cade-cli 0.3.4__tar.gz → 0.4.0__tar.gz

{cade_cli-0.3.4 → cade_cli-0.4.0}/.gitignore

@@ -41,3 +41,6 @@ build/
 .venv
 activate.sh
 docs/
+
+# uv
+uv.lock

{cade_cli-0.3.4 → cade_cli-0.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cade-cli
-Version: 0.3.4
+Version: 0.4.0
 Summary: Cade - The CLI Agent from Arcade.dev
 Project-URL: Homepage, https://arcade.dev
 Project-URL: Documentation, https://docs.arcade.dev
@@ -24,8 +24,8 @@ Classifier: Typing :: Typed
 Requires-Python: >=3.11
 Requires-Dist: anthropic<1.0.0,>=0.34.0
 Requires-Dist: arcade-core<5.0.0,>=4.1.0
+Requires-Dist: arcade-mcp-server>=1.0.0
 Requires-Dist: arcade-tdk>=2.0.0
-Requires-Dist: arcadepy>=1.3.1
 Requires-Dist: authlib<2.0.0,>=1.6.0
 Requires-Dist: httpx<1.0.0,>=0.27.0
 Requires-Dist: openai<2.0.0,>=1.0.0
@@ -75,7 +75,7 @@ brew install cade
 
 ```bash
 curl -LsSf https://astral.sh/uv/install.sh | sh
-uv pip install cade-cli
+uv tool install cade-cli
 ```
 
 ### Install with pip
@@ -115,6 +115,27 @@ cade -r                        # Resume most recent
 cade resume "my-project"       # Resume by name
 ```
 
+### Authentication
+
+Cade uses Arcade Cloud for authentication and shares credentials with arcade-cli.
+
+```bash
+cade login                      # Log in to Arcade Cloud
+cade logout                     # Log out
+cade whoami                     # Show current login status
+```
+
+### Context Management
+
+Switch between organizations and projects for Arcade Cloud features.
+
+```bash
+cade context show               # Show current org/project
+cade context list               # List available orgs and projects
+cade context switch -i          # Interactive selection
+cade context switch --org my-org --project my-project
+```
+
 ### Single Message Mode
 
 ```bash
@@ -128,6 +149,7 @@ cat error.log | cade -m "What went wrong?"
 |--------|-------------|
 | `-r`, `--resume` | Resume the most recent thread |
 | `-m`, `--message` | Single message mode (non-interactive) |
+| `-L`, `--local-only` | Disable remote tools (use only local tools) |
 | `-v`, `--verbose` | Enable debug logging |
 | `--version` | Show version |
 
@@ -205,9 +227,11 @@ Config is stored in `~/.cadecoder/`:
 | Variable | Description |
 |----------|-------------|
 | `OPENAI_API_KEY` | OpenAI API key |
+| `OPENAI_BASE_URL` | Custom OpenAI-compatible API endpoint |
 | `ANTHROPIC_API_KEY` | Anthropic API key |
 | `ARCADE_API_KEY` | Arcade API key (alternative to OAuth) |
 | `ARCADE_BASE_URL` | Custom Arcade API endpoint |
+| `CADE_LOCAL_ONLY` | Set to `1` to disable remote tools |
 | `CADECODER_HOME` | Override config directory |
 
