orbitchat 3.1.0 → 3.1.2

package/README.md

@@ -22,19 +22,26 @@ Installed CLI commands:
 
 ## Quick Start
 
-1. Define your adapters (agents) via the `ORBIT_ADAPTERS` or `VITE_ADAPTERS` environment variable:
+1. Define your adapter secrets via the `ORBIT_ADAPTER_KEYS` or `VITE_ADAPTER_KEYS` environment variable:
    ```bash
-   export ORBIT_ADAPTERS='[
-     {"name":"Simple Chat","apiKey":"my-key","apiUrl":"http://localhost:3000","description":"Default conversational agent."}
-   ]'
+   # Mapping of Agent Name -> API Key
+   export ORBIT_ADAPTER_KEYS='{"Simple Chat":"my-secret-key"}'
    ```
 
-2. Run the CLI:
+2. (Optional) Configure adapter URLs and metadata in `orbitchat.yaml`:
+   ```yaml
+   adapters:
+     - name: "Simple Chat"
+       apiUrl: "http://localhost:3000"
+       description: "Default conversational agent."
+   ```
+
+3. Run the CLI:
    ```bash
    orbitchat --config ./orbitchat.yaml --port 5173
    ```
 
-3. Open `http://localhost:5173` — select an agent and start chatting.
+4. Open `http://localhost:5173` — select an agent and start chatting.
 
 ## Architecture
 
@@ -46,7 +53,7 @@ Browser  ──X-Adapter-Name──▶  Express proxy  ──X-API-Key──▶
 The frontend never handles API keys. Instead:
 - The browser sends an `X-Adapter-Name` header with every API request.
 - The Express proxy looks up the adapter, injects the real `X-API-Key`, and forwards the request to the configured backend URL.
-- `GET /api/adapters` returns only adapter names and descriptions — never keys or URLs.
+- `GET /api/adapters` returns non-secret adapter metadata (name, description, notes, model) — never keys or backend URLs.
 
 ## CLI Options
 
@@ -70,8 +77,8 @@ Options:
 # Start with a custom config file
 orbitchat --config /path/to/orbitchat.yaml
 
-# Start with adapters defined inline
-ORBIT_ADAPTERS='[{"name":"Chat","apiKey":"mykey","apiUrl":"https://api.example.com"}]' orbitchat
+# Start with adapter keys defined inline
+ORBIT_ADAPTER_KEYS='{"Chat":"mykey"}' orbitchat
 
 # API proxy only — no UI, no build required
 orbitchat --api-only --port 5174
@@ -101,7 +108,7 @@ Your frontend needs to do two things:
    ```json
    {
      "adapters": [
-       { "name": "Simple Chat", "description": "...", "notes": "..." }
+       { "name": "Simple Chat", "description": "...", "notes": "...", "model": "gpt-4o-mini" }
      ]
    }
    ```
@@ -116,7 +123,7 @@ Your frontend needs to do two things:
 
 | Method | Path | Headers | Description |
 |--------|------|---------|-------------|
-| GET | `/api/adapters` | — | List available adapter names and descriptions |
+| GET | `/api/adapters` | — | List available adapter metadata (name, description, notes, model) |
 | POST | `/api/v1/chat` | `X-Adapter-Name`, `X-Session-ID` | Send a chat message (SSE streaming response) |
 | POST | `/api/files/upload` | `X-Adapter-Name` | Upload a file (multipart/form-data) |
 | GET | `/api/files` | `X-Adapter-Name` | List uploaded files |
@@ -124,103 +131,80 @@ Your frontend needs to do two things:
 | DELETE | `/api/files/:id` | `X-Adapter-Name` | Delete a file |
 | GET | `/api/v1/autocomplete?q=...&limit=5` | `X-Adapter-Name` | Autocomplete suggestions |
 
-### Example: calling from a custom React app
-
-```js
-// Discover adapters
-const res = await fetch('http://localhost:5174/api/adapters');
-const { adapters } = await res.json();
-
-// Send a chat message (SSE stream)
-const response = await fetch('http://localhost:5174/api/v1/chat', {
-  method: 'POST',
-  headers: {
-    'Content-Type': 'application/json',
-    'X-Adapter-Name': adapters[0].name,
-    'X-Session-ID': crypto.randomUUID(),
-  },
-  body: JSON.stringify({ message: 'Hello!' }),
-});
-
-// Read the SSE stream
-const reader = response.body.getReader();
-const decoder = new TextDecoder();
-while (true) {
-  const { done, value } = await reader.read();
-  if (done) break;
-  console.log(decoder.decode(value));
-}
-```
-
 ## Configuring Adapters
 
-Adapters map a user-visible name to a backend API key and URL. Configure them via the `ORBIT_ADAPTERS` (or `VITE_ADAPTERS`) environment variable as a JSON array:
+Configuration is split between **metadata/URLs** (in `orbitchat.yaml`) and **secrets** (in environment variables).
 
