Update app.py

app.py

@@ -1,22 +1,25 @@
 #!/usr/bin/env python3
 """
-JusticeAI Backend — merged app.py with Image Editor endpoint
+JusticeAI Backend — merged app.py
 
-This is a consolidated FastAPI backend that:
+This is a consolidated, ready-to-run FastAPI backend that:
 - Keeps two databases: DATABASE_URL (personal user memory) and KNOWLEDGEDATABASE_URL (global knowledge).
-- Exposes /chat, /response, /add, /add-bulk, /leaderboard, /reembed, /model-status, /health, /metrics_stream,
+- Exposes /chat, /response, /add, /add-bulk, /leaderboard, /reembed, /model-status, /health, /metrics, /metrics_stream,
   /metrics_recent, /verify-admin, /cleardatabase.
-- Adds TTS endpoints (/speak and /speak_json) with preloading and optimizations when Coqui TTS is installed.
-- Adds image editor endpoints:
-    - /image_edit_json (JSON input with base64 image) — always available
-    - /image_edit (multipart/form-data with file upload and JSON "operations") — available only if python-multipart is installed
-  The editor supports operations: resize, rotate, crop, flip, flop, grayscale, blur, overlay_text.
-- Ensures user chats are stored only in DATABASE_URL.user_memory and never used to modify global knowledge.
+- Provides TTS voice-cloning via Coqui TTS with optimizations: preloaded model, GPU/half precision (if available),
+  speaker-sample caching, and a JSON fallback endpoint so the app can run without python-multipart installed.
+- Adds image editing endpoints: /image_edit_json (JSON with base64 image) and /image_edit (multipart if python-multipart installed).
+- Performs topic inference (Ollama-first if available), uses embeddings when available, and translates replies
+  into the user's detected language (language.py preferred; Helsinki transformers fallback).
+- Ensures user chats are stored only in DATABASE_URL.user_memory and never used to influence global knowledge.
 - Keeps last 10 chats per user (pruned on insert).
+- Adds /language.bin and /metrics endpoints to reduce 404 noise.
 
 Notes:
-- For multipart endpoints install python-multipart: pip install python-multipart
+- If you want multipart file upload support for /speak or /image_edit (form+file), install python-multipart:
+    pip install python-multipart
 - For image editing the Pillow package is required: pip install pillow
+- For TTS: install Coqui TTS and (optionally) torch for GPU/half precision speedups.
 """
 
 from sqlalchemy.pool import NullPool
@@ -36,7 +39,7 @@ import base64
 import io
 from datetime import datetime, timezone
 from collections import deque
-from typing import Optional, Dict, Any, List, Tuple
+from typing import Optional, Dict, Any, List
 
 from fastapi import (
     FastAPI, Request, Body, Query, Header, BackgroundTasks,
@@ -679,8 +682,7 @@ def _save_upload_file_tmp(upload_file: UploadFile) -> str:
 if TTS_AVAILABLE:
     threading.Thread(target=lambda: (get_tts_model_blocking()), daemon=True).start()
 
-# /speak_json and /speak endpoints handled earlier in previous merge; they remain present below.
-
+# /speak_json and /speak endpoints
 @app.post("/speak_json")
 async def speak_json(background_tasks: BackgroundTasks, payload: dict = Body(...)):
     text = payload.get("text", "")
@@ -883,7 +885,6 @@ def _perform_image_operations(input_path: str, operations: List[Dict[str, Any]],
             anchor = op.get("anchor", "lt")
             draw = ImageDraw.Draw(img)
             try:
-                # try to use a truetype font if available
                 font_path = op.get("font_path") or "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
                 font = ImageFont.truetype(font_path, size=size)
             except Exception:
@@ -894,7 +895,6 @@ def _perform_image_operations(input_path: str, operations: List[Dict[str, Any]],
             if mode:
                 img = img.convert(mode)
         # ignore unknown ops
-    # Save result. Determine format from out_path extension
     ext = os.path.splitext(out_path)[1].lower()
     fmt = None
     if ext in (".jpg", ".jpeg"):
@@ -905,7 +905,6 @@ def _perform_image_operations(input_path: str, operations: List[Dict[str, Any]],
         fmt = "WEBP"
     else:
         fmt = "PNG"
-    # If saving JPEG, convert to RGB
     save_img = img
     if fmt == "JPEG":
         save_img = img.convert("RGB")
@@ -938,7 +937,6 @@ async def image_edit_json(background_tasks: BackgroundTasks, payload: dict = Bod
     if not image_b64 and not image_url:
         raise HTTPException(status_code=400, detail="Provide either image_b64 or image_url")
 
