Skip to content

CHAT_README.md

Latest commit

 

History

History
111 lines (77 loc) · 2.78 KB

CHAT_README.md

File metadata and controls

111 lines (77 loc) · 2.78 KB

Agent Engine Chat Client

Interactive command-line client for chatting with your deployed agent on Vertex AI Agent Engine.

Features

  • Clean turn-based conversation interface
  • Color-coded user and agent responses
  • Automatic session management and memory persistence
  • Streaming responses from the remote agent
  • Session saved to memory on exit

Quick Start

ae -p PROJECT_ID -l us-central1 chat AGENT_ID

Command-Line Options

ae chat --help

Usage: ae chat [OPTIONS] AGENT_ID

  Start an interactive chat session with an agent.

Arguments:
  AGENT_ID  Agent ID or full resource name [required]

Options:
  -u, --user TEXT       User ID for the chat session [default: cli-user]
  -d, --debug           Enable verbose HTTP debug logging
  --help                Show this message and exit.

Global Options (can be placed before 'chat'):
  -p, --project TEXT    Google Cloud project ID
  -l, --location TEXT   Google Cloud region
  --base-url TEXT       Override the Vertex AI base URL
  --api-version TEXT    Override the API version

Usage Examples

# Basic chat session
ae -p my-project -l us-central1 chat 172357243746910208

# With custom user ID
ae -p my-project -l us-central1 chat 172357243746910208 --user john@example.com

# With debug logging to see HTTP requests/responses
ae -p my-project -l us-central1 chat 172357243746910208 --debug

Interactive Session

  1. Start the client:

    ae -p PROJECT_ID -l LOCATION chat AGENT_ID

  2. Chat with your agent:

    • Type your message and press Enter
    • Agent responses stream in real-time
    • Tools used by the agent are displayed

  3. Exit gracefully:

    • Type quit or exit
    • Or press Ctrl+C

Example Session

Ready. User: cli-user, Session: abc123...

Type your message and press Enter. Type 'quit' or 'exit' to end.

You: Hello! How are you?

Agent: Hello! I'm doing well, thank you for asking. How can I help you today?

You: What can you do?

Tools: [search_documents] [get_user_info]

Agent: I can help you with several things including searching documents,
retrieving user information, and answering questions. What would you like to do?

You: quit
Chat session ended.

How It Works

  1. Connects to your deployed Agent Engine on Vertex AI
  2. Creates a new session (or resumes existing one)
  3. Sends your messages to the remote agent
  4. Streams responses back in real-time
  5. Saves conversation to memory bank on exit

Troubleshooting

Error: "Error in chat session: ..."

  • Check your GCP credentials: gcloud auth application-default login
  • Verify agent is deployed: ae -p PROJECT_ID -l LOCATION list
  • Ensure the agent ID is correct

No response from agent:

  • Check your network connection
  • Verify agent is running: ae -p PROJECT_ID -l LOCATION get AGENT_ID