File size: 10,548 Bytes
732e77c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 | # logged_in_bot/tools.py
"""
Utilities for the logged-in chatbot flow.
Features
- PII redaction (optional) via guardrails.pii_redaction
- Sentiment (Azure via importlib if configured; falls back to heuristic)
- Tiny intent router: help | remember | forget | list memory | summarize | echo | chat
- Session history capture via memory.sessions
- Lightweight RAG context via memory.rag.retriever (TF-IDF)
- Deterministic, dependency-light; safe to import in any environment
"""
from __future__ import annotations
from dataclasses import asdict, dataclass
from typing import Any, Dict, List, Optional, Tuple
import os
import re
# -------------------------
# Optional imports (safe)
# -------------------------
# Guardrails redaction (optional)
try: # pragma: no cover
from guardrails.pii_redaction import redact as pii_redact # type: ignore
except Exception: # pragma: no cover
pii_redact = None # type: ignore
# Fallback PlainChatResponse if core.types is absent
try: # pragma: no cover
from core.types import PlainChatResponse # dataclass with .to_dict()
except Exception: # pragma: no cover
@dataclass
class PlainChatResponse: # lightweight fallback shape
reply: str
meta: Optional[Dict[str, Any]] = None
def to_dict(self) -> Dict[str, Any]:
return asdict(self)
# Sentiment (unified; Azure if configured; otherwise heuristic)
try:
from agenticcore.providers_unified import analyze_sentiment_unified as _sent
except Exception:
def _sent(t: str) -> Dict[str, Any]:
t = (t or "").lower()
if any(w in t for w in ["love","great","awesome","good","thanks","glad","happy"]): return {"label":"positive","score":0.9,"backend":"heuristic"}
if any(w in t for w in ["hate","terrible","awful","bad","angry","sad"]): return {"label":"negative","score":0.9,"backend":"heuristic"}
return {"label":"neutral","score":0.5,"backend":"heuristic"}
# Memory + RAG (pure-Python, no extra deps)
try:
from memory.sessions import SessionStore
except Exception as e: # pragma: no cover
raise RuntimeError("memory.sessions is required for logged_in_bot.tools") from e
try:
from memory.profile import Profile
except Exception as e: # pragma: no cover
raise RuntimeError("memory.profile is required for logged_in_bot.tools") from e
try:
from memory.rag.indexer import DEFAULT_INDEX_PATH
from memory.rag.retriever import retrieve, Filters
except Exception as e: # pragma: no cover
raise RuntimeError("memory.rag.{indexer,retriever} are required for logged_in_bot.tools") from e
History = List[Tuple[str, str]] # [("user","..."), ("bot","...")]
# -------------------------
# Session store shim
# -------------------------
def _get_store():
"""Some versions expose SessionStore.default(); others don’t. Provide a shim."""
try:
if hasattr(SessionStore, "default") and callable(getattr(SessionStore, "default")):
return SessionStore.default()
except Exception:
pass
# Fallback: module-level singleton
if not hasattr(_get_store, "_singleton"):
_get_store._singleton = SessionStore()
return _get_store._singleton
# -------------------------
# Helpers
# -------------------------
_WHITESPACE_RE = re.compile(r"\s+")
def sanitize_text(text: str) -> str:
"""Basic sanitize/normalize; keep CPU-cheap & deterministic."""
text = (text or "").strip()
text = _WHITESPACE_RE.sub(" ", text)
max_len = int(os.getenv("MAX_INPUT_CHARS", "4000"))
if len(text) > max_len:
text = text[:max_len] + "…"
return text
def redact_text(text: str) -> str:
"""Apply optional PII redaction if available; otherwise return text."""
if pii_redact:
try:
return pii_redact(text)
except Exception:
return text
return text
def _simple_sentiment(text: str) -> Dict[str, Any]:
t = (text or "").lower()
pos = any(w in t for w in ["love", "great", "awesome", "good", "thanks", "glad", "happy"])
neg = any(w in t for w in ["hate", "terrible", "awful", "bad", "angry", "sad"])
label = "positive" if pos and not neg else "negative" if neg and not pos else "neutral"
score = 0.8 if label != "neutral" else 0.5
return {"label": label, "score": score, "backend": "heuristic"}
def _sentiment_meta(text: str) -> Dict[str, Any]:
try:
res = _sent(text)
# Normalize common shapes
if isinstance(res, dict):
label = str(res.get("label", "neutral"))
score = float(res.get("score", 0.5))
backend = str(res.get("backend", res.get("provider", "azure")))
return {"label": label, "score": score, "backend": backend}
except Exception:
pass
return _simple_sentiment(text)
def intent_of(text: str) -> str:
"""Tiny intent classifier."""
