Skip to content

added adapted scripts. added supports for both gpt and open-weights models. added supported for web search and better markdown rendering.#4

Open
jaaack-wang wants to merge 21 commits into1rgs:masterfrom
jaaack-wang:master
Open

added adapted scripts. added supports for both gpt and open-weights models. added supported for web search and better markdown rendering.#4
jaaack-wang wants to merge 21 commits into1rgs:masterfrom
jaaack-wang:master

Conversation

@jaaack-wang
Copy link

added adapted scripts. added supports for both gpt and open-weights models. added supported for web search and better markdown rendering.

Zhengxiang Wang and others added 21 commits January 13, 2026 00:11
Now each tool has danger level, it is one of:
- safe: nor writes, nor reads user data
- sensitive: reads user data
- dangerous: writes user data
New options `--safe_tools` allow to specify which tools are considered
to be safe and can run without any confirmation. Any other tool call
will require confirmation from user (yes/no prompt).

Also improved tool call preview printing with formatter for all
arguments and custom formatters for "write" and "edit" tools (they now
print full context of what will be changed)
NOTE: Only basic testing eas done, because I does not have access to
that AI
NOTE: Only basic testing was done because I does not have access to this
AI's API
NOTE: Not really tested because my laptop can't really handle local AI
Added `SENSITIVE_COMMANDS` array and check if command is one of them. If
so, call's danger is demoted to "sensitive"
Added tool safety feature and oauth2 log in for gemini
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants