Feature chat

.agent/tasks/task.md

@@ -0,0 +1,10 @@
+# Task: Re-apply Model Fixes
+
+## Observations
+1. **Gemini Error**: `models/gemini-1.5-pro` 404s. Using `gemini-1.5-pro-latest` previously resolved this (passing the error to the next provider).
+2. **OpenAI Error**: `gpt-4-turbo` error ("model does not exist"). Only appeared after Gemini was fixed.
+
+## Plan
+- [ ] Update `app/api/chat/provider.ts` again to use correct model identifiers. <!-- id: 0 -->
+    - Gemini: `gemini-1.5-flash-latest`, `gemini-1.5-pro-latest`
+    - OpenAI: `gpt-4o` (widely available and valid)

.env.example

@@ -25,3 +25,19 @@ UPSTASH_REDIS_REST_TOKEN=xxxxxxxxx
 # # Upstash
 # UPSTASH_REDIS_REST_URL=xxxxxxxxx
 # UPSTASH_REDIS_REST_TOKEN=xxxxxxxxx
+
+# AI Gateway (routing / unified LLM access)
+AI_GATEWAY_API_KEY=xxxxxxxxx
+
+# Google Gemini / Generative AI
+GEMINI_API_KEY=xxxxxxxxx
+GOOGLE_GENERATIVE_AI_API_KEY=xxxxxxxxx
+
+# OpenAI (GPT models)
+OPENAI_API_KEY=xxxxxxxxx
+
+# Anthropic (Claude models)
+ANTHROPIC_API_KEY=xxxxxxxxx
+
+# Tavily (Search / Web RAG)
+TAVILY_API_KEY=xxxxxxxxx

README.md

@@ -13,6 +13,7 @@ A modern SaaS template built with Next.js 15, Tailwind CSS v4, Shadcn UI v2, Ups
 - 🧩 **React Hook Form** - Flexible form validation
 - ⚙️ **Zod** - Schema validation
 - 🛡️ **Enhanced Security** - Robust authentication with rate limiting using Upstash
+- 📚 **RAG (Chat with Data)** - Upload and chat with PDF, DOCX, TXT, and Images using Google Gemini embeddings
 - 🔒 **Security Headers** - CSP and other security headers (Coming Soon)
 - 🚫 **Anti-Brute Force** - Protection against authentication attacks (Coming Soon)
 
@@ -62,6 +63,7 @@ supabase start
 ```
 
 > After the contianer starts, you will be provided with some credentials like the following example:
+>
 > ```
 >         API URL: http://127.0.0.1:54321
 >     GraphQL URL: http://127.0.0.1:54321/graphql/v1
@@ -76,8 +78,18 @@ supabase start
 >   S3 Secret Key: 850181e4652dd023b7a98c58ae0d2d34bd487ee0cc3254aed6eda37307425907
 >       S3 Region: local
 > ```
+>
 > **Copy those credentials from you terminal** and save them in your notes or create a `supabase-local-credentials.txt` file in this repo (it is already added to `.gitignore` so that it is not pushed into the repository.)
 
+#### Run RAG Migrations
+
+To enable vector search, you need to apply the RAG database schema:
+
+1. Go to your Supabase Dashboard (http://127.0.0.1:54323).
+2. Open the **SQL Editor**.
+3. Open the file `supabase/migrations/20251217_add_rag_columns.sql` from your project.
+4. Copy the content and run it in the SQL Editor.
+
 ### 4. Set up environment variables
 
 1. Copy the `.env.example` file to `.env.local`:
@@ -99,6 +111,11 @@ NEXT_PUBLIC_SUPABASE_ANON_KEY=xxxxxxxxx
 # Upstash
 UPSTASH_REDIS_REST_URL=xxxxxxxxx
 UPSTASH_REDIS_REST_TOKEN=xxxxxxxxx
+
+# RAG & AI (Google Gemini is required for RAG)
+GOOGLE_GENERATIVE_AI_API_KEY=xxxxxxxxx
+TAVILY_API_KEY=xxxxxxxxx # Optional, for web search
+# OPENAI_API_KEY=xxxxxxxxx # Optional, but Google is preferred for this RAG implementation
 ```
 
 ### 5. Run the development server
@@ -114,7 +131,48 @@ yarn dev
 
 Your application should now be running at [http://localhost:3000](http://localhost:3000).
 
+### 6. How to use RAG (Chat with Data)
+
+1. **Upload Documents**: Click the paperclip icon in the chat input area.
+2. **Select Files**: Choose PDF, Word (.docx), Text (.txt), or Image files.
+3. **Chat**: Once uploaded, ask questions about your documents. The AI will retrieve relevant context to answer.
+
+## RAG Architecture
+
+### Document Ingestion Flow
+
+```mermaid
+graph TD
+    A["User Uploads File"] --> B["API: /api/process-document"]
+    B --> C{"File Type?"}
+    C -- "PDF/DOCX/TXT" --> D["Extract Text"]
+    C -- "Image" --> E["Gemini Vision Analysis"]
+    D --> F["Chunk Text"]
+    E --> F
+    F --> G["Generate Embeddings (Google/OpenAI)"]
+    G --> H["Store in Supabase (pgvector)"]
+```
+
+### Chat Retrieval Flow
+
+```mermaid
+sequenceDiagram
+    participant User
+    participant API as Chat API
+    participant DB as Supabase (Vector DB)
+    participant LLM as Google Gemini
+
+    User->>API: Send Message
+    API->>API: Generate Embedding for Query
+    API->>DB: match_documents(embedding)
+    DB-->>API: Return Relevant Chunks
+    API->>API: Inject Context into System Prompt
+    API->>LLM: Generate Response (Query + Context)
+    LLM-->>User: Stream Response
+```
+
 ## Some **Features**
+
 - Email/password authentication
 - Google OAuth integration
 - Strong password requirements
@@ -134,13 +192,18 @@ Your application should now be running at [http://localhost:3000](http://localho
 │   │   └── reset-password/
 │   ├── (public)/           # Public routes
 │   ├── (authenticated)/    # Protected routes
+│   │   └── chat/           # Chat interface
+│   ├── api/                # API Routes
+│   │   ├── chat/           # Chat API
+│   │   └── process-document/# RAG Ingestion API
 │   ├── actions/            # Server actions
 │   └── globals.css         # Global styles
 ├── assets/                 # Project assets
 │   ├── images/             # Image assets
 │   └── logos/              # Logo files
 ├── components/             # React components
 │   ├── ui/                 # Shadcn UI components
+│   ├── chat/               # Chat components
 │   ├── mode-toggle.tsx     # Dark/light mode toggle
 │   └── theme-provider.tsx  # Theme context provider
 ├── hooks/                  # Custom React hooks

app/(authenticated)/chat/[[...sessionId]]/layout.tsx

@@ -0,0 +1,17 @@
+"use client";
+
+import ChatSidebar from "@/components/chat/ChatSidebar";
+
+export default function ChatLayout({ children }: { children: React.ReactNode }) {
+  return (
+    <div className="flex h-screen">
+      <div>
+        <ChatSidebar />
+      </div>
+
+      <main className="flex flex-col flex-1">
+        {children}
+      </main>
+    </div>
+  );
+}