ko

.idea/misc.xml

@@ -3,5 +3,5 @@
   <component name="Black">
     <option name="sdkName" value="Python 3.14" />
   </component>
-  <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.14" project-jdk-type="Python SDK" />
+  <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.11 (son-of-anton)" project-jdk-type="Python SDK" />
 </project>
\ No newline at end of file

.idea/son-of-anton.iml

 <?xml version="1.0" encoding="UTF-8"?>
 <module type="PYTHON_MODULE" version="4">
   <component name="NewModuleRootManager">
-    <content url="file://$MODULE_DIR$" />
-    <orderEntry type="jdk" jdkName="Python 3.14" jdkType="Python SDK" />
+    <content url="file://$MODULE_DIR$">
+      <excludeFolder url="file://$MODULE_DIR$/.venv" />
+    </content>
+    <orderEntry type="jdk" jdkName="Python 3.11 (son-of-anton)" jdkType="Python SDK" />
     <orderEntry type="sourceFolder" forTests="false" />
   </component>
 </module>
\ No newline at end of file

backend/main.py

@@ -21,10 +21,30 @@ from backend.routes.files_routes import router as files_router
 from backend.services.bedrock_service import close_http_client
 
 
+def _run_migrations():
+    """Add new columns to existing tables if they're missing (lightweight migration)."""
+    from sqlalchemy import inspect, text
+    try:
+        inspector = inspect(engine)
+        if "chats" in inspector.get_table_names():
+            columns = {c["name"] for c in inspector.get_columns("chats")}
+            with engine.connect() as conn:
+                if "max_tokens" not in columns:
+                    conn.execute(text("ALTER TABLE chats ADD COLUMN max_tokens INTEGER DEFAULT 4096"))
+                    print("  ✅ Added chats.max_tokens column")
+                if "reasoning_budget" not in columns:
+                    conn.execute(text("ALTER TABLE chats ADD COLUMN reasoning_budget INTEGER DEFAULT 0"))
+                    print("  ✅ Added chats.reasoning_budget column")
+                conn.commit()
+    except Exception as e:
+        print(f"  ⚠️  Migration note: {e}")
+
+
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     # --- Startup ---
     Base.metadata.create_all(bind=engine)
+    _run_migrations()
     seed_superadmin()
     print("🔥 Son of Anton is online.")
     yield

backend/models.py

@@ -49,6 +49,8 @@ class Chat(Base):
     title = Column(String(200), default="New Chat")
     model = Column(String(100), default="eu.anthropic.claude-opus-4-6-v1")
     knowledge_base_id = Column(String(36), nullable=True)
+    max_tokens = Column(Integer, default=4096)
+    reasoning_budget = Column(Integer, default=0)
     created_at = Column(DateTime, default=datetime.utcnow)
     updated_at = Column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
 

backend/routes/chat_routes.py

@@ -24,10 +24,16 @@ class CreateChatBody(BaseModel):
     title: str = "New Chat"
     model: str = "eu.anthropic.claude-opus-4-6-v1"
     knowledge_base_id: Optional[str] = None
+    max_tokens: int = 4096
+    reasoning_budget: int = 0
 
 
-class RenameChatBody(BaseModel):
-    title: str
+class UpdateChatBody(BaseModel):
+    title: Optional[str] = None
+    model: Optional[str] = None
+    max_tokens: Optional[int] = None
+    reasoning_budget: Optional[int] = None
+    knowledge_base_id: Optional[str] = None
 
 
 class SendMessageBody(BaseModel):
@@ -57,7 +63,9 @@ def create_chat(body: CreateChatBody, user: User = Depends(get_current_user), db
         user_id=user.id,
         title=body.title,
         model=body.model,
-        knowledge_base_id=body.knowledge_base_id,
+        knowledge_base_id=body.knowledge_base_id or None,
+        max_tokens=body.max_tokens,
+        reasoning_budget=body.reasoning_budget,
     )
     db.add(chat)
     db.commit()
@@ -74,11 +82,21 @@ def get_chat(chat_id: str, user: User = Depends(get_current_user), db: Session =
 