 ### Example Config
@@ -228,9 +252,73 @@ provider = "openai"
 model = "gpt-4.1"
 
 [tool_settings]
-disabled_tools = []
+# Tool filtering is managed via MCP server configuration
+# See: cade mcp add --help
 ```
 
+## Using Local or Custom LLMs
+
+Cade works with any OpenAI-compatible API, including local servers (Ollama, vLLM, llama.cpp) and alternative cloud providers (Together AI, Groq, Fireworks).
+
+### Local-Only Mode
+
+When using local LLMs, you can skip Arcade Cloud authentication entirely with `--local-only`:
+
+```bash
+# Local Ollama server without Arcade Cloud
+cade chat --local-only --endpoint "http://localhost:11434/v1" --model "llama3"
+
+# Or via environment variable
+CADE_LOCAL_ONLY=1 cade chat --endpoint "http://localhost:11434/v1" --model "llama3"
+```
+
+This disables remote tools and uses only local tools. Cade will also gracefully fall back to local-only mode if Arcade Cloud authentication is not configured.
+
+### Via CLI Flags
+
+```bash
+# Local Ollama server
+cade chat --endpoint "http://localhost:11434/v1" --model "glm-4.7-flash:latest"
+
+# vLLM server
+cade chat -e http://localhost:8000/v1 -m mistral-7b
+```
+
+### Via Environment Variables
+
+```bash
+export OPENAI_BASE_URL="http://localhost:11434/v1"
+export OPENAI_API_KEY="ollama"  # Dummy key for local model
+cade chat --model glm-4.7-flash:latest
+```
+
+### Via Config File
+
+```toml
+# ~/.cadecoder/cadecoder.toml
+
+default_model = "glm-4.7-flash:latest"
+
+[model_settings]
+host = "http://localhost:11434/v1"
+api_key = "ollama"
+```
+
+After configuring the TOML file:
+
+```bash
+cade chat
+```
+
+### `cade chat` Configuration Precedence
+
+Settings are resolved in this order (first is used):
+
+1. CLI flags (`--endpoint`, `--model`)
+2. Environment variables (`OPENAI_BASE_URL`, `OPENAI_API_KEY`)
+3. Config file (`model_settings.host`, `model_settings.api_key`)
+4. Hardcoded defaults
+
 ## Contributing
 
 ### Development Setup
@@ -238,9 +326,7 @@ disabled_tools = []
 ```bash
 git clone https://github.com/arcadeai-labs/cade.git
 cd cade
-uv venv --python 3.11
-source .venv/bin/activate
-uv pip install -e '.[dev]'
+uv sync --extra dev
 ```
 
 ### Run Tests

{cade_cli-0.3.4 → cade_cli-0.4.0}/README.md

@@ -21,7 +21,7 @@ brew install cade
 
 ```bash
 curl -LsSf https://astral.sh/uv/install.sh | sh
-uv pip install cade-cli
+uv tool install cade-cli
 ```
 
 ### Install with pip
@@ -61,6 +61,27 @@ cade -r                        # Resume most recent
 cade resume "my-project"       # Resume by name
 ```
 
+### Authentication
+
+Cade uses Arcade Cloud for authentication and shares credentials with arcade-cli.
+
+```bash
+cade login                      # Log in to Arcade Cloud
+cade logout                     # Log out
+cade whoami                     # Show current login status
+```
+
+### Context Management
+
+Switch between organizations and projects for Arcade Cloud features.
+
+```bash
+cade context show               # Show current org/project
+cade context list               # List available orgs and projects
+cade context switch -i          # Interactive selection
+cade context switch --org my-org --project my-project
+```
+
 ### Single Message Mode
 
 ```bash
@@ -74,6 +95,7 @@ cat error.log | cade -m "What went wrong?"
 |--------|-------------|
 | `-r`, `--resume` | Resume the most recent thread |
 | `-m`, `--message` | Single message mode (non-interactive) |
+| `-L`, `--local-only` | Disable remote tools (use only local tools) |
 | `-v`, `--verbose` | Enable debug logging |
 | `--version` | Show version |
 
@@ -151,9 +173,11 @@ Config is stored in `~/.cadecoder/`:
 | Variable | Description |
 |----------|-------------|
 | `OPENAI_API_KEY` | OpenAI API key |
+| `OPENAI_BASE_URL` | Custom OpenAI-compatible API endpoint |
 | `ANTHROPIC_API_KEY` | Anthropic API key |
 | `ARCADE_API_KEY` | Arcade API key (alternative to OAuth) |
 | `ARCADE_BASE_URL` | Custom Arcade API endpoint |
+| `CADE_LOCAL_ONLY` | Set to `1` to disable remote tools |
 | `CADECODER_HOME` | Override config directory |
 
 ### Example Config
@@ -174,9 +198,73 @@ provider = "openai"
 model = "gpt-4.1"
 
 [tool_settings]
-disabled_tools = []
+# Tool filtering is managed via MCP server configuration
+# See: cade mcp add --help
 ```
 