t = (text or "").lower().strip()
if not t:
return "empty"
if t in {"help", "/help", "capabilities"}:
return "help"
if t.startswith("remember ") and ":" in t:
return "memory_remember"
if t.startswith("forget "):
return "memory_forget"
if t == "list memory":
return "memory_list"
if t.startswith("summarize ") or t.startswith("summarise ") or " summarize " in f" {t} ":
return "summarize"
if t.startswith("echo "):
return "echo"
return "chat"
def summarize_text(text: str, target_len: int = 120) -> str:
m = re.split(r"(?<=[.!?])\s+", (text or "").strip())
first = m[0] if m else (text or "").strip()
if len(first) <= target_len:
return first
return first[: target_len - 1].rstrip() + "…"
def capabilities() -> List[str]:
return [
"help",
"remember <key>: <value>",
"forget <key>",
"list memory",
"echo <text>",
"summarize <paragraph>",
"sentiment tagging (logged-in mode)",
]
def _handle_memory_cmd(user_id: str, text: str) -> Optional[str]:
prof = Profile.load(user_id)
m = re.match(r"^\s*remember\s+([^:]+)\s*:\s*(.+)$", text, flags=re.I)
if m:
key, val = m.group(1).strip(), m.group(2).strip()
prof.remember(key, val)
return f"Okay, I'll remember **{key}**."
m = re.match(r"^\s*forget\s+(.+?)\s*$", text, flags=re.I)
if m:
key = m.group(1).strip()
return "Forgot." if prof.forget(key) else f"I had nothing stored as **{key}**."
if re.match(r"^\s*list\s+memory\s*$", text, flags=re.I):
keys = prof.list_notes()
return "No saved memory yet." if not keys else "Saved keys: " + ", ".join(keys)
return None
def _retrieve_context(query: str, k: int = 4) -> List[str]:
passages = retrieve(query, k=k, index_path=DEFAULT_INDEX_PATH, filters=None)
return [p.text for p in passages]
# -------------------------
# Main entry
# -------------------------
def handle_logged_in_turn(message: str, history: Optional[History], user: Optional[dict]) -> Dict[str, Any]:
"""
Process one user turn in 'logged-in' mode.
Returns a PlainChatResponse (dict) with:
- reply: str
- meta: { intent, sentiment: {...}, redacted: bool, input_len: int }
"""
history = history or []
user_text_raw = message or ""
user_text = sanitize_text(user_text_raw)
# Redaction (if configured)
redacted_text = redact_text(user_text)
redacted = (redacted_text != user_text)
it = intent_of(redacted_text)
# Compute sentiment once (always attach — satisfies tests)
sentiment = _sentiment_meta(redacted_text)
# ---------- route ----------
if it == "empty":
reply = "Please type something. Try 'help' for options."
meta = _meta(redacted, it, redacted_text, sentiment)
return PlainChatResponse(reply=reply, meta=meta).to_dict()
if it == "help":
reply = "I can:\n" + "\n".join(f"- {c}" for c in capabilities())
meta = _meta(redacted, it, redacted_text, sentiment)
return PlainChatResponse(reply=reply, meta=meta).to_dict()
if it.startswith("memory_"):
user_id = (user or {}).get("id") or "guest"
mem_reply = _handle_memory_cmd(user_id, redacted_text)
reply = mem_reply or "Sorry, I didn't understand that memory command."
# track in session
sess = _get_store().get(user_id)
sess.append({"role": "user", "text": user_text})
sess.append({"role": "assistant", "text": reply})
meta = _meta(redacted, "memory_cmd", redacted_text, sentiment)
return PlainChatResponse(reply=reply, meta=meta).to_dict()
if it == "echo":
payload = redacted_text.split(" ", 1)[1] if " " in redacted_text else ""
reply = payload or "(nothing to echo)"
meta = _meta(redacted, it, redacted_text, sentiment)
return PlainChatResponse(reply=reply, meta=meta).to_dict()
if it == "summarize":
if redacted_text.lower().startswith("summarize "):
payload = redacted_text.split(" ", 1)[1]
elif redacted_text.lower().startswith("summarise "):
payload = redacted_text.split(" ", 1)[1]
else:
payload = redacted_text
reply = summarize_text(payload)
meta = _meta(redacted, it, redacted_text, sentiment)
return PlainChatResponse(reply=reply, meta=meta).to_dict()
# default: chat (with RAG)
user_id = (user or {}).get("id") or "guest"
ctx_chunks = _retrieve_context(redacted_text, k=4)
if ctx_chunks:
reply = "Here's what I found:\n- " + "\n- ".join(
c[:220].replace("\n", " ") + ("…" if len(c) > 220 else "") for c in ctx_chunks
)
else:
reply = "I don’t see anything relevant in your documents. Ask me to index files or try a different query."
# session trace
sess = _get_store().get(user_id)
sess.append({"role": "user", "text": user_text})
sess.append({"role": "assistant", "text": reply})
meta = _meta(redacted, it, redacted_text, sentiment)
return PlainChatResponse(reply=reply, meta=meta).to_dict()
# -------------------------
# Internals
# -------------------------
def _meta(redacted: bool, intent: str, redacted_text: str, sentiment: Dict[str, Any]) -> Dict[str, Any]:
return {
"intent": intent,
"redacted": redacted,
"input_len": len(redacted_text),
"sentiment": sentiment,
}
__all__ = [
"handle_logged_in_turn",
"sanitize_text",
"redact_text",
"intent_of",
"summarize_text",
"capabilities",
]
|