-```bash
-export ORBIT_ADAPTERS='[
-  {
-    "name": "Simple Chat",
-    "apiKey": "default-key",
-    "apiUrl": "http://localhost:3000",
-    "description": "Basic chat interface using the default conversational agent."
-  },
-  {
-    "name": "Document QA",
-    "apiKey": "doc-qa-key",
-    "apiUrl": "http://localhost:3000",
-    "description": "Chat with uploaded documents.",
-    "notes": "Supports PDF, DOCX, and plain text uploads."
-  }
-]'
+### 1. Metadata in `orbitchat.yaml`
+
+Define your adapters list in the YAML file:
+
+```yaml
+adapters:
+  - name: "Simple Chat"
+    apiUrl: "http://localhost:3000"
+    description: "Basic chat interface using the default conversational agent."
+  - name: "Document QA"
+    apiUrl: "http://localhost:3000"
+    description: "Chat with uploaded documents."
+    notes: "Supports PDF, DOCX, and plain text uploads."
 ```
 
-Each adapter object supports:
+| Field | Description |
+|-------|-------------|
+| `name` | Display name shown in the agent selector (must match the key in `.env`) |
+| `apiUrl` | Backend URL (defaults to `api.url`, then `http://localhost:3000`) |
+| `description` | Short summary shown in dropdowns |
+| `notes` | Markdown content shown in the chat empty state |
 
-| Field | Required | Description |
-|-------|----------|-------------|
-| `name` | Yes | Display name shown in the agent selector |
-| `apiKey` | Yes | Backend API key (never exposed to the browser) |
-| `apiUrl` | No | Backend URL (defaults to `api.url` in `orbitchat.yaml`, then `http://localhost:3000`) |
-| `description` | No | Short summary shown in dropdowns |
-| `notes` | No | Markdown content shown in the chat empty state |
+### 2. Secrets in `.env`
 
-If `api.defaultAdapter` is not set (or left as `default-key`), the first adapter in the list is used.
+Provide the API keys via `ORBIT_ADAPTER_KEYS` (or `VITE_ADAPTER_KEYS`) as a JSON object:
 
-### Agent Selector UX
+```bash
+VITE_ADAPTER_KEYS='{
+  "Simple Chat": "secret-key-1",
+  "Document QA": "secret-key-2"
+}'
+```
 
-- When a conversation has no messages, the chat canvas shows a centered agent selector with the adapter's notes rendered beneath it.
-- Once an adapter is selected, the input field unlocks.
-- Sidebar cards display the agent assigned to each conversation.
-- To change the adapter after messages exist, use the "Change agent" action in the sidebar.
+The system deep-merges these sources at runtime.
 
 ## Configuration
 
 ### Runtime Config File
 
-Runtime settings are loaded from `orbitchat.yaml` (see `orbitchat.yaml.example`).
+Runtime settings are loaded from `orbitchat.yaml` (see `orbitchat.yaml.example`). The configuration uses a nested structure:
 
-Config lookup:
-1. `--config /path/to/orbitchat.yaml` if provided
-2. `./orbitchat.yaml` (current working directory)
+```yaml
+application:
+  name: "ORBIT Chat"
+api:
+  url: "http://localhost:3000"
+features:
+  enableUpload: true
+```
 
-### Environment Variables
+Header logos (`header.logoUrl`, `header.logoUrlLight`, `header.logoUrlDark`) support:
+- Remote URLs, for example `https://example.com/logo.png`
+- Local file paths (absolute or relative to `orbitchat.yaml`), for example `./public/logo.png`
 
-Adapter secrets are provided via environment variables:
+Theme-aware logo selection order:
+- Light theme: `header.logoUrlLight` -> `header.logoUrl` -> `header.logoUrlDark`
+- Dark theme: `header.logoUrlDark` -> `header.logoUrl` -> `header.logoUrlLight`
 
-```bash
-ORBIT_ADAPTERS='[{"name":"Simple Chat","apiKey":"default-key","apiUrl":"http://localhost:3000"}]'
-```
+### Environment Variables
 
-- `ORBIT_ADAPTERS` is preferred.
-- `VITE_ADAPTERS` is also supported for compatibility.
-- If both are set, `ORBIT_ADAPTERS` takes precedence.
+Adapter secrets are provided via:
+- `ORBIT_ADAPTER_KEYS` (Preferred)
+- `VITE_ADAPTER_KEYS`
 
 Auth secrets are read from:
 - `VITE_AUTH_DOMAIN`
 - `VITE_AUTH_CLIENT_ID`
 - `VITE_AUTH_AUDIENCE`
 
-The CLI also loads `.env` and `.env.local` from the current working directory on startup.
+The CLI loads `.env` and `.env.local` from the current working directory on startup.
 
 ## Development
 
@@ -233,18 +217,6 @@ npm install
 npm run dev
 ```
 