+## Using Local or Custom LLMs
+
+Cade works with any OpenAI-compatible API, including local servers (Ollama, vLLM, llama.cpp) and alternative cloud providers (Together AI, Groq, Fireworks).
+
+### Local-Only Mode
+
+When using local LLMs, you can skip Arcade Cloud authentication entirely with `--local-only`:
+
+```bash
+# Local Ollama server without Arcade Cloud
+cade chat --local-only --endpoint "http://localhost:11434/v1" --model "llama3"
+
+# Or via environment variable
+CADE_LOCAL_ONLY=1 cade chat --endpoint "http://localhost:11434/v1" --model "llama3"
+```
+
+This disables remote tools and uses only local tools. Cade will also gracefully fall back to local-only mode if Arcade Cloud authentication is not configured.
+
+### Via CLI Flags
+
+```bash
+# Local Ollama server
+cade chat --endpoint "http://localhost:11434/v1" --model "glm-4.7-flash:latest"
+
+# vLLM server
+cade chat -e http://localhost:8000/v1 -m mistral-7b
+```
+
+### Via Environment Variables
+
+```bash
+export OPENAI_BASE_URL="http://localhost:11434/v1"
+export OPENAI_API_KEY="ollama"  # Dummy key for local model
+cade chat --model glm-4.7-flash:latest
+```
+
+### Via Config File
+
+```toml
+# ~/.cadecoder/cadecoder.toml
+
+default_model = "glm-4.7-flash:latest"
+
+[model_settings]
+host = "http://localhost:11434/v1"
+api_key = "ollama"
+```
+
+After configuring the TOML file:
+
+```bash
+cade chat
+```
+
+### `cade chat` Configuration Precedence
+
+Settings are resolved in this order (first is used):
+
+1. CLI flags (`--endpoint`, `--model`)
+2. Environment variables (`OPENAI_BASE_URL`, `OPENAI_API_KEY`)
+3. Config file (`model_settings.host`, `model_settings.api_key`)
+4. Hardcoded defaults
+
 ## Contributing
 
 ### Development Setup
@@ -184,9 +272,7 @@ disabled_tools = []
 ```bash
 git clone https://github.com/arcadeai-labs/cade.git
 cd cade
-uv venv --python 3.11
-source .venv/bin/activate
-uv pip install -e '.[dev]'
+uv sync --extra dev
 ```
 
 ### Run Tests

{cade_cli-0.3.4 → cade_cli-0.4.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "cade-cli"
-version = "0.3.4"
+version = "0.4.0"
 description = "Cade - The CLI Agent from Arcade.dev"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -28,11 +28,11 @@ dependencies = [
     "pydantic[email]>=2.0.0,<3.0.0",
     "toml>=0.10.0,<1.0.0",
     "pyyaml>=6.0,<7.0.0",
-    "arcadepy>=1.3.1",
     "openai>=1.0.0,<2.0.0",
     "anthropic>=0.34.0,<1.0.0",
     "ulid==1.1",
     "arcade-tdk>=2.0.0",
+    "arcade-mcp-server>=1.0.0",
     "arcade-core>=4.1.0,<5.0.0",
     "authlib>=1.6.0,<2.0.0",
     "pyperclip>=1.8.0,<2.0.0",

cade_cli-0.4.0/src/cadecoder/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.4.0"

{cade_cli-0.3.4 → cade_cli-0.4.0}/src/cadecoder/cli/app.py

@@ -13,6 +13,7 @@ from rich.console import Console
 
 from cadecoder import __version__
 from cadecoder.cli.commands import auth, chat
+from cadecoder.cli.commands.context import context_app
 from cadecoder.cli.commands.mcp import mcp_app
 from cadecoder.cli.commands.model import model_app
 from cadecoder.cli.commands.thread import thread_app
@@ -41,6 +42,12 @@ app.command(name="resume", help="Resume the most recent thread or a specific thr
 )
 
 # Register sub-command groups
+app.add_typer(
+    context_app,
+    name="context",
+    help="Manage organization and project context",
+    rich_help_panel="User",
+)
 app.add_typer(mcp_app, name="mcp", help="Manage MCP servers", rich_help_panel="Tools")
 app.add_typer(model_app, name="model", help="Manage AI models", rich_help_panel="Tools")
 app.add_typer(tool_app, name="tools", help="View available tools", rich_help_panel="Tools")

{cade_cli-0.3.4 → cade_cli-0.4.0}/src/cadecoder/cli/commands/chat.py

@@ -1,16 +1,18 @@
 """Interactive chat command for CadeCoder CLI."""
 
