merge: chore/cleanup-remove-bloat-and-secrets into main

app/routers/moderation.py

@@ -0,0 +1,116 @@
+"""Content Moderation Pipeline — AI-powered spam/scam/NSFW detection for user content."""
+
+import os
+import re
+
+import httpx
+from fastapi import APIRouter
+from pydantic import BaseModel
+
+router = APIRouter(prefix="/api/v1/moderation", tags=["moderation"])
+
+OLLAMA = os.getenv("OLLAMA_HOST", "http://localhost:11434")
+
+BLOCKED_PATTERNS = [
+    (r"(?i)(buy|sell|trade).*\b(signal|call)\b", "trading_signal"),
+    (r"(?i)(\b\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}\b)", "ip_address"),
+    (r"(?i)\b(0x[a-fA-F0-9]{40})\b.*\b(private.?key|seed.?phrase|mnemonic|password)\b", "wallet_phishing"),
+    (r"(?i)(airdrop|giveaway|free.*token).*\b(claim|connect|verify)\b", "scam_airdrop"),
+    (r"(https?://(?!rugmunch\.io|polymarket\.com|dexscreener\.com)[^\s]+)", "external_link"),
+]
+
+
+class ModerationRequest(BaseModel):
+    text: str
+    user_id: str = "anonymous"
+    context: str = "comment"  # comment, post, review, chat
+
+
+class ModerationResult(BaseModel):
+    approved: bool
+    risk_score: int  # 0-100
+    flags: list[str]
+    reason: str = ""
+
+
+async def _ai_classify(text: str) -> dict:
+    """Use Ollama to classify content."""
+    try:
+        async with httpx.AsyncClient(timeout=15) as c:
+            r = await c.post(
+                f"{OLLAMA}/api/generate",
+                json={
+                    "model": "qwen2.5-coder:7b",
+                    "prompt": f"Classify this crypto-related content as SAFE, SPAM, SCAM, or NSFW. Answer with one word only.\n\nContent: {text[:500]}\n\nClassification:",
+                    "stream": False,
+                    "options": {"num_predict": 5, "temperature": 0.1},
+                },
+            )
+            if r.status_code == 200:
+                response = r.json().get("response", "").strip().upper()
+                return {"ai_verdict": response, "ai_used": True}
+    except Exception:
+        pass
+    return {"ai_verdict": "UNKNOWN", "ai_used": False}
+
+
+@router.post("/check")
+async def moderate_content(req: ModerationRequest):
+    """Check content for spam, scams, NSFW, and policy violations."""
+    flags = []
+    risk = 0
+
+    # Pattern-based detection
+    for pattern, flag_type in BLOCKED_PATTERNS:
+        if re.search(pattern, req.text):
+            flags.append(flag_type)
+            risk += 25
+
+    # Length checks
+    if len(req.text) < 5:
+        flags.append("too_short")
+        risk += 10
+    if len(req.text) > 10000:
+        flags.append("too_long")
+        risk += 5
+
+    # AI classification
+    ai = await _ai_classify(req.text)
+    ai_verdict = ai["ai_verdict"]
+    if "SCAM" in ai_verdict:
+        flags.append("ai_scam")
+        risk += 40
+    elif "SPAM" in ai_verdict:
+        flags.append("ai_spam")
+        risk += 25
+    elif "NSFW" in ai_verdict:
+        flags.append("ai_nsfw")
+        risk += 50
+
+    approved = risk < 40
+    reason = "Content approved" if approved else f"Flagged: {', '.join(flags)}"
+
+    return {
+        "approved": approved,
+        "risk_score": min(100, risk),
+        "flags": flags,
+        "reason": reason,
+        "ai_classification": ai_verdict,
+    }
+
+
+@router.get("/stats")
+async def moderation_stats():
+    return {
+        "patterns_checked": len(BLOCKED_PATTERNS),
+        "ai_model": "qwen2.5-coder:7b",
+        "categories": [
+            "trading_signal",
+            "wallet_phishing",
+            "scam_airdrop",
+            "external_link",
+            "ai_scam",
+            "ai_spam",
+            "ai_nsfw",
+        ],
+    }