coder-agent 2.9.13 → 2.9.14

package/README.md

@@ -1,10 +1,13 @@
 # 🤖 Coder
 
-A powerful, zero-dependency, cross-platform CLI coding agent powered by **Google Gemini** (gemini-2.5-flash / gemini-2.5-pro) with native tool execution and a clean, minimal Apple-inspired design.
+A powerful, zero-dependency, cross-platform CLI coding agent powered by **Google Gemini, Groq, Mistral, OpenRouter, NVIDIA NIM, and GitHub Models** with native tool execution and a clean, minimal Apple-inspired design.
 
 ## Features
 
 - **Apple-Inspired Design**: A clean, high-contrast, minimalist terminal UI styled in elegant monochrome greys and whites.
+- **Multi-Provider AI Engine**: Query models from Gemini API, Groq, Mistral, OpenRouter, NVIDIA NIM, or GitHub Models.
+- **Auto-Rotation & Fallbacks**: Automatically rotates to backup models within a provider family if a server error or rate-limit is hit.
+- **Smart Rate-Limit Throttling**: Implements smart background cooldown timers between requests to prevent hitting provider Requests Per Minute (RPM) limits without disrupting the user experience.
 - **Global CLI Execution**: Install once and run anywhere via `coder-agent` (or the `coder` / `gemini-agent` / `groq-agent` compatibility aliases).
 - **System Automation Tools**:
   - `read_file` / `write_file` / `patch_file` — Segmented reading and patch-based editing.
@@ -36,23 +39,30 @@ npx coder-agent
 
 ## Setup & Configuration
 
-The agent requires a **Gemini API Key**.
+The agent requires at least one API Key to function. By default, it uses the **Gemini API Key**.
 
 ### 1. Interactive Bootstrapping
-Simply run `coder-agent`. If no key is configured, the CLI will interactively prompt you for the key and offer to save it globally.
+Simply run `coder-agent`. If no key is configured, the CLI will interactively prompt you for a Gemini key and offer to save it globally.
 
-### 2. Command Line Flags
-You can save your key globally at any time using flags:
+### 2. Environment Variables
+You can export provider API keys as environment variables:
 
 ```bash
-coder-agent --set-key YOUR_GEMINI_API_KEY
+export GEMINI_API_KEY=AIzaSy...
+export GROQ_API_KEY=gsk_...
+export MISTRAL_API_KEY=...
+export OPENROUTER_API_KEY=sk-or-...
+export NVIDIA_API_KEY=nvapi-...
+export GITHUB_TOKEN=github_pat_...
 ```
 
-### 3. Environment Variables
-Alternatively, you can export it as an environment variable:
+### 3. REPL Commands (Config File)
+Inside the interactive REPL session, use the `/key` command to save API keys globally:
 
 ```bash
-export GEMINI_API_KEY=AIzaSy...
+/key                          # View active API key status for all providers
+/key <api_key>                # Save your Gemini API Key globally
+/key <provider> <api_key>     # Save key for a specific provider (e.g. /key groq gsk_...)
 ```
 
 ---
@@ -74,7 +84,7 @@ coder-agent "write a binary search script in Python"
 ### CLI Command Options
 - `-h, --help` — Show the help screen.
 - `-v, --version` — Show version information.
-- `--model <name>` — Set the default Gemini model globally.
+- `--model <name>` — Set the default model globally.
 - `--set-key <key>` — Persist your Gemini API Key globally.
 - `--memory <scope>` — Define memory scope (`project`, `user`, or `local`).
 
@@ -86,7 +96,8 @@ Type these commands directly inside the interactive session:
 
 | Command | Description |
 |---------|-------------|
-| `/model [name]` | View active model or switch active model dynamically (e.g. `/model gemini-2.5-pro`) |
+| `/model [name]` | View active model or switch active model dynamically (e.g. `/model groq/llama-3.3-70b-specdec`) |
+| `/key [provider] [key]` | View stored API keys or save keys globally |
 | `/clear` | Wipe conversation memory |
 | `/status` | Show active model and memory usage details |
 | `/help` | Show command summary |
@@ -95,10 +106,27 @@ Type these commands directly inside the interactive session:
 ---
 
 ## Popular Models Supported
-- `gemini-2.5-flash` (Default, highly capable, extremely fast, massive 1M+ token context)
-- `gemini-2.5-pro` (Reasoning model, superb for complex coding tasks)
-- `gemini-2.0-flash` (Ultra-fast, lightweight)
-- `gemini-2.0-pro-exp` (Experimental reasoning model)
+
+To use a model from a different provider, prefix the model name with `provider/`:
+
+*   **Google Gemini (Default)**:
+    *   `gemini-2.5-flash` — Default, highly capable & fast (1M+ context).
+    *   `gemini-2.5-pro` — High reasoning model, superb for complex coding.
+    *   `gemini-3.5-flash` — Newest generation, ultra-fast.
+*   **Groq (`groq/`)**:
+    *   `groq/llama-3.3-70b-specdec` — Blazing fast reasoning.
+    *   `groq/gemma-2-9b-it` — Lightweight, fast open-weights.
+*   **Mistral (`mistral/`)**:
+    *   `mistral/codestral` — SOTA open coding model.
+*   **OpenRouter (`openrouter/`)**:
+    *   `openrouter/qwen/qwen-2.5-coder-32b-instruct:free` — Popular free-tier coding model.
+    *   `openrouter/deepseek/deepseek-r1:free` — Powerful reasoning/thinking model.
+*   **NVIDIA NIM (`nvidia/`)**:
+    *   `nvidia/qwen-2.5-coder-32b-instruct` — NVIDIA-optimized coding model.
+    *   `nvidia/deepseek-r1` — High-speed reasoning model.
+*   **GitHub Models (`github/`)**:
+    *   `github/claude-3.5-sonnet` — Claude 3.5 Sonnet (requires GitHub token).
+    *   `github/gpt-4o` — OpenAI GPT-4o.
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "coder-agent",
-  "version": "2.9.13",
+  "version": "2.9.14",
   "description": "CLI coding agent powered by Google Gemini",
   "type": "module",
   "main": "dist/index.js",