 
 @router.put("/{chat_id}")
-def rename_chat(chat_id: str, body: RenameChatBody, user: User = Depends(get_current_user), db: Session = Depends(get_db)):
+def update_chat(chat_id: str, body: UpdateChatBody, user: User = Depends(get_current_user), db: Session = Depends(get_db)):
     chat = db.query(Chat).filter(Chat.id == chat_id, Chat.user_id == user.id).first()
     if not chat:
         raise HTTPException(404)
-    chat.title = body.title
+    if body.title is not None:
+        chat.title = body.title
+    if body.model is not None:
+        chat.model = body.model
+    if body.max_tokens is not None:
+        chat.max_tokens = body.max_tokens
+    if body.reasoning_budget is not None:
+        chat.reasoning_budget = body.reasoning_budget
+    if body.knowledge_base_id is not None:
+        # Empty string means "clear the KB"
+        chat.knowledge_base_id = body.knowledge_base_id or None
     db.commit()
     return _chat_dict(chat)
 
@@ -219,6 +237,12 @@ async def send_message(
             db.add(assistant_msg)
 
             db_user.tokens_used_this_month += input_tokens + output_tokens
+
+            # Persist the generation settings used for this message onto the chat
+            chat.model = model_id
+            chat.max_tokens = body.max_tokens
+            chat.reasoning_budget = body.reasoning_budget
+            chat.knowledge_base_id = body.knowledge_base_id or None
             chat.updated_at = datetime.utcnow()
             db.commit()
 
@@ -264,6 +288,8 @@ def _chat_dict(c: Chat) -> dict:
         "title": c.title,
         "model": c.model,
         "knowledge_base_id": c.knowledge_base_id,
+        "max_tokens": c.max_tokens or 4096,
+        "reasoning_budget": c.reasoning_budget or 0,
         "created_at": str(c.created_at),
         "updated_at": str(c.updated_at),
     }

backend/routes/knowledge_routes.py

 """
-Knowledge base management and document upload.
+Knowledge base management and document upload (supports multiple files at once).
 """
 
 from pydantic import BaseModel
@@ -73,50 +73,80 @@ def delete_kb(kb_id: str, user: User = Depends(get_current_user), db: Session =
 
 
 @router.post("/{kb_id}/upload")
-async def upload_document(
+async def upload_documents(
     kb_id: str,
-    file: UploadFile = File(...),
+    files: list[UploadFile] = File(...),
     user: User = Depends(get_current_user),
     db: Session = Depends(get_db),
 ):
+    """Upload one or more documents to a knowledge base."""
     kb = _get_kb(kb_id, user, db)
-    content_bytes = await file.read()
 
-    if len(content_bytes) > MAX_UPLOAD_BYTES:
-        raise HTTPException(413, f"File too large. Max {MAX_UPLOAD_BYTES // 1024 // 1024}MB")
+    results = []
+    total_new_docs = 0
+    total_new_chunks = 0
+    total_new_chars = 0
 
-    filename = file.filename or "document.txt"
-    text = _extract_text(filename, content_bytes)
-
-    if not text.strip():
-        raise HTTPException(400, "Could not extract text from file")
-
-    chunks = _chunk_text(text, chunk_size=3000, overlap=300)
-
-    rag_service.add_documents(
-        collection_id=kb_id,
-        documents=chunks,
-        metadatas=[{"filename": filename, "chunk_index": i} for i in range(len(chunks))],
-    )
-
-    doc = KnowledgeDocument(
-        knowledge_base_id=kb_id,
-        filename=filename,
-        file_size=len(content_bytes),
-        chunk_count=len(chunks),
-    )
-    db.add(doc)
+    for file in files:
+        filename = file.filename or "document.txt"
+        try:
+            content_bytes = await file.read()
+
+            if len(content_bytes) > MAX_UPLOAD_BYTES:
+                results.append({
+                    "filename": filename,
+                    "error": f"File too large. Max {MAX_UPLOAD_BYTES // 1024 // 1024}MB",
+                })
+                continue
+
+            text = _extract_text(filename, content_bytes)
+
+            if not text.strip():
+                results.append({"filename": filename, "error": "Could not extract text from file"})
+                continue
+
+            chunks = _chunk_text(text, chunk_size=3000, overlap=300)
+
+            rag_service.add_documents(
+                collection_id=kb_id,
+                documents=chunks,
+                metadatas=[{"filename": filename, "chunk_index": i} for i in range(len(chunks))],
+            )
+
+            doc = KnowledgeDocument(
+                knowledge_base_id=kb_id,
+                filename=filename,
+                file_size=len(content_bytes),
+                chunk_count=len(chunks),
+            )
+            db.add(doc)
+
+            total_new_docs += 1
+            total_new_chunks += len(chunks)
+            total_new_chars += len(text)
+
+            results.append({
+                "filename": filename,
+                "chunks_added": len(chunks),
+                "characters": len(text),
+                "estimated_tokens": len(text) // 4,
+            })
+        except HTTPException as e:
+            results.append({"filename": filename, "error": str(e.detail)})
+        except Exception as e:
+            results.append({"filename": filename, "error": str(e)})
 
