vrr/agent-zero
2
0
Fork 0
mirror of https://github.com/agent0ai/agent-zero.git synced 2026-05-23 04:17:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4
agent-zero/knowledge/main/about/architecture.md
Alessandro daf95ec3ab Normalize tool contracts and slim prompt surface 
Standardize multi-action tools around tool_args.action while keeping parser compatibility for older tool/args, tool_name:action, and method-shaped requests. This keeps new prompts clean without breaking agents that learned the previous dialect.

Move A0 connector remote execution/file tools into stable standard prompts, make remote targeting independent of the active chat context, and skill-gate beta computer-use remote so it no longer weighs down the always-on tool list.

Align text editor, scheduler, skills, office artifact, memory, notify, and browser prompts/tools around the canonical action contract. Add scheduler update/timezone handling, skills_tool read_file, text editor patch coverage, and fixes for memory_forget, behaviour_adjustment, and code execution progress warnings.

Reduce default prompt pressure by compacting browser and scheduler prompts into skill-backed manifests, shortening skill catalog descriptions, and pruning noisy framework knowledge. Remove obsolete connector prompt stubs and root tool-call knowledge examples.

Tests: conda run -n a0 pytest tests/test_a0_connector_prompt_gating.py tests/test_tool_action_contracts.py tests/test_task_scheduler_timezone.py tests/test_text_editor_context_patch.py tests/test_tool_request_normalization.py tests/test_office_document_store.py::test_odf_is_advertised_and_docx_remains_explicit_compatibility tests/test_office_document_store.py::test_document_artifact_accepts_method_alias_for_ods_create tests/test_skills_runtime.py tests/test_default_prompt_budget.py::test_a0_small_profile_removed_and_prompt_text_generic -q
2026-05-09 21:54:43 +02:00

1 KiB
Raw Permalink Blame History

Agent Zero Architecture

The agent loop builds a system prompt, appends conversation history, asks the model for one JSON tool request, executes that tool, records the result, and repeats until response ends the task.

Key runtime files:

  • agent.py: Agent, AgentContext, loop state, tool dispatch
  • initialize.py: framework initialization
  • run_ui.py: Web UI entry point
  • helpers/: shared framework helpers
  • tools/: core tools
  • plugins/: framework plugins
  • usr/: user data, custom plugins, settings, workdir

Prompt assembly is file-based. Main prompts come from prompts/, profile overrides from agents/<profile>/prompts/, and plugin prompts from plugins/<plugin>/prompts/.

Plugins can add tools, prompts, API handlers, Web UI components, extensions, and hooks. User plugins live in usr/plugins/ and should survive updates.

Memory and knowledge use the memory plugin and vector search. Knowledge files are indexed for recall; they should be concise because irrelevant recall can steer behavior badly.