Skip to content

Pepebot v0.5.7

Choose a tag to compare

@github-actions github-actions released this 07 Mar 02:52
· 95 commits to main since this release
0e71799

🐸 Pepebot v0.5.6 - Google Vertex AI Support

Release Date: 2026-03-02

πŸŽ‰ What's New

☁️ Google Vertex AI Provider

Pepebot now supports Google Cloud Vertex AI natively using service account credentials (no API key needed).

  • Authenticate via service account JSON file (account_services.json)
  • Configure credentials_file, project_id, and region
  • Full Chat and Streaming support with Gemini models
  • Tool calling (function declarations) fully supported
  • Supports both regional and global endpoints

Quick Setup:

{
  "agents": {
    "defaults": {
      "model": "vertex/gemini-2.0-flash",
      "provider": "vertex"
    }
  },
  "providers": {
    "vertex": {
      "credentials_file": "~/.config/service-accounts.json",
      "project_id": "your-gcp-project-id",
      "region": "global"
    }
  }
}

Or via environment variables:

export PEPEBOT_PROVIDERS_VERTEX_CREDENTIALS_FILE=~/.config/service-accounts.json
export PEPEBOT_PROVIDERS_VERTEX_PROJECT_ID=your-gcp-project-id
export PEPEBOT_PROVIDERS_VERTEX_REGION=global
export PEPEBOT_AGENTS_DEFAULTS_MODEL=vertex/gemini-2.0-flash

🎯 Explicit Provider Option

New optional provider field in agent config β€” override automatic provider detection from model name prefix.

{
  "agents": {
    "defaults": {
      "model": "gemini-2.0-flash",
      "provider": "vertex"
    }
  }
}

Supported values: vertex, maiarouter, openrouter, anthropic, openai, gemini, zhipu, groq, vllm

Environment variable: PEPEBOT_AGENTS_DEFAULTS_PROVIDER

πŸ€– Better Multi-Agent Controls (manage_agent)

manage_agent is now more complete for agent lifecycle and delegation:

  • remove action to delete an agent from registry (optional file cleanup with remove_files)
  • call action to directly invoke another named agent and get its response
  • assign_skill action to persist skill assignments into per-agent memory (<agent_dir>/memory/MEMORY.md)
  • register now supports optional provider, so each agent can use different provider/model settings

Prompt behavior has also been tightened so normal requests like "panggil agent" are less likely to be misinterpreted as workflow creation.


πŸ“¦ Installation

Using Install Script (Recommended)

curl -fsSL https://raw.githubusercontent.com/pepebot-space/pepebot/main/install.sh | bash

Using Homebrew

brew tap pepebot-space/pepebot
brew install pepebot

Using Docker

docker pull ghcr.io/pepebot-space/pepebot:latest
docker run -it --rm pepebot:latest

Manual Download

Download the appropriate binary for your platform from the releases page.

πŸš€ Quick Start

  1. Create a service account at Google Cloud Console
  2. Download the JSON key file
  3. Configure Pepebot:
    pepebot onboard
    # Or manually set environment variables
  4. Start chatting:
    pepebot agent -m "Hello from Vertex AI!"

πŸ“š Documentation

πŸ”— Links

πŸ“ Full Changelog

For a complete list of changes, see CHANGELOG.md.

πŸ“¦ Platform-Specific Downloads

Choose the appropriate binary for your system:

Linux

  • x86_64: pepebot-linux-amd64.tar.gz
  • ARM64: pepebot-linux-arm64.tar.gz
  • ARMv7: pepebot-linux-armv7.tar.gz (Raspberry Pi 3/4)
  • ARMv6: pepebot-linux-armv6.tar.gz (Raspberry Pi 1/Zero)
  • RISC-V 64: pepebot-linux-riscv64.tar.gz
  • MIPS: pepebot-linux-mips.tar.gz / pepebot-linux-mipsle.tar.gz
  • MIPS64: pepebot-linux-mips64.tar.gz / pepebot-linux-mips64le.tar.gz

macOS

  • Intel: pepebot-darwin-amd64.tar.gz
  • Apple Silicon (M1/M2/M3): pepebot-darwin-arm64.tar.gz

Windows

  • x86_64: pepebot-windows-amd64.tar.gz
  • ARM64: pepebot-windows-arm64.tar.gz

BSD

  • FreeBSD x86_64: pepebot-freebsd-amd64.tar.gz
  • FreeBSD ARM64: pepebot-freebsd-arm64.tar.gz

Android

  • ARM64: pepebot-android-arm64.tar.gz (Most devices, Android 5.0+)

πŸ” Verification

All binaries include SHA256 checksums. Verify your download:

sha256sum -c pepebot-*.tar.gz.sha256

Built with ❀️ using Go