-    kb.document_count = (kb.document_count or 0) + 1
-    kb.chunk_count = (kb.chunk_count or 0) + len(chunks)
-    kb.total_characters = (kb.total_characters or 0) + len(text)
+    # Update KB aggregate stats
+    kb.document_count = (kb.document_count or 0) + total_new_docs
+    kb.chunk_count = (kb.chunk_count or 0) + total_new_chunks
+    kb.total_characters = (kb.total_characters or 0) + total_new_chars
     db.commit()
 
     return {
-        "filename": filename,
-        "chunks_added": len(chunks),
-        "characters": len(text),
-        "estimated_tokens": len(text) // 4,
+        "files": results,
+        "total_files": len(results),
+        "total_chunks_added": total_new_chunks,
+        "total_characters": total_new_chars,
     }
 
 

backend/services/memory_service.py

@@ -3,31 +3,40 @@ Build the `messages` list for the Bedrock/Anthropic API from chat history.
 
 Keeps the most recent messages that fit within a character budget
 (rough proxy for tokens — 1 token ≈ 4 chars).
+
+Limits the total number of DB rows loaded to prevent hangs on very long chats.
 """
 
 from sqlalchemy.orm import Session
 from backend.models import Chat, Message
 
-# ~180 000 tokens budget → ~720 000 characters
-MAX_CONTEXT_CHARS = 720_000
+# ~100 000 tokens budget → ~400 000 characters
+MAX_CONTEXT_CHARS = 400_000
+
+# Hard cap: never load more than this many messages from the DB
+MAX_MESSAGES = 80
 
 
 def build_messages(chat: Chat, db: Session) -> list[dict]:
     """
     Return a list of {"role": ..., "content": ...} ready for the API.
     Messages are oldest-first (chronological).
+    Only the most recent MAX_MESSAGES are considered, then trimmed by char budget.
     """
+    # Fetch the most recent N messages (descending) then reverse to chronological
     rows: list[Message] = (
         db.query(Message)
         .filter(Message.chat_id == chat.id)
-        .order_by(Message.created_at.asc())
+        .order_by(Message.created_at.desc())
+        .limit(MAX_MESSAGES)
         .all()
     )
+    rows.reverse()
 
     if not rows:
         return []
 
-    # --- trim from the oldest to fit budget ---
+    # --- trim from the oldest to fit character budget ---
     total_chars = sum(len(m.content or "") for m in rows)
     idx = 0
     while total_chars > MAX_CONTEXT_CHARS and idx < len(rows) - 2:

frontend/src/api.js

@@ -37,8 +37,11 @@ export const listChats = (token) =>
 export const createChat = (token, data = {}) =>
   request("POST", "/chats", token, data);
 