-    # Save input image to temp file
     in_fd, in_path = tempfile.mkstemp(suffix=".input")
     os.close(in_fd)
     try:
@@ -950,7 +948,6 @@ async def image_edit_json(background_tasks: BackgroundTasks, payload: dict = Bod
             with open(in_path, "wb") as f:
                 f.write(data)
         else:
-            # download image_url (quick fetch)
             try:
                 resp = requests.get(image_url, timeout=10)
                 if resp.status_code != 200:
@@ -972,14 +969,12 @@ async def image_edit_json(background_tasks: BackgroundTasks, payload: dict = Bod
             pass
         raise HTTPException(status_code=500, detail="Failed to save input image")
 
-    # Prepare output path
     ext = "." + out_format if not out_format.startswith(".") else out_format
     out_fd, out_path = tempfile.mkstemp(suffix=ext, prefix="img_edit_out_")
     os.close(out_fd)
     background_tasks.add_task(lambda p: os.path.exists(p) and os.remove(p), out_path)
     background_tasks.add_task(lambda p: os.path.exists(p) and os.remove(p), in_path)
 
-    # Run ops in threadpool
     try:
         await _run_image_ops_in_thread(in_path, operations, out_path)
     except Exception as e:
@@ -1001,12 +996,6 @@ if HAVE_MULTIPART:
         image_url: Optional[str] = Form(None),
         format: Optional[str] = Form("png"),
     ):
-        """
-        Multipart endpoint to edit uploaded image with form-data.
-        operations: JSON string of operations (see image_edit_json)
-        image: uploaded file
-        image_url: alternative to upload
-        """
         if not PIL_AVAILABLE:
             raise HTTPException(status_code=503, detail="Image editing requires Pillow. Install with pip install pillow")
 
@@ -1073,7 +1062,7 @@ else:
         )
 
 # -------------------------
-# Metrics & small helpers
+# Metrics, language.bin, and small helpers
 # -------------------------
 recent_request_times = deque()
 recent_learning_timestamps = deque()
@@ -1097,13 +1086,40 @@ def record_learn_event():
     while recent_learning_timestamps and recent_learning_timestamps[0] < ts - 3600:
         recent_learning_timestamps.popleft()
 
+@app.get("/metrics")
+async def metrics():
+    try:
+        with engine_knowledge.connect() as c:
+            k = c.execute(sql_text("SELECT COUNT(*) FROM knowledge")).scalar() or 0
+    except Exception:
+        k = -1
+    try:
+        with engine_user.connect() as c:
+            u = c.execute(sql_text("SELECT COUNT(*) FROM user_memory")).scalar() or 0
+    except Exception:
+        u = -1
+    reqs_last_hour = sum(1 for ts, _ in recent_request_times if ts >= time.time() - 3600) if 'recent_request_times' in globals() else 0
+    return {
+        "ok": True,
+        "uptime_s": round(time.time() - app_start_time, 2) if 'app_start_time' in globals() else None,
+        "knowledge_count": int(k),
+        "user_memory_count": int(u),
+        "requests_last_hour": int(reqs_last_hour)
+    }
+
+@app.get("/language.bin")
+async def language_bin():
+    path = "language.bin"
+    if os.path.exists(path):
+        return FileResponse(path, media_type="application/octet-stream")
+    return JSONResponse(status_code=404, content={"error": "language.bin not found", "hint": "Place file at ./language.bin or upload it"})
+
 # -------------------------
 # Startup warmups
 # -------------------------
 @app.on_event("startup")
 async def startup_event():
     logger.info("[JusticeAI] startup: warming optional components")
-    # Warm embeddings in background
     if SentenceTransformer is not None:
         def warm_embed():
             try:
@@ -1111,7 +1127,6 @@ async def startup_event():
             except Exception as e:
                 logger.debug(f"[startup] embed warmup error: {e}")
         threading.Thread(target=warm_embed, daemon=True).start()
-    # Attempt Ollama pull if requested (best-effort)
     if OLLAMA_AUTO_PULL and ollama_cli_available():
         try:
             subprocess.run(["ollama", "pull", OLLAMA_MODEL], timeout=300)