docs(ai): document the core LLM-tools preset#3774
Conversation
Add a guide for the three-tool core preset (agent_inspect, agent_recipe, execute_code), where SuperDoc keeps control of targeting, validation, edits, verification, and receipts. Wire it into the AI overview, the LLM Tools page, the docs nav, and llms.txt. Tool names and the recipe list are taken from the SDK source so the page matches what the model actually sees.
|
📖 Docs preview: https://superdoc-caio-pizzol-docs-ai-core-preset.mintlify.app |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6a20c7d9a5
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| } from '@superdoc-dev/sdk'; | ||
|
|
||
| // Provider-shaped tool definitions for the core preset. | ||
| const { tools } = await chooseTools({ provider: 'openai', preset: 'core' }); |
There was a problem hiding this comment.
When readers copy this example, chooseTools({ preset: 'core' }) resolves through the SDK preset registry, but the Node registry currently only contains legacy (packages/sdk/langs/node/src/presets.ts:162) and the Python registry mirrors that. That means the documented call throws PRESET_NOT_FOUND instead of returning the three advertised tools; dispatchSuperDocTool also still dispatches only the default preset. Please either land/register the core preset and preset-aware dispatch first, or avoid publishing these docs as a working API.
Useful? React with 👍 / 👎.
1 participant
Documents the core LLM-tools preset: a compact three-tool surface (
agent_inspect,agent_recipe,execute_code) where SuperDoc keeps control of targeting, argument validation, edits, verification, and receipts. It sits next to the existing LLM Tools page, which covers the broader grouped intent catalog, and gives embedded agents a smaller surface that is easier to audit and verify.The new page is linked from the AI overview, the LLM Tools page, the docs nav, and
llms.txt. Tool names and the 29-recipe list are taken from the SDK source so the page matches what the model actually sees, rather than a hand-maintained list that drifts.Verified: recipe names cross-checked against the SDK recipe registry (29, exact match);
docs.jsonparses; internal doc links resolve to existing pages.