+import os
 from typing import Annotated
 
 import typer
 from rich.console import Console
 from rich.markup import escape
 
-from cadecoder.core.config import get_config
-from cadecoder.core.constants import DEFAULT_AI_MODEL
-from cadecoder.core.errors import CadeCoderError, StorageError
+from cadecoder.core.config import get_config, is_local_only_mode
+from cadecoder.core.errors import CadeCoderError, StorageError, get_provider_config_help
 from cadecoder.core.logging import log
 from cadecoder.execution.orchestrator import create_orchestrator
+from cadecoder.providers import initialize_providers
+from cadecoder.providers.base import provider_registry
 from cadecoder.storage.threads import get_thread_history
 from cadecoder.tools.local.git import get_current_branch_name
 from cadecoder.ui.session import main as run_tui_main
@@ -18,19 +20,59 @@ from cadecoder.ui.session import main as run_tui_main
 console = Console(stderr=True)
 
 
+def _configure_custom_endpoint(endpoint: str) -> None:
+    """Configure a custom API endpoint for local LLMs.
+
+    Automatically detects Ollama endpoints and uses the native OllamaProvider
+    for full tool calling support. Falls back to OpenAI-compatible mode for
+    other endpoints.
+
+    Args:
+        endpoint: The custom API endpoint URL (e.g., http://localhost:11434/v1)
+    """
+    # Detect Ollama endpoint (port 11434 or contains "ollama")
+    is_ollama = ":11434" in endpoint or "ollama" in endpoint.lower()
+
+    if is_ollama:
+        # Use native Ollama provider for full tool calling support
+        # Remove /v1 suffix if present (native API doesn't use it)
+        ollama_base = endpoint.rstrip("/").replace("/v1", "").replace("/api", "")
+        os.environ["OLLAMA_BASE_URL"] = ollama_base
+        log.info(f"Detected Ollama endpoint, using native OllamaProvider at {ollama_base}")
+    else:
+        # Use OpenAI-compatible provider for other endpoints
+        os.environ["OPENAI_BASE_URL"] = endpoint
+        if not os.environ.get("OPENAI_API_KEY"):
+            os.environ["OPENAI_API_KEY"] = "local"
+        log.info(f"Using OpenAI-compatible provider with endpoint {endpoint}")
+
+    # Re-initialize providers with the new endpoint
+    provider_registry._providers.clear()
+    provider_registry._default_provider = None
+    initialize_providers()
+
+
 def chat(
     thread_or_name: Annotated[
         str | None,
         typer.Argument(help="Thread ID or Thread Name to resume (optional)"),
     ] = None,
     model: Annotated[
-        str,
+        str | None,
         typer.Option(
             "--model",
             "-m",
             help="AI model to use for the conversation.",
         ),
-    ] = DEFAULT_AI_MODEL,
+    ] = None,
+    endpoint: Annotated[
+        str | None,
+        typer.Option(
+            "--endpoint",
+            "-e",
+            help="Custom API endpoint URL (for OpenAI-compatible APIs like Ollama, vLLM, etc.)",
+        ),
+    ] = None,
     name: Annotated[
         str | None,
         typer.Option(
@@ -47,6 +89,14 @@ def chat(
             help="System prompt to guide the AI assistant's behavior.",
         ),
     ] = None,
+    local_only: Annotated[
+        bool,
+        typer.Option(
+            "--local-only",
+            "-L",
+            help="Disable Arcade Cloud tools (use only local tools).",
+        ),
+    ] = False,
 ) -> None:
     """
     Start an interactive chat session with AI.
@@ -56,17 +106,18 @@ def chat(
     """
     command_name = "chat"
     try:
+        # Figure out which model to use: CLI flag > config > constant
+        resolved_model = model or get_config().settings.default_model
+        if endpoint:
+            _configure_custom_endpoint(endpoint)
+
+        effective_local_only = local_only or is_local_only_mode()
+
         # Preflight: ensure provider/API keys are configured before entering TUI
         try:
-            _ = create_orchestrator()
-        except Exception as e:
-            console.print(
-                "[bold red]Provider configuration error:[/bold red] "
-                "Failed to initialize the AI provider.\n"
-                "Set required API keys (e.g., OPENAI_API_KEY or ANTHROPIC_API_KEY) "
-                "or configure the provider, then try again.\n"
-                f"Details: {str(e)}"
-            )
+            _ = create_orchestrator(local_only=effective_local_only)
+        except Exception:
+            console.print(f"[bold red]Error:[/bold red] {get_provider_config_help()}")
             raise typer.Exit(code=1)
 
