Changelog
What shipped in Compresr, month by month — models, SDKs, Gateway, and dashboard.
July 2026
Enterprise workflow completed end-to-end.
NewEnterprise email-whitelist auto-approval — customers from allowlisted company domains get their first API key without waiting for support approval.
NewSDK CLI with browser-based login and logout — no more copy-pasting API keys locally.
NewAgentCore joins Tavily and Brave as a first-class web-search provider inside the research agent.
SecuritySOC 2 Type I audit kicked off with Vanta.
SecurityMonthly security audit and hardening.
June 2026
Latte v2 shipped, and framework integrations moved into the base SDK.
ReleaseLatte v2 — the next-generation query-specific compression model. Up to 5x faster than latte v1.
NewDynamic compression ratio — latte v2 picks the ratio automatically per input, so you no longer have to tune it upfront.
NewStructured compression — JSON and Markdown inputs keep their structure through compression, so downstream parsers keep working.
NewResearch agent in the SDK — one call runs a multi-step web-search loop with per-snippet compression and a typed result, using Brave, Tavily, or AgentCore as the search provider.
NewLiteLLM integration in the SDK — shim-free CLI wrapper with per-role compression ratios and caching.
NewConfigurable HTTP transport for corporate proxies, custom CA bundles, and mTLS.
ImprovedThe Compresr Gateway project is on hold — we’re focusing on the SDK and model releases. Existing integrations continue to work; no new Gateway features are planned.
May 2026
Framework integrations shipped in the base SDK.
NewNative integrations with LangChain, LangGraph, and LlamaIndex — included in the base SDK.
NewDrop-in support for Anthropic-style and OpenAI-style calls in the SDK — use the client pattern you already know, and Compresr handles compression underneath.
FixedAuthentication flow fixes for edge cases around session refresh.
April 2026
TypeScript SDK went live and the SDK gained on-prem support.
ReleaseTypeScript SDK first public release (v1.1) with full feature parity to the Python SDK.
NewOn-prem SDK support — a base_url parameter lets customers point the SDK at self-hosted Compresr instances.
March 2026
Enterprise and User accounts, batch compression, and a faster API.
NewTwo account types introduced — Enterprise (postpaid, invoices, monthly budgets) and User (prepaid wallet, $10 free credit), each with their own key scopes, billing path, and dashboard tabs.
NewBatch compression API on the SDK (v2.1) — compress multiple documents in a single request.
ImprovedFaster API — reworked the backend to cut request latency across all endpoints.
ImprovedRedesigned dashboard with per-key usage analytics.
February 2026
Latte v1 replaced cmprsr-v1 and Python SDK v2.0 shipped.
ReleaseLatte v1 — a much faster and higher-quality query-specific compression model. cmprsr-v1 is deprecated in favor of latte v1.
ReleasePython SDK v2.0 — the compression client available on your own machine.
NewInteractive demo section on the landing page — try compression on your own text.
NewStripe integration for billing and payments.
ImprovedRefreshed landing page, docs, and pricing.
SecurityMonthly security audit and hardening.
January 2026
The platform came online and the first compression model shipped.
ReleaseCompresr platform first live — dashboard, account management, and billing UX.
Releasecmprsr-v1 released — the first compression model in production.
ReleasePython SDK first public release.
Subscribe to the Atom feed or follow @Compresr_yc_w26 for release updates.