Defines tool schemas and helper functions for LLM-generated artifact creation.
# Source: https://github.com/mindsdb/anton | License: MIT | Copyright 2026 MindsDB
# Provided via justclaw.it.com; keep this notice with the code.
"""Sub-tools exposed to the inner generation LLM.
The write loop offers four:
- ``write_file(path, mode)`` — produce one file inside the artifact folder;
the body travels as plain text in the reply (see the file-body protocol).
- ``read_text_file`` — the main agent's tool; the engine reads paths
relative to the artifact folder and never returns line numbers.
- ``finish(summary)`` — terminal tool; signals the loop to stop.
- ``scratchpad`` — the main agent's scratchpad, unchanged.
Each handler accepts the artifact ``root`` plus the sub-agent's input dict and
returns a string the engine forwards back t
... (truncated -- full source via MCP)
See the full source, get the GitHub permalink, and search for more like it.
Get a free API key