-### Development with Express Proxy
-
-To run both the Express proxy and Vite dev server together:
-
-```bash
-node bin/dev-server.js
-```
-
-This starts:
-- Express proxy on port 5174 (handles `/api/*` routes)
-- Vite dev server on port 5173 (proxies API requests to Express)
-
 ### Building for Production
 
 ```bash
@@ -257,133 +229,28 @@ The output is written to `dist/`. Serve it with:
 orbitchat --port 8080
 ```
 
-### Running as a Daemon
-
-For npm package installs, use:
+## Troubleshooting
 
-```bash
-orbitchat-daemon --start        # Start in background
-orbitchat-daemon --start 8080   # Start on custom port
-orbitchat-daemon --stop         # Stop
-orbitchat-daemon --status       # Check status
-```
+### No Adapters Available
 
-From a source checkout, you can also run:
+If the agent selector shows no adapters:
+1. Ensure `VITE_ADAPTER_KEYS` is set and contains valid JSON.
+2. Verify that the adapter `name` in `orbitchat.yaml` exactly matches the key used in `VITE_ADAPTER_KEYS`.
+3. Check the CLI startup logs for "Available Adapters: ...".
 
-```bash
-./orbitchat.sh --start
-```
+### Stale Configuration
 
-## Available Scripts
-
-- `npm run dev` — Start Vite dev server
-- `npm run build` — Build for production
-- `npm run preview` — Preview production build
-- `npm run dev:local` — Start dev server with local API build
-- `npm run dev:with-api` — Build API from `../node-api` and start dev server
-- `npm run build:local` — Build for production with local API
-- `npm run build:api` — Build and copy API from `../node-api`
-
-## Features
-
-- **Streaming Responses**: Real-time streaming of AI responses via SSE
-- **Agent Selection**: Choose from configured adapters per conversation
-- **File Upload**: Upload and attach files (PDF, DOCX, TXT, CSV, JSON, HTML, images, audio) to conversations
-- **File Context**: Query uploaded files — they are chunked, embedded, and included in chat context
-- **Autocomplete**: Optional type-ahead suggestions via `/api/v1/autocomplete`
-- **Conversation Threads**: Branch conversations into focused sub-threads
-- **Session Management**: Automatic session ID generation and persistence
-- **Conversation Persistence**: Chat history saved to localStorage
-- **Audio Output**: Optional text-to-speech for AI responses
-- **Feedback Buttons**: Optional thumbs-up/down per message
+If you've updated `orbitchat.yaml` but don't see changes:
+1. The CLI watches the YAML file and should restart automatically.
+2. Clear browser site data/localStorage for the app origin to ensure no stale session state is being used.
 
