Skip to content

Latest commit

 

History

History
48 lines (37 loc) · 1.8 KB

File metadata and controls

48 lines (37 loc) · 1.8 KB

Agent Runtime (LangGraph)

Status: Partial
Package: beep/agent/

Purpose

Autonomous multi-step coding agent: tool loop, approvals, LangGraph graph, provider packs, subagents, structured output, portable bundles.

Code locations

Area Path
Graph graph.py, graph_runner*.py, graph_streaming.py
Loop (legacy/bridge) loop.py
Tools agent/tools/*, tools/factory.py
Providers provider_*.py, provider_plugins.py
Subagents subagents/
Planning planning/, tools/todo_tool.py
Verification verification/
Bundles bundle_contract.py, bundle_store.py
Commands beep/commands/agent*.py

User surfaces

  • beep agent "<goal>" and agent admin subcommands
  • REPL /agent
  • Provider setup: beep agent setup, providers, configure

Current behavior

  • LangGraph runtime with streaming events (Phase 11, 15)
  • Sandbox modes: read-only, workspace-write, full-trust
  • Parallel tools, todo list, dispatch_agent subagents, auto-verify hooks
  • Multimodal and structured output CLI flags (Phase 16)
  • Portable bundle import/export/deploy (Phase 17–18)

Enhancement backlog

ID Enhancement Priority Verification
AG-1 Coordinated multi-agent blackboard (shared state across workers) P2 Design doc + integration test
AG-2 Concurrent agent workers with resource limits P2 Stress test
AG-3 Agent run replay from JSONL in beep agent resume P1 Fixture replay test
AG-4 Cost/token budget per run with hard stop P1 Agent test
AG-5 Tool result cache for idempotent reads P2 Unit test
AG-6 Offline mode: local tools only with clear banner P3 CLI test
AG-7 Graph visualization export (--graph-dot) P3 Smoke test