+export const updateChat = (token, chatId, data) =>
+  request("PUT", `/chats/${chatId}`, token, data);
+
 export const renameChat = (token, chatId, title) =>
-  request("PUT", `/chats/${chatId}`, token, { title });
+  updateChat(token, chatId, { title });
 
 export const deleteChat = (token, chatId) =>
   request("DELETE", `/chats/${chatId}`, token);
@@ -106,9 +109,11 @@ export const getKnowledgeBase = (token, kbId) =>
 export const deleteKnowledgeBase = (token, kbId) =>
   request("DELETE", `/knowledge/${kbId}`, token);
 
-export async function uploadDocument(token, kbId, file) {
+export async function uploadDocuments(token, kbId, files) {
   const form = new FormData();
-  form.append("file", file);
+  for (const file of files) {
+    form.append("files", file);
+  }
   const res = await fetch(`${BASE}/knowledge/${kbId}/upload`, {
     method: "POST",
     headers: { Authorization: `Bearer ${token}` },
@@ -121,6 +126,10 @@ export async function uploadDocument(token, kbId, file) {
   return res.json();
 }
 
+// Backward-compat wrapper for single file
+export const uploadDocument = (token, kbId, file) =>
+  uploadDocuments(token, kbId, [file]);
+
 /* ── Admin ─────────────────────────────────── */
 export const adminStats = (token) =>
   request("GET", "/admin/stats", token);

frontend/src/components/ChatView.jsx

 import React, { useState, useEffect, useRef, useCallback } from "react";
 import { useApp } from "../store";
 import {
-  getMessages, streamMessage, downloadZip, listKnowledgeBases,
+  getMessages, streamMessage, downloadZip, listKnowledgeBases, updateChat,
 } from "../api";
 import MessageBubble from "./MessageBubble";
 import {
@@ -15,27 +15,47 @@ const MODELS = [
 
 export default function ChatView({ chatId }) {
   const { state, dispatch } = useApp();
+
+  // ── Load persisted settings from the chat object ──
+  const currentChat = state.chats.find((c) => c.id === chatId);
   const [messages, setMessages] = useState([]);
   const [input, setInput] = useState("");
   const [streaming, setStreaming] = useState(false);
   const [showSettings, setShowSettings] = useState(false);
-  const [model, setModel] = useState(MODELS[0].id);
-  const [maxTokens, setMaxTokens] = useState(4096);
-  const [reasoningBudget, setReasoningBudget] = useState(0);
-  const [selectedKbId, setSelectedKbId] = useState(null);
+  const [model, setModel] = useState(currentChat?.model || MODELS[0].id);
+  const [maxTokens, setMaxTokens] = useState(currentChat?.max_tokens || 4096);
+  const [reasoningBudget, setReasoningBudget] = useState(currentChat?.reasoning_budget ?? 0);
+  const [selectedKbId, setSelectedKbId] = useState(currentChat?.knowledge_base_id || null);
   const [kbs, setKbs] = useState([]);
   const [streamText, setStreamText] = useState("");
   const [streamThinking, setStreamThinking] = useState("");
   const [isThinking, setIsThinking] = useState(false);
 
-  const bottomRef = useRef(null);
+  const scrollContainerRef = useRef(null);
   const inputRef = useRef(null);
   const abortRef = useRef(null);
+  const shouldAutoScrollRef = useRef(true);
+  const rafRef = useRef(null);
+
+  // ── Scroll helpers ──
+  function handleContainerScroll() {
+    const el = scrollContainerRef.current;
+    if (!el) return;
+    const { scrollHeight, scrollTop, clientHeight } = el;
+    shouldAutoScrollRef.current = scrollHeight - scrollTop - clientHeight < 200;
+  }
 
-  const scroll = useCallback(() => {
-    bottomRef.current?.scrollIntoView({ behavior: "smooth" });
+  const scrollToBottom = useCallback(() => {
+    if (!shouldAutoScrollRef.current) return;
+    if (rafRef.current) return; // already scheduled
+    rafRef.current = requestAnimationFrame(() => {
+      const el = scrollContainerRef.current;
+      if (el) el.scrollTop = el.scrollHeight;
+      rafRef.current = null;
+    });
   }, []);
 
+  // ── Load messages & KBs on mount ──
   useEffect(() => {
     (async () => {
       try {
@@ -47,12 +67,39 @@ export default function ChatView({ chatId }) {
     })();
   }, [chatId, state.token]);
 