         # Get the thread history manager
@@ -110,7 +161,7 @@ def chat(
                     thread = history_manager.create_thread(
                         name=thread_or_name,
                         git_branch=current_branch,
-                        model=model,
+                        model=resolved_model,
                         user_id=user_id,
                     )
                     log.info(
@@ -136,7 +187,7 @@ def chat(
                 thread = history_manager.create_thread(
                     name=name,
                     git_branch=current_branch,
-                    model=model,
+                    model=resolved_model,
                     user_id=user_id,
                 )
                 selected_thread_id = thread.thread_id
@@ -157,9 +208,10 @@ def chat(
         # Launch TUI
         run_tui_main(
             thread_id_to_run=str(selected_thread_id),
-            model=model,
+            model=resolved_model,
             stream=True,
             system_prompt=prompt,
+            local_only=effective_local_only,
         )
     except (StorageError, CadeCoderError) as e:
         console.print(":x: [bold red]Error:[/bold red] " + escape(str(e)))
@@ -179,13 +231,21 @@ def resume(
         typer.Argument(help="Thread name to resume (optional)"),
     ] = None,
     model: Annotated[
-        str,
+        str | None,
         typer.Option(
             "--model",
             "-m",
             help="AI model to use for the conversation.",
         ),
-    ] = DEFAULT_AI_MODEL,
+    ] = None,
+    endpoint: Annotated[
+        str | None,
+        typer.Option(
+            "--endpoint",
+            "-e",
+            help="Custom API endpoint URL (for OpenAI-compatible APIs like Ollama, vLLM).",
+        ),
+    ] = None,
     prompt: Annotated[
         str | None,
         typer.Option(
@@ -194,6 +254,14 @@ def resume(
             help="System prompt to guide the AI assistant's behavior.",
         ),
     ] = None,
+    local_only: Annotated[
+        bool,
+        typer.Option(
+            "--local-only",
+            "-L",
+            help="Only use local tools.",
+        ),
+    ] = False,
 ) -> None:
     """Resume a saved chat thread.
 
@@ -208,17 +276,18 @@ def resume(
     """
     command_name = "resume"
     try:
+        # Figure out which model to use: CLI flag > config > constant
+        resolved_model = model or get_config().settings.default_model
+        if endpoint:
+            _configure_custom_endpoint(endpoint)
+
+        effective_local_only = local_only or is_local_only_mode()
+
         # Preflight provider before entering TUI
         try:
-            _ = create_orchestrator()
-        except Exception as e:
-            console.print(
-                "[bold red]Provider configuration error:[/bold red] "
-                "Failed to initialize the AI provider.\n"
-                "Set required API keys (e.g., OPENAI_API_KEY or ANTHROPIC_API_KEY) "
-                "or configure the provider, then try again.\n"
-                f"Details: {str(e)}"
-            )
+            _ = create_orchestrator(local_only=effective_local_only)
+        except Exception:
+            console.print(f"[bold red]Error:[/bold red] {get_provider_config_help()}")
             raise typer.Exit(code=1)
 
         history_manager = get_thread_history()
@@ -267,9 +336,10 @@ def resume(
 
         run_tui_main(
             thread_id_to_run=str(target_thread.thread_id),
-            model=model,
+            model=resolved_model,
             stream=True,
             system_prompt=prompt,
+            local_only=effective_local_only,
         )
     except (StorageError, CadeCoderError) as e:
         console.print(":x: [bold red]Error:[/bold red] " + escape(str(e)))