Humanize bot responses for greetings, catalog questions, and empty RAG retrieval.

app/admin/templates/admin.html

@@ -678,7 +678,7 @@
         <div class="form-group"><label class="form-label">Response Style</label>
           <select class="form-input" id="p-style"><option value="balanced">Balanced</option><option value="formal">Formal</option><option value="casual">Casual</option><option value="enthusiastic">Enthusiastic</option></select>
         </div>
-        <div class="form-group"><label class="form-label">Fallback Message (when bot can't answer)</label><textarea class="form-input" id="p-fallback" placeholder="I want to give you the most accurate answer..."></textarea></div>
+        <div class="form-group"><label class="form-label">Fallback Message (when bot can't answer)</label><textarea class="form-input" id="p-fallback" placeholder="I'm not sure about that detail — but I can tell you plenty about the books. What would you like to know?"></textarea></div>
         <div class="form-group"><label class="form-label">Out-of-Scope Message (non-book questions)</label><textarea class="form-input" id="p-outofscope" placeholder="I'm specialized in discussing books..."></textarea></div>
         <button class="btn btn-primary" id="save-personality-btn">Save Personality Settings</button>
       </div>

app/models/user.py

@@ -44,7 +44,7 @@ class User(Base, TimestampMixin):
     )
     fallback_message: Mapped[str] = mapped_column(
         String(500),
-        default="I want to give you the most accurate answer. Could you rephrase that?",
+        default="I'm not sure about that specific detail — but I can tell you plenty about the books. What would you like to know?",
         nullable=False,
     )
     response_style: Mapped[str] = mapped_column(

app/services/prompter.py

@@ -109,6 +109,9 @@ Rules:
 
 INTENT_CLASSIFICATION_PROMPT = """Classify this reader message for a book sales chatbot.
 
+RECENT CONVERSATION:
+{history}
+
 MESSAGE: {query}
 
 Output ONLY a JSON object:
@@ -148,14 +151,24 @@ COMPETITOR_RESPONSE = """I'm specifically focused on {author_name}'s work, \
 so I can't speak to other authors. But I'd love to show you what makes \
 {author_name}'s approach different — what are you hoping a book will help you with?"""
 
-NO_CONTEXT_RESPONSE = """I want to give you the most accurate answer I can. \
-That specific detail isn't in what I have handy right now — but I'd love to \
-point you toward what *is* covered. What aspect of {topic} matters most to you?"""
+NO_CONTEXT_RESPONSE = """Hmm, I don't have that exact detail in front of me right now. \
+{book_hint}What would you like to know — the story, the themes, or who it's perfect for?"""
+
+HALLUCINATION_FALLBACK_RESPONSE = """I want to stay accurate rather than guess. \
+{book_hint}Ask me something specific — a character, a theme, or what the book helps with — \
+and I'll answer from what's actually in the text."""
+
+GREETING_RESPONSE = """Hey! I'm {bot_name}, {author_name}'s book advisor.
+
+{book_teaser}
+
+What are you in the mood for — a quick recommendation, or curious about a specific book?"""
+
+CATALOG_RESPONSE = """Here's what's in the library right now:
+
+{book_list}
 
-HALLUCINATION_FALLBACK_RESPONSE = """I want to make sure I'm giving you \
-accurate information. Let me point you to exactly where you can find this — \
-the answer is in {author_name}'s book, and I'd hate to paraphrase it poorly. \
-Is there something more specific I can help you find?"""
+Tell me which one catches your eye, or describe what you're looking for and I'll point you to the best fit."""
 
 TOKEN_EXHAUSTED_RESPONSE = "I'm taking a short break to recharge! Check back soon."
 

app/services/rag_pipeline.py

@@ -41,6 +41,7 @@ from app.services.prompter import (
     MASTER_SYSTEM_PROMPT,
     JAILBREAK_RESPONSE, OFF_TOPIC_RESPONSE,
     NO_CONTEXT_RESPONSE, HALLUCINATION_FALLBACK_RESPONSE,
+    GREETING_RESPONSE, CATALOG_RESPONSE,
 )
 from app.services.reranker import rerank_chunks
 from app.services.vector_store import retrieve_chunks
@@ -178,6 +179,13 @@ async def run_pipeline(
     if not active_books:
         return _no_books_response(start_ms)
 
+    # Short-circuit: greetings and catalog questions (no vector search needed)
+    if intent_result.intent == "greeting" or _is_greeting(query):
+        return _greeting_response(author, active_books, start_ms)
+
+    if intent_result.intent in ("meta", "comparison") or _is_catalog_question(query):
+        return _catalog_response(author, active_books, start_ms)
+
     # Resolve which book to search
     target_book_id = await _resolve_book(
         intent_result, session_context, active_books, author.id
@@ -208,7 +216,7 @@ async def run_pipeline(
 
     if not raw_chunks:
         log.warning("No chunks retrieved")
-        return _no_context_response(query, author, start_ms)
+        return _no_context_response(query, author, active_books, start_ms)
 
     # ── Step 6: Re-ranking ────────────────────────────────────────────────────
     top_chunks = await rerank_chunks(
@@ -219,7 +227,11 @@ async def run_pipeline(
     )
 
     if not top_chunks:
-        return _no_context_response(query, author, start_ms)
+        log.warning("Re-ranker filtered all chunks — using top retrieval results")
+        top_chunks = raw_chunks[: cfg.RAG_RERANK_TOP_N]
+
+    if not top_chunks:
+        return _no_context_response(query, author, active_books, start_ms)
 