-  useEffect(scroll, [messages, streamText, streamThinking, scroll]);
+  // Scroll when messages change or stream updates
+  useEffect(scrollToBottom, [messages, streamText, streamThinking, scrollToBottom]);
 
   useEffect(() => {
     inputRef.current?.focus();
   }, [chatId]);
 
+  // ── Save settings to backend ──
+  async function saveSettings() {
+    const data = {
+      model,
+      max_tokens: maxTokens,
+      reasoning_budget: reasoningBudget,
+      knowledge_base_id: selectedKbId || "",
+    };
+    try {
+      await updateChat(state.token, chatId, data);
+      dispatch({
+        type: "UPDATE_CHAT",
+        chat: { id: chatId, model, max_tokens: maxTokens, reasoning_budget: reasoningBudget, knowledge_base_id: selectedKbId },
+      });
+    } catch { /* ignore */ }
+  }
+
+  function toggleSettings() {
+    const closing = showSettings;
+    setShowSettings(!showSettings);
+    if (closing) {
+      saveSettings();
+    }
+  }
+
+  // ── Send message ──
   async function handleSend() {
     const content = input.trim();
     if (!content || streaming) return;
@@ -64,6 +111,7 @@ export default function ChatView({ chatId }) {
     setStreamText("");
     setStreamThinking("");
     setIsThinking(false);
+    shouldAutoScrollRef.current = true; // Force scroll for user's own message
 
     const ac = new AbortController();
     abortRef.current = ac;
@@ -126,6 +174,12 @@ export default function ChatView({ chatId }) {
         created_at: new Date().toISOString(),
       };
       setMessages((p) => [...p, assistantMsg]);
+
+      // Sync settings to store (backend already saved them from the message)
+      dispatch({
+        type: "UPDATE_CHAT",
+        chat: { id: chatId, model, max_tokens: maxTokens, reasoning_budget: reasoningBudget, knowledge_base_id: selectedKbId },
+      });
     } catch (err) {
       setMessages((p) => [
         ...p,
@@ -163,7 +217,11 @@ export default function ChatView({ chatId }) {
   return (
     <div className="flex-1 flex flex-col min-h-0">
       {/* Messages */}
-      <div className="flex-1 overflow-y-auto px-4 py-4 space-y-4">
+      <div
+        ref={scrollContainerRef}
+        onScroll={handleContainerScroll}
+        className="flex-1 overflow-y-auto px-4 py-4 space-y-4"
+      >
         {messages.map((m) => (
           <MessageBubble key={m.id} message={m} />
         ))}
@@ -192,8 +250,6 @@ export default function ChatView({ chatId }) {
             <span className="text-anton-muted text-sm">Son of Anton is thinking…</span>
           </div>
         )}
-
-        <div ref={bottomRef} />
       </div>
 
       {/* Input area */}
@@ -205,7 +261,7 @@ export default function ChatView({ chatId }) {
               <h3 className="text-sm font-semibold text-white flex items-center gap-1.5">
                 <Settings2 size={14} className="text-anton-accent" /> Generation Settings
               </h3>
-              <button onClick={() => setShowSettings(false)} className="text-anton-muted hover:text-white"><X size={14} /></button>
+              <button onClick={toggleSettings} className="text-anton-muted hover:text-white"><X size={14} /></button>
             </div>
 
             {/* Model */}
@@ -270,7 +326,7 @@ export default function ChatView({ chatId }) {
         )}
 
         <div className="flex items-end gap-2">
-          <button onClick={() => setShowSettings(!showSettings)}
+          <button onClick={toggleSettings}
             className={`p-2.5 rounded-xl transition shrink-0 ${
               showSettings ? "bg-anton-accent/20 text-anton-accent" : "text-anton-muted hover:text-white hover:bg-anton-card"
             }`}

frontend/src/components/MessageBubble.jsx

@@ -4,7 +4,7 @@ import remarkGfm from "remark-gfm";
 import CodeBlock from "./CodeBlock";
 import { User, Flame, ChevronDown, ChevronRight, Brain, Copy, Check } from "lucide-react";
 
