docs: document the Anthropic automatic prompt caching env vars

[tin-berri]

Summary

Documents automatic Anthropic prompt caching across the environment settings reference and prompt caching tutorial. The reference adds LITELLM_ENABLE_ANTHROPIC_PROMPT_CACHING and LITELLM_ANTHROPIC_PROMPT_CACHING_TTL; the tutorial explains the provider-credential sharing boundary, gateway-wide checkpoint injection for Claude models, supported models and endpoints, TTL and precedence rules, and when to use manual injection points instead

These variables were added in BerriAI/litellm#33573. Its documentation job checks this repo's default branch and stays red until every environment variable read by the codebase has a row in docs/proxy/config_settings.md. The validation exits 0 against this branch and reports both keys as undocumented without the new rows

The rows remain alphabetized among the existing LITELLM_ANTHROPIC_* settings. A local Docusaurus production build also completes successfully

Screenshots

Rendered from current head 40174df1 using the local Docusaurus production build

Shared cache boundary and automatic checkpoints	Injection behavior, cache lifetime, and manual fallback
Environment settings reference rows

Link to Devin session: https://app.devin.ai/sessions/f03da2725ec94d28b3facf766871b102

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Jul 17, 2026 1:02am