 ## Security
 
 - The browser **never** sees real API keys. The Express proxy maps adapter names to keys server-side.
-- `GET /api/adapters` only exposes names and descriptions — never keys or backend URLs.
-- Keep `ORBIT_ADAPTERS` / `VITE_ADAPTERS` out of source control.
-- Run the proxy behind HTTPS (or another reverse proxy) in production so users cannot intercept traffic.
-- Secure the host running the CLI — a compromised host can leak the adapters config or intercept proxied traffic.
-
-## File Upload
-
-### Supported File Types
-
-| Type | Formats | Processing |
-|------|---------|------------|
-| Documents | PDF, DOCX, PPTX, XLSX | Text extraction, chunking, vector indexing |
-| Text | TXT, MD, HTML | Direct chunking and indexing |
-| Data | CSV, JSON | Chunking and indexing |
-| Code | PY, JS, TS, Java, Go, Rust, C/C++, and more | Direct indexing |
-| Images | PNG, JPEG, TIFF | OCR via vision service |
-| Audio | WAV, MP3, MP4, OGG, FLAC, WebM, M4A, AAC | ASR (Automatic Speech Recognition) |
-| Subtitles | VTT | Direct indexing |
-
-### Limits
-
-- Maximum file size: 50 MB (configurable via `--max-file-size-mb`)
-- Maximum files per conversation: 5 (configurable via `--max-files-per-conversation`)
-
-### Processing Pipeline
-
-1. **Upload** — File uploaded via the Express proxy to `/api/files/upload`
-2. **Validation** — File type and size validated client-side and server-side
-3. **Storage** — File saved to filesystem (or S3 in production)
-4. **Extraction** — Text and metadata extracted using format-specific processors
-5. **Chunking** — Content chunked using configured strategy (fixed or semantic)
-6. **Indexing** — Chunks indexed in vector store for semantic search
-7. **Status Polling** — Client polls until processing completes
-
-## Integration Details
-
-The application uses:
-- **Zustand** for state management
-- **Express** + `http-proxy-middleware` for the API proxy layer
-- **@schmitech/markdown-renderer** ([GitHub](https://github.com/schmitech/markdown-renderer) | [NPM](https://www.npmjs.com/package/@schmitech/markdown-renderer)) for rich markdown rendering
-- **localStorage** for persistent session and conversation storage
-- **TypeScript** for type safety throughout
-
-## Troubleshooting
-
-### No Adapters Available
-
-If the agent selector shows no adapters:
-1. Ensure `ORBIT_ADAPTERS` or `VITE_ADAPTERS` is set and valid JSON
-2. Check the CLI startup logs for "Available Adapters: ..."
-3. Verify each adapter has a `name` and `apiKey` field
-
-If adapters load but descriptions/notes are missing in packaged installs (`npm pack` + install), while `npm run dev` works:
-1. Prefer `ORBIT_ADAPTERS` (it takes precedence over `VITE_ADAPTERS` when both are set)
-2. Ensure `orbitchat.yaml` contains adapter metadata and adapter `name` values exactly match `ORBIT_ADAPTERS`
-3. Rebuild and repack from the updated source: `npm run build && npm pack`
-4. Reinstall the newly generated tarball
-5. Restart with a clean process/port: `orbitchat-daemon --force-restart` (or `./orbitchat.sh --force-restart` in source checkout)
-6. Verify runtime output:
-   - Startup log shows `Available Adapters: ...`
-   - `GET /api/adapters` returns `description`/`notes` for each adapter
-
-### File Upload Issues
-
-- **File size exceeded** — Check file size against the configured limit
-- **Unsupported format** — Verify file type is in the supported list above
-- **Upload fails** — Check server logs and adapter configuration
-- **Processing fails** — Ensure the file processing service is initialized on the backend
-
-### Debug Mode
-
-Enable debug logging:
-```bash
-# in orbitchat.yaml
-debug:
-  consoleDebug: true
-```
-This enables detailed runtime logging from the CLI server.
+- `GET /api/adapters` only exposes non-secret metadata (name, description, notes, model) — never keys or backend URLs.
+- Keep `VITE_ADAPTER_KEYS` out of source control.
+- Run the proxy behind HTTPS in production.
 
-## Deployment Checklist
+## License
 
-1. **Build the app**: `npm run build`
-2. **Set `ORBIT_ADAPTERS`** with your production adapter configs (keep out of git)
-3. **Run behind HTTPS** — use a reverse proxy like nginx or Caddy in front of `orbitchat`
-4. **Bind to the right interface**: use `--host 0.0.0.0` to allow external access, or keep the default `localhost` for local-only
-5. **Tune limits** — set `--max-conversations`, `--max-message-length`, etc. appropriate for your deployment
-6. **Monitor logs** — use `orbitchat-daemon --start` for daemon mode with log file, or run directly and pipe to your log aggregator
+MIT