Skip to content

CLAUDE.md

Latest commit

 

History

History
133 lines (103 loc) · 4.18 KB

CLAUDE.md

File metadata and controls

133 lines (103 loc) · 4.18 KB
Raw

CLAUDE.md

Guidance for Claude Code when working with this repository.

Project Overview

lyzr-kit is a Python SDK for managing AI agents via the Lyzr platform.

CLI Commands

lk auth                    # Configure API credentials
lk ls                      # List agents (two tables)
lk get <source>            # Clone and deploy agent (creates copy-of-<name>)
lk set <identifier>        # Update agent on platform
lk chat <identifier>       # Interactive chat session
lk rm <identifier>         # Delete local agent
lk tree [identifier]       # Show agent dependency tree
lk doctor                  # Validate all local agents

Note: agent resource is optional. lk ls = lk agent ls = lk a ls

Serial numbers: Context-aware

  • get → Built-in agents
  • set/chat/rm → Your agents

Agent Get Flow

The get command uses a plan-based approach:

  1. Build clone plan (collects all dependencies recursively)
  2. Display plan with single confirmation prompt
  3. Execute plan (creates agents in dependency order)
$ lk get project-manager

Creating 'copy-of-project-manager' with 3 sub-agent(s):

  AGENT                        ACTION         FROM
  copy-of-task-planner         create         task-planner
  copy-of-data-analyst         create         data-analyst
  copy-of-summarizer           create         summarizer
  copy-of-project-manager      create         project-manager

Proceed? [Y/n]: y

No ID prompts - users rename via YAML edit + lk set.

Storage

Location Purpose
src/lyzr_kit/collection/agents/ Built-in agents (bundled)
agents/ User agents (via lk agent get)

Project Structure

src/lyzr_kit/
├── main.py              # CLI entry point
├── schemas/             # Pydantic models (agent.py, tool.py, feature.py)
├── collection/agents/   # Built-in agent YAMLs
├── commands/            # CLI implementations
│   ├── _console.py      # Shared Rich console
│   ├── _resolver.py     # Serial number resolver
│   ├── _websocket.py    # WebSocket event streaming
│   ├── agent.py         # Agent Typer app
│   ├── agent_list.py    # ls command
│   ├── agent_get.py     # get command (plan-based cloning)
│   ├── agent_set.py     # set command
│   ├── agent_rm.py      # rm command (with --tree flag)
│   ├── agent_tree.py    # tree command
│   ├── agent_doctor.py  # doctor command
│   ├── agent_chat.py    # chat command (SSE + WebSocket)
│   ├── auth.py          # auth command
│   ├── tool.py          # stub
│   └── feature.py       # stub
├── storage/             # StorageManager, serialization, validation
└── utils/               # auth.py, platform.py

tests/
├── unit/commands/       # Command tests
├── unit/storage/        # Storage tests
└── integration/         # E2E tests

Chat Implementation

Key files for chat functionality:

  • commands/agent_chat.py - Main chat loop, UI boxes, SSE streaming
  • commands/_websocket.py - WebSocket client, event parsing

Features:

  • Session box (agent info, model, session ID, timestamp)
  • Real-time WebSocket events (tool calls, memory, artifacts)
  • SSE streaming for responses
  • Metrics footer (latency, tokens)
  • prompt_toolkit for keyboard shortcuts

Sub-agent Features

  • Cycle detection - Prevents circular dependencies (A → B → A)
  • Dependency plan - Shows all agents to create before confirmation
  • Tree view - lk tree shows dependency graph
  • Doctor - lk doctor validates all local agents
  • Recursive delete - lk rm --tree deletes agent + sub-agents

Build Commands

pip install -e .        # Install dev mode
pytest tests/ -v        # Run tests
ruff check src/         # Lint
mypy src/               # Type check

Phases

Phase Focus Status
1 Agents, CLI, storage ✅ Done
2 Chat experience, WebSocket ✅ Done
3 Sub-agents ✅ Done
4 Tools Stub
5 Features Stub

Specs

  • specs/concepts/ - Entity definitions
  • specs/implementation/ - Technical details
  • specs/phases/ - Roadmap