Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
Commits on main
fix(proxy): share one compression deadline across a Responses request (#3938)
4d27d02 gglucass committed 1 month ago
fix(proxy): bill the whole prompt on the gateway path so savings read true (#3632)
befdb52 Tejas Chopra committed 1 month ago
perf(proxy): lighter request path: alias the message snapshot, keep the token cache on re-counts, decompress off the event loop (#3909)
c873d13 Bill committed 1 month ago
fix(cache): keep prefix lineage across a replaced system tail (#3934)
16eaec1 gglucass committed 1 month ago
fix(ccr): only proactively expand compressions present in the requesting conversation (#3924)
9b9a883 Parideboy committed 1 month ago
fix(proxy): take a streamed Anthropic turn's uncached input from usage (#3946)
7f608ea gglucass committed 1 month ago
perf(image): OCR and SigLIP each image once, not once per turn (#3941)
a8561fb gglucass committed 1 month ago
fix(proxy): let an explicit compress-user-messages off beat the profile (#3936)
822c886 gglucass committed 1 month ago
fix(proxy): quarantine compression only once half the pool is stuck (#3932)
be2b205 gglucass committed 1 month ago
fix(proxy): do not cache error replies delivered as http 200 (#3930)
1132a54 gglucass committed 1 month ago
feat(learn): add agy (Antigravity CLI) as an analysis backend (#3939)
4ba231a Trîggered committed 1 month ago
fix(learn): drop the echoed prompt from a failed cli's error (#3928)
f00d425 gglucass committed 1 month ago
fix(learn): run the claude-cli analysis with hooks disabled (#3926)