deploy（部署）

Deploy an agent to Pipecat Cloud. This skill walks through the full deployment process interactively, confirming each step with the user.

Arguments

/deploy [--config <PATH>] [--env <PATH>]

• --config (optional): Path to pcc-deploy.toml. Defaults to pcc-deploy.toml in the current directory.
• --env (optional): Path to .env file for creating secrets. If not provided, the skill will look for a .env file in the same directory as the config file.

Examples:

• /deploy
• /deploy --config examples/mybot/pcc-deploy.toml
• /deploy --config examples/mybot/pcc-deploy.toml --env examples/mybot/.env

Prerequisites

Before starting, verify these prerequisites and inform the user about any that are missing:

1. Pipecat Cloud CLI: Check if pc is installed by running pc --version. If not installed, tell the user to install it with uv tool install pipecat-ai-cli and stop.
2. Pipecat Cloud authentication: Check if pc cloud auth whoami succeeds. If not authenticated, run pc cloud auth login --headless as a background task, then read the output file to extract the URL and six-digit code. Share both with the user so they can authenticate in their browser. Wait for the background task to complete before proceeding.
3. Configuration file: Read the pcc-deploy.toml file (from --config argument or current directory). If it doesn't exist, tell the user they need a pcc-deploy.toml and stop. Parse and display the configuration to the user (agent_name, image, secret_set, etc.).

If any prerequisite fails, stop and clearly explain what needs to be fixed. Do not proceed to the next steps.

Deployment Method

After prerequisites pass, ask the user how they want to deploy using AskUserQuestion:

• Cloud Build (Recommended): Pipecat Cloud builds your Docker image from source. No local Docker required.
• Self-managed image: Build and push your own Docker image from your machine, then deploy.

Important: Running Commands

All pc cloud commands must be run from the directory containing the pcc-deploy.toml file. Use cd <dir> && <command> to ensure the correct working directory.

Several pc cloud commands prompt for interactive confirmation which doesn't work in this environment. Use --force to skip confirmation prompts:

• pc cloud deploy: Use --force (e.g., pc cloud deploy --force)
• pc cloud docker build-push: Use --force (e.g., pc cloud docker build-push --force)
• pc cloud secrets set: Use the --skip flag (e.g., pc cloud secrets set NAME --file.env --skip)

Cloud Build Path

This path uses Pipecat Cloud to build your Docker image from source. No local Docker installation is required.

Step 1: Verify Dockerfile and Lockfile

Check that a Dockerfile exists in the build context directory (the directory containing pcc-deploy.toml, or the context_dir from the [build] section of the config). If no Dockerfile is found, tell the user they need one and stop.

If the Dockerfile references uv (e.g., uv sync, uv pip install, COPY uv.lock), check that a uv.lock file exists in the build context directory. If it doesn't exist, run uv lock to generate it. If it already exists, run uv lock to ensure it's up to date. The lockfile must be present and current because cloud builds run remotely and cannot generate it.

Step 2: Secrets Setup

Ask the user if they need to create or update secrets for this deployment.

• If yes, determine the env file path (from --env argument, or look for .env in the same directory as the config file, or ask the user).
• Read the secret_set name from the pcc-deploy.toml configuration.
• Run: pc cloud secrets set {SECRET_SET_NAME} --file {ENV_FILE_PATH} --skip
• Show the output to the user.
• If no, skip this step.

Step 3: Deploy

Ask the user to confirm they want to deploy the agent.

• Show a summary of what will be deployed (agent_name, secret_set from the config).
• Run from the config directory: pc cloud deploy --force
• Use a generous timeout (10 minutes) as the cloud build and deployment can take a while.
• This command handles both the cloud build and the deployment automatically.
• If the command times out but no error occurred, retrieve the build ID from the command output and run pc cloud build logs {BUILD_ID} to check build progress. Share the output with the user.
• If the build succeeds but deployment times out, check agent logs with pc cloud agent logs {AGENT_NAME} and share with the user.
• Show the deployment output and status to the user.

Cloud Build Error Handling

• If the build fails, retrieve the build ID from the command output and run pc cloud build logs {BUILD_ID} to show the user what went wrong.
• Common cloud build issues:

- Missing or invalid Dockerfile - Build context too large (500MB limit) — check for large files that should be in .dockerignore - Build timeout — the build exceeded the maximum duration

Self-Managed Image Path

This path builds and pushes the Docker image from your machine. Requires Docker to be installed and running.

Step 1: Docker Prerequisites

Check these additional prerequisites:

• Docker: Check if docker info succeeds (daemon running). If not, tell the user to start Docker and stop.
• Docker login: Check if docker login succeeds. If not logged in, tell the user to run docker login and stop.

Step 2: Secrets Setup

Ask the user if they need to create or update secrets for this deployment.

• If yes, determine the env file path (from --env argument, or look for .env in the same directory as the config file, or ask the user).
• Read the secret_set name from the pcc-deploy.toml configuration.
• Run: pc cloud secrets set {SECRET_SET_NAME} --file {ENV_FILE_PATH} --skip
• Show the output to the user.
• If no, skip this step.

Step 3: Build and Push Docker Image

Ask the user if they want to build and push the Docker image.

• If yes:

- First, check if uv.lock exists in the config directory. If so, run uv lock to ensure it's up to date before building. - Run from the config directory: pc cloud docker build-push --force - Use a generous timeout (5 minutes) as builds can take a while. - If the build fails due to a stale lockfile, run uv lock in the config directory and retry. - If the build fails for other reasons, show the error and ask the user how to proceed.

• If no, skip this step (image may already be pushed).

Step 4: Deploy

Ask the user to confirm they want to deploy the agent.

• Show a summary of what will be deployed (agent_name, image, secret_set from the config).
• Run from the config directory: pc cloud deploy --force
• Use a generous timeout (5 minutes) as deployment may take time to reach ready state.
• If deployment times out but no error occurred, check logs with pc cloud agent logs {AGENT_NAME} and share with the user — the deployment may still be starting up.
• Show the deployment output and status to the user.

Self-Managed Image Error Handling

• Common issues:

- Docker not logged in to the image registry (docker login) - Missing image field in pcc-deploy.toml - Stale uv.lock file — run uv lock to fix

General Error Handling

• If any pc command fails, show the full error output and explain what might have gone wrong.
• Common issues for both paths:

- Invalid or expired Pipecat Cloud authentication - Missing or malformed pcc-deploy.toml - Secret set name mismatch between config and what exists in Pipecat Cloud

Completion

After a successful deployment, summarize what was done:

• Deployment method used (cloud build or self-managed image)
• Secrets created/updated (if applicable)
• Image built and pushed (if applicable)
• Agent deployed with name from config