Cloudflare

Add documentation specifying maximum retention period for Agent Memory sessions and hard caps on storage per agent (e.g., in limits/pricing pages or the Agent Memory API reference).

The retrieved chunks do not contain any information about Cloudflare's Certificate Transparency log shard size limits or STH (signed tree head) frequency. The ct.cloudflare.com pages mention CT logs and monitoring but lack technical specifications about log internals. Consider adding documentation about CT log operational parameters such as shard size limits and STH publication frequency for transparency and monitoring purposes.

The retrieved chunks discuss Cloudflare Email Service for agents being in public beta and technical implementation details, but contain no information about daily sending limits per domain or reputation thresholds for throttling/suspension. These operational limits and policies need to be documented, likely on a dedicated limits, quotas, or pricing page for Email Service.

Add documentation on account-level limits for feature flags, targeting rules per flag, and segments. Buyers evaluating rollout strategy need hard numbers on maximum flags per account and any caps on rules/segments per flag to assess if the platform can scale to their needs.

The retrieved chunks from the 'Agents Week: network performance update' blog post discuss general network performance improvements (new PoP locations in Constantine, Malang, and Wroclaw; software improvements via HTTP/3 and congestion window management; Rust-based FL2 architecture) but do not mention specific latency improvements for AI inference traffic or which PoP regions received upgraded backbone capacity. The chunks appear to be from a blog post that exists, but the specific content about AI inference latency improvements and backbone capacity upgrades is not present in the retrieved text.

Add explicit compatibility information stating whether Unweight-compressed models can run on standard GGUF/ONNX runtimes or require Cloudflare's custom inference stack (Infire/Omni).

The retrieved chunks mention rate limiting as a feature of AI Gateway but provide no specifics on default rate limits per provider (OpenAI, Anthropic) or whether burst handling is configurable. The chunks link to a 'Use Rate limiting' documentation page at https://developers.cloudflare.com/ai-gateway/ai-gateway/features/rate-limiting/ but the actual content of that page was not retrieved. Need to retrieve the specific rate limiting documentation page for AI Gateway to answer this technical configuration question.

The blog post should explicitly state what score threshold (e.g., 80/100, 90/100) or 'level' constitutes an 'optimized' or 'agent-ready' site, and whether there are tiered ratings (good/great/excellent).

The chunks mention that AI Crawl Control 'Track robots.txt' feature helps 'identify which crawlers are violating your directives' and that it 'Manage[s] crawler activity via block or allow rules', but never explicitly states whether AI Crawl Control overrides existing robots.txt rules or requires removing conflicting rules. Need documentation on interaction between AI Crawl Control rules and robots.txt directives.

The documentation should clarify whether 308 redirects are supported (for preserving POST bodies) and whether users can configure redirect chain limits or behavior for multi-hop canonical chains.

Add explicit pricing for Git LFS large object storage, clarifying whether it uses the standard $0.50/GB-mo rate or a different tier.

Add specific documentation about compression ratios for JSON API responses (not just JS bundles) and clearly document any dictionary size limits or cache variant limits that affect hit rates.

The chunks mention that Hyperdrive 'manages database connection pools' but do not specify the default concurrent connection limit or whether manual configuration is required. Add explicit documentation on default pool size, max connections, and whether limits are automatic or configurable.