Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.
Mirrored from https://github.com/Camo67/composio.git
Synced Nov 21, 11:00 UTC
docs: clarify tool discovery and direct execution guidance (#4583)
## Summary The legacy direct-execution guides overpromise search relevance and omit key tool-selection details. Correct those examples while making sessions the recommended path for new integrations and encouraging existing users to migrate. Legacy methods remain supported. ## Changes - Replace semantic-search and "top N" promises with action-keyword examples, schema inspection, and explicit tool selection guidance. - Explain how TypeScript's `limit` disables automatic `important` filtering and show `important: true`; distinguish Python's available options. - Preserve legacy page and navigation labels. Recommend sessions in the quickstart, comparison guide, direct-execution callouts, and agent-readable documentation; link to migration guidance and the session direct-tools preset. - Surface `link()` for managed OAuth near the top of authentication setup. - Generate `toolkits.get({})` in the reference usage example and retain only that generated correction. ## Type of change - [x] Documentation ## How Has This Been Tested? - `bun run lint` in `docs/`: passed with existing warnings. - `bun run lint:links` in `docs/`: passed, no broken links. - `bun run types:check` in `docs/`: passed. - `bun run test` in `docs/`: 590 tests passed. - `bun run test test/scripts/generate-docs.test.ts` in `ts/packages/core/`: 15 tests passed; the new Toolkits usage assertion also fails when the old `list()` output is restored. - `bun run scripts/generate-docs.ts` in `ts/packages/core/`: regenerated the corrected usage example; unrelated generated API changes excluded. - Agent-skill and skill-routing validators: passed. - `bun run build` in `docs/`: passed before the final prose-only legacy/session guidance revision. - All six TypeScript examples in the fetching-tools guide also passed a focused Twoslash check. - After the legacy/session guidance revision: lint, type-checking, links, both guidance validators, and 14 focused navigation/session-execution tests passed. Code examples are unchanged by this revision. ## Checklist - [x] I updated documentation as needed. - [x] Added a focused generator-output regression test for `toolkits.get({})`; runtime SDK tests are not applicable. - [x] No changeset needed: no published SDK behavior changes. ## Additional context SDK behavior, backend validation, search ranking, and toolkit bugs from the feedback are outside this PR. Search guidance does not assert an unverified backend matching algorithm.
No changes.