inconvo 1.9.0 → 2.0.0

package/README.md

@@ -1,41 +1,71 @@
-# `inconvo CLI`
+# Inconvo CLI
 
-Install Inconvo-flavored assistant-ui components into any project.
+Try out Inconvo without cloning the repository.
 
-## Usage
+## Quick Start
 
-Add the assistant-ui tool components to the current working directory:
+```bash
+npx inconvo dev
+```
+
+This will:
+1. Run the setup wizard (first time only)
+2. Pull Docker images for the dev server
+3. Start the dev server in Docker and the sandbox on your machine
+
+Then open http://localhost:26686 to start chatting with your data.
+
+## Requirements
+
+- Docker (Docker Desktop or Docker daemon running)
+- Node.js 18+ (only for running `npx`)
+
+## Commands
+
+### `inconvo dev`
+
+Start the Inconvo development environment.
 
 ```bash
-npx inconvo@latest add assistant-ui-tool-components
+npx inconvo dev
 ```
 
 Options:
+- `--version <version>` - Use a specific release version
+
+### `inconvo configure`
+
+Re-run the configuration wizard to update your settings.
+
+```bash
+npx inconvo configure
+```
+
+## Configuration
+
+On first run, the CLI will prompt you for:
+
+- **Database connection** - PostgreSQL, MySQL, or SQL Server
+- **OpenAI API key** - For the LLM
 
-- `--cwd <path>` – run the installer against a different project directory
-- `--path <relative>` – override the root folder used for file output (defaults to `src`)
-- `--overwrite` – replace any files that already exist without prompting
-- `--yes` – answer "yes" to interactive prompts
-- `--skip-install` – copy files only and skip dependency installation
+Configuration is stored in `~/.inconvo/config.env`.
+The CLI stores the dev-server SQLite database at `~/.inconvo/data/inconvo.db`.
 
-By default the installer writes everything under `src/`, including the tool components and their supporting `src/lib/inconvo/types.ts`. It also installs external dependencies such as `react-vega`, `vega`, `vega-embed`, `vega-lite`, and `@tanstack/react-table` unless you opt out with `--skip-install`.
+> **Note**: When developing locally from a cloned repo, the dev-server uses `.inconvo.env` in the `apps/dev-server/` directory instead. These are separate configurations.
 
-After the files are copied, the CLI attempts to update `src/components/assistant-ui/thread.tsx` by importing `InconvoTools` and rendering it inside `ThreadPrimitive.Root`. If the file is missing or has been customized heavily, you'll see a warning and can complete the insertion manually.
+## For Development
 
-## Development
+If you want to contribute or develop locally, clone the repository instead:
 
 ```bash
-npm install
-npm run build
+git clone https://github.com/inconvoai/inconvo.git
+cd inconvo
+pnpm install
+pnpm dev
 ```
 
-The component templates live under `templates/assistant-ui/tools`. Update those files whenever the upstream components change and cut a new CLI release.
+See the main repository README for development instructions.
 
-## Releasing
+## License
 
-1. In the npm package settings for `inconvo`, connect this repository as a *Trusted Publisher* (OIDC) so GitHub Actions can publish without storing an npm token.
-2. Follow conventional commit syntax (`feat:`, `fix:`, `chore:`…) when merging into `main`. Semantic Release reads those messages to determine the next version.
-3. Every push to `main` (or a manual *workflow_dispatch*) runs `.github/workflows/release.yml`, which installs dependencies, builds, and executes `npx semantic-release`. When it finds a release-worthy change it will:
-   - publish the package to npm with provenance,
-   - update the GitHub release notes.
-4. Verify with `npx inconvo@latest --help` after the workflow reports success.
+MIT

package/assets/sandbox/Dockerfile

@@ -0,0 +1,17 @@
+FROM docker.io/cloudflare/sandbox:0.7.0-python
+
+# On a Mac with Apple Silicon, you might need to specify the platform:
+# FROM --platform=linux/arm64 docker.io/cloudflare/sandbox:0.4.3
+
+# Install altair for chart generation (outputs Vega-Lite specs)
+RUN --mount=type=cache,target=/root/.cache/pip \
+    pip3 install --no-cache-dir \
+    altair==6.0.0 \
+    jsonschema==4.24.1
+
+# Precompile Python bytecode to speed up imports (~10-20% faster)
+RUN python3 -m compileall -q /usr/lib/python3* /usr/local/lib/python3* 2>/dev/null || true
+
+# Pre-install inconvo helper module so `from inconvo import ...` works
+COPY scripts/inconvo.py /workspace/inconvo.py
+

package/assets/sandbox/scripts/inconvo.py

@@ -0,0 +1,67 @@
+"""Inconvo response helpers for sandbox code execution."""
+import json
+
+def text(message: str) -> None:
+    """Output a text response.
+
+    Args:
+        message: The message to display to the user
+    """
+    print(json.dumps({"type": "text", "message": message}))
+
+def table(df, message: str, columns: list = None) -> None:
+    """Output a table response from a DataFrame.
+
+    Args:
+        df: pandas DataFrame containing the data
+        message: Explanation for the user
+        columns: Optional list of columns to include (default: all)
+    """
+    if columns:
+        df = df[columns]
+
+    # Convert all values to strings for consistent JSON output
+    body = []
+    for row in df.values.tolist():
+        body.append([str(v) if v is not None else "" for v in row])
+
+    print(json.dumps({
+        "type": "table",
+        "message": message,
+        "table": {
+            "head": [str(c) for c in df.columns.tolist()],
+            "body": body
+        }
+    }))
+
+def chart(alt_chart, message: str) -> None:
+    """Output a chart response. MUST pass an Altair Chart object, NOT a dict.
+
+    Args:
+        alt_chart: Altair Chart object created with alt.Chart(df).mark_*().encode(...)
+        message: Explanation highlighting key insights
+
+    Example:
+        import altair as alt
+        c = alt.Chart(df).mark_bar().encode(x='category:N', y='value:Q')
+        chart(c, "Sales by category")
+    """
+    import altair as alt
+    alt.data_transformers.enable('default', max_rows=None)
+    spec = alt_chart.to_dict()
+
+    # Remove $schema if present
+    spec.pop('$schema', None)
+
+    # Convert named datasets to inline data.values
+    if 'datasets' in spec and 'data' in spec:
+        data_name = spec['data'].get('name')
+        if data_name and data_name in spec['datasets']:
+            spec['data'] = {'values': spec['datasets'][data_name]}
+            del spec['datasets']
+
+    print(json.dumps({
+        "type": "chart",
+        "message": message,
+        "spec": spec
+    }))