer Scientist. "
-        "A user is asking you a question. Respond as yourself, using the first person, in a friendly and concise manner. "
-        "For questions about your CV, base your answer *exclusively* on the provided CV information below and do not add any details not explicitly stated. "
-        "For casual questions not covered by the CV, respond naturally but limit answers to general truths about yourself (e.g., your current location is Paris, France) "
-        "and say 'I don't have specific details to share about that' if pressed for specifics beyond the CV or FAQs. "
-        f"Today's date is {current_date}. CV: {full_cv_text}"
-    )
+    game_context_text = _format_game_context_for_prompt(game_context)
+    is_game_mode = (mode == 'game') or bool(game_context_text)
+    if is_game_mode:
+        system_prompt = (
+            "You are Tim Luka Horstmann as a friendly in-game 'Station Guide'. "
+            "Stay in first person. Prioritize the current station context provided. "
+            "Use the CV for personal facts, roles and dates. "
+            "For non-personal or general questions (e.g., about cities or institutions), you may use general world knowledge. "
+            "Do not invent personal details beyond the CV/FAQs; if specifics are unknown, say so briefly. "
+            f"Today's date is {current_date}. CV: {full_cv_text}" + game_context_text
+        )
+    else:
+        system_prompt = (
+            "You are Tim Luka Horstmann, a Computer Scientist. "
+            "A user is asking you a question. Respond as yourself, using the first person, in a friendly and concise manner. "
+            "For questions about your CV, base your answer *exclusively* on the provided CV information below and do not add any details not explicitly stated. "
+            "For casual questions not covered by the CV, respond naturally but limit answers to general truths about yourself (e.g., your current location is Paris, France) "
+            "and say 'I don't have specific details to share about that' if pressed for specifics beyond the CV or FAQs. "
+            f"Today's date is {current_date}. CV: {full_cv_text}"
+        )
 
     # 2) Build only user/model history as `contents`
     contents = []
@@ -272,7 +315,7 @@ async def stream_response_gemini(query, history):
         yield "data: [DONE]\n\n"
 
 
-async def stream_response_local(query, history):
+async def stream_response_local(query, history, game_context=None, mode: Optional[str] = None):
     """Stream response using local model"""
     logger.info(f"Processing query with local model: {query}")
     start_time = time.time()
@@ -280,14 +323,25 @@ async def stream_response_local(query, history):
 
     current_date = datetime.now().strftime("%Y-%m-%d")
     
-    system_prompt = (
-        "/no_think You are Tim Luka Horstmann, a Computer Scientist. A user is asking you a question. Respond as yourself, using the first person, in a friendly and concise manner. "
-        "For questions about your CV, base your answer *exclusively* on the provided CV information below and do not add any details not explicitly stated. "
-        "For casual questions not covered by the CV, respond naturally but limit answers to general truths about yourself (e.g., your current location is Paris, France, or your field is AI) "
-        "and say 'I don't have specific details to share about that' if pressed for specifics beyond the CV or FAQs. Do not invent facts, experiences, or opinions not supported by the CV or FAQs. "
-        f"Today's date is {current_date}. "
-        f"CV: {full_cv_text}"
-    )
+    game_context_text = _format_game_context_for_prompt(game_context)
+    is_game_mode = (mode == 'game') or bool(game_context_text)
+    if is_game_mode:
+        system_prompt = (
+            "/no_think You are Tim Luka Horstmann as a friendly in-game 'Station Guide'. "
+            "Stay in first person. Prioritize the current station context provided. "
+            "Use the CV for personal facts, roles and dates. "
+            "For non-personal or general questions (e.g., about cities or institutions), you may use general world knowledge. "
+            "Do not invent personal details beyond the CV/FAQs; if specifics are unknown, say so briefly. "
+            f"Today's date is {current_date}. CV: {full_cv_text}" + game_context_text
+        )
+    else:
+        system_prompt = (
+            "/no_think You are Tim Luka Horstmann, a Computer Scientist. A user is asking you a question. Respond as yourself, using the first person, in a friendly and concise manner. "
+            "For questions about your CV, base your answer *exclusively* on the provided CV information below and do not add any details not explicitly stated. "
+            "For casual questions not covered by the CV, respond naturally but limit answers to general truths about yourself (e.g., your current location is Paris, France, or your field is AI) "
+            "and say 'I don't have specific details to share about that' if pressed for specifics beyond the CV or FAQs. Do not invent facts, experiences, or opinions not supported by the CV or FAQs. "
+            f"Today's date is {current_date}. CV: {full_cv_text}"
+        )
 
     if not isinstance(system_prompt, str):
         system_prompt = str(system_prompt)
@@ -340,6 +394,8 @@ async def stream_response_local(query, history):
 class QueryRequest(BaseModel):
     query: str
     history: list
+    game_context: Optional[Union[str, Dict[str, Any]]] = None
+    mode: Optional[str] = None
 
 class TTSRequest(BaseModel):
     text: str
@@ -363,7 +419,9 @@ def get_ram_usage():
 async def predict(request: Request, query_request: QueryRequest):
     query = query_request.query
     history = query_request.history
-    return StreamingResponse(stream_response(query, history), media_type="text/event-stream")
+    game_context = query_request.game_context
+    mode = (query_request.mode or '').lower() or None
+    return StreamingResponse(stream_response(query, history, game_context, mode), media_type="text/event-stream")
 
 @app.post("/api/tts")
 @limiter.limit("5/minute")  # Allow 5 TTS requests per minute per IP
@@ -490,4 +548,4 @@ async def keep_model_warm():
             logger.error(f"Error in periodic warm-up: {str(e)}")
         
         # Wait for 13 minutes before the next warm-up
-        await asyncio.sleep(13 * 60)
+        await asyncio.sleep(13 * 60)