     # ── Step 7: Context Assembly ───────────────────────────────────────────────
     context_str, context_tokens = build_context(top_chunks)
@@ -264,7 +276,7 @@ async def run_pipeline(
         is_faithful2, faithfulness_score = await check_faithfulness(raw_response, top_chunks)
         if not is_faithful2:
             raw_response = HALLUCINATION_FALLBACK_RESPONSE.format(
-                author_name=author.full_name or "the author"
+                book_hint=_book_hint(active_books),
             )
 
     # ── Step 10: Scope Check ──────────────────────────────────────────────────
@@ -426,9 +438,122 @@ def _boundary_response(text: str, start_ms: float, violation_type: str) -> Pipel
     )
 
 
-def _no_context_response(query: str, author: User, start_ms: float) -> PipelineResult:
+_CATALOG_PHRASES = (
+    "uploaded book", "your book", "what book", "which book", "list book",
+    "books do you", "books you have", "books available", "about the book",
+    "tell me about", "what is the book", "what's the book", "book about",
+    "in the catalog", "in your catalog",
+)
+
+_GREETINGS = frozenset({
+    "hi", "hello", "hey", "hiya", "howdy", "yo", "sup",
+    "good morning", "good afternoon", "good evening", "good day",
+})
+
+
+def _is_greeting(query: str) -> bool:
+    q = query.strip().lower().rstrip("!.?")
+    return q in _GREETINGS or (len(q) <= 12 and q.startswith(("hi ", "hey ", "hello ")))
+
+
+def _is_catalog_question(query: str) -> bool:
+    q = query.lower()
+    return any(phrase in q for phrase in _CATALOG_PHRASES)
+
+
+def _format_book_list(books: list) -> str:
+    lines = []
+    for book in books:
+        line = f"📖 **{book.title}**"
+        if book.tagline:
+            line += f" — {book.tagline}"
+        elif book.ai_summary:
+            summary = book.ai_summary.strip().replace("\n", " ")
+            line += f" — {summary[:160]}{'…' if len(summary) > 160 else ''}"
+        elif book.description:
+            desc = book.description.strip().replace("\n", " ")
+            line += f" — {desc[:160]}{'…' if len(desc) > 160 else ''}"
+        if book.status != "ready":
+            line += " *(indexing in progress)*"
+        lines.append(line)
+    return "\n".join(lines)
+
+
+def _book_hint(books: list) -> str:
+    if not books:
+        return ""
+    if len(books) == 1:
+        return f"I do know **{books[0].title}** well. "
+    return f"I have {len(books)} books I can walk you through. "
+
+
+def _greeting_response(author: User, books: list, start_ms: float) -> PipelineResult:
     elapsed_ms = int((time.monotonic() - start_ms) * 1000)
-    text = NO_CONTEXT_RESPONSE.format(topic=query[:50])
+    if len(books) == 1:
+        book_teaser = f"I can tell you everything about **{books[0].title}** — themes, characters, who it's for, all of it."
+    else:
+        titles = ", ".join(b.title for b in books[:3])
+        extra = f" and {len(books) - 3} more" if len(books) > 3 else ""
+        book_teaser = f"I know {len(books)} books inside out — including {titles}{extra}."
+    text = GREETING_RESPONSE.format(
+        bot_name=author.bot_name,
+        author_name=author.full_name or "the author",
+        book_teaser=book_teaser,
+    )
+    return PipelineResult(
+        response={"text": text, "links": [], "has_links": False},
+        intent="greeting",
+        response_ms=elapsed_ms,
+    )
+
+
+def _catalog_response(author: User, books: list, start_ms: float) -> PipelineResult:
+    elapsed_ms = int((time.monotonic() - start_ms) * 1000)
+    text = CATALOG_RESPONSE.format(book_list=_format_book_list(books))
+    return PipelineResult(
+        response={"text": text, "links": [], "has_links": False},
+        intent="meta",
+        response_ms=elapsed_ms,
+    )
+
+
+def _no_context_response(query: str, author: User, books: list, start_ms: float) -> PipelineResult:
+    elapsed_ms = int((time.monotonic() - start_ms) * 1000)
+
+    # Prefer book metadata when vector search came up empty
+    if len(books) == 1:
+        book = books[0]
+        if book.ai_summary and _is_catalog_question(query):
+            text = (
+                f"**{book.title}** — here's the gist:\n\n{book.ai_summary.strip()}\n\n"
+                "Want to go deeper on characters, themes, or who it's best for?"
+            )
+            return PipelineResult(
+                response={"text": text, "links": [], "has_links": False},
+                intent="question",
+                response_ms=elapsed_ms,
+            )
+        if book.status != "ready":
+            text = (
+                f"**{book.title}** is still being indexed — give it a minute and ask again. "
+                "I'll be able to answer detailed questions once processing finishes."
+            )
+            return PipelineResult(
+                response={"text": text, "links": [], "has_links": False},
+                intent="question",
+                response_ms=elapsed_ms,
+            )
+
+    if len(books) > 1 and _is_catalog_question(query):
+        return _catalog_response(author, books, start_ms)
+
+    default_robotic = "I want to give you the most accurate answer"
+    fallback = author.fallback_message or ""
+    if fallback and default_robotic not in fallback:
+        text = fallback
+    else:
+        text = NO_CONTEXT_RESPONSE.format(book_hint=_book_hint(books))
+
     return PipelineResult(
         response={"text": text, "links": [], "has_links": False},
         response_ms=elapsed_ms,