-export default function MessageBubble({ message, isStreaming, isThinking }) {
+const MessageBubble = React.memo(function MessageBubble({ message, isStreaming, isThinking }) {
   const { role, content, thinking_content, input_tokens, output_tokens } = message;
   const isUser = role === "user";
   const [showThinking, setShowThinking] = useState(false);
@@ -133,4 +133,6 @@ export default function MessageBubble({ message, isStreaming, isThinking }) {
       )}
     </div>
   );
-}
+});
\ No newline at end of file
+
+export default MessageBubble;
\ No newline at end of file

frontend/src/components/Sidebar.jsx

@@ -3,7 +3,7 @@ import { useNavigate } from "react-router-dom";
 import { useApp } from "../store";
 import {
   createChat, deleteChat, renameChat,
-  listKnowledgeBases, createKnowledgeBase, deleteKnowledgeBase, uploadDocument,
+  listKnowledgeBases, createKnowledgeBase, deleteKnowledgeBase, uploadDocuments,
 } from "../api";
 import {
   Plus, Trash2, Flame, LogOut, Shield, PanelLeftClose, PanelLeftOpen,
@@ -20,6 +20,7 @@ export default function Sidebar({ onRefresh }) {
   const [showNewKb, setShowNewKb] = useState(false);
   const [expandedKb, setExpandedKb] = useState(null);
   const [uploading, setUploading] = useState(false);
+  const [uploadCount, setUploadCount] = useState(0);
   const [renamingId, setRenamingId] = useState(null);
   const [renameVal, setRenameVal] = useState("");
 
@@ -75,15 +76,25 @@ export default function Sidebar({ onRefresh }) {
     } catch { /* */ }
   }
 
-  async function handleUpload(kbId, file) {
+  async function handleUpload(kbId, files) {
     setUploading(true);
+    setUploadCount(files.length);
     try {
-      await uploadDocument(state.token, kbId, file);
+      const result = await uploadDocuments(state.token, kbId, files);
+      // Check for per-file errors
+      const errors = (result.files || []).filter((f) => f.error);
+      if (errors.length > 0) {
+        alert(
+          `Uploaded ${result.files.length - errors.length} of ${result.files.length} files.\n\nErrors:\n` +
+          errors.map((e) => `• ${e.filename}: ${e.error}`).join("\n")
+        );
+      }
       loadKbs();
     } catch (e) {
       alert(e.message);
     } finally {
       setUploading(false);
+      setUploadCount(0);
     }
   }
 
@@ -221,9 +232,17 @@ export default function Sidebar({ onRefresh }) {
                     </div>
                     <label className={`flex items-center gap-1.5 px-2 py-1.5 rounded border border-dashed border-anton-border text-xs text-anton-muted hover:text-anton-accent hover:border-anton-accent transition cursor-pointer ${uploading ? "opacity-50 pointer-events-none" : ""}`}>
                       <Upload size={12} />
-                      {uploading ? "Uploading…" : "Upload file (.txt, .pdf, .md, .json, .csv …)"}
-                      <input type="file" className="hidden" accept=".txt,.md,.pdf,.json,.csv,.py,.js,.ts,.cs,.html,.css,.xml,.yaml,.yml,.toml"
-                        onChange={(e) => e.target.files[0] && handleUpload(kb.id, e.target.files[0])}
+                      {uploading
+                        ? `Uploading ${uploadCount} file${uploadCount !== 1 ? "s" : ""}…`
+                        : "Upload files (.txt, .pdf, .md, .json, .csv …)"}
+                      <input type="file" className="hidden" multiple
+                        accept=".txt,.md,.pdf,.json,.csv,.py,.js,.ts,.cs,.html,.css,.xml,.yaml,.yml,.toml"
+                        onChange={(e) => {
+                          if (e.target.files && e.target.files.length > 0) {
+                            handleUpload(kb.id, Array.from(e.target.files));
+                          }
+                          e.target.value = "";
+                        }}
                       />
                     </label>
                     <button onClick={() => handleDeleteKb(kb.id)}