docs: add proactive compression docs

[opieter-aws]

Description

Add docs for proactive compression and token counting features

Related Issues

strands-agents/sdk-python#2239

Type of Change

• New content

Checklist

• I have read the CONTRIBUTING document
• My changes follow the project's documentation style
• I have tested the documentation locally using npm run dev
• Links in the documentation are valid and working

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[github-actions]

Documentation Preview Ready

Your documentation preview has been successfully deployed!

Preview URL: https://d3ehv1nix5p99z.cloudfront.net/pr-cms-806/docs/user-guide/quickstart/overview/

Updated at: 2026-05-13T18:43:41.048Z

[github-actions]

Assessment: Comment

The documentation is well-written and clearly explains a complex feature (proactive context compression). The code snippets follow existing patterns, the section structure is logical, and cross-references between token counting and proactive compression are solid.

Review Notes

• Correctness: The token counting snippet in index.ts references an undefined agent variable — see inline comment for details.
• Design doc alignment: The design doc (0008) uses threshold as the parameter name while these docs use compressionThreshold. If the implementation landed with compressionThreshold, this is fine (design doc is aspirational), but worth confirming.

Nice work documenting both the high-level concept and the internals (token estimation strategy, provider support table).

[github-actions]

Assessment: Approve

This PR adds clear, well-organized documentation for proactive context compression. The code snippets follow established patterns (function scoping, snippet markers), cross-references between pages are consistent, and the external link to the SDK source code is valid. The prior review comment about the undefined agent variable in the token counting snippet has been resolved. No new issues identified.

[github-actions]

Assessment: Approve

The PR has been significantly improved since the prior review. The restructuring — distributing token counting docs to individual provider pages and using proactiveCompression: true | { compressionThreshold } API — is cleaner and more discoverable. The documentation is thorough and well-organized.

Review Notes

• Structure: Good decision to add Token Counting subsections on each provider page rather than a centralized section. The cross-references between pages (relative links from conversation-management, /docs/-prefixed links from provider pages) follow existing repo patterns.
• API clarity: The dual-form proactiveCompression option (true for default, object for custom threshold) is clearly documented with both the shorthand and explicit usages.
• Best-effort semantics: The explanation that proactive compression never breaks a working agent (line 354) is an important detail for users.
• Prior feedback addressed: The undefined agent variable issue is resolved (token counting snippet removed and replaced by per-provider docs).

[github-actions]

Assessment: Approve

All previously raised feedback has been addressed. The addition of Python code examples, simplified caution section, and generalized token counting intro text ("context management strategies") are good improvements. No new issues found.

[github-actions]

Assessment: Approve (one minor note inline)

The PR is in great shape. All prior feedback has been addressed — typos fixed, Python examples added, caution section simplified, token counting intro generalized. The only remaining item is the "(TypeScript)" label on line 99 of conversation-management.mdx which is now outdated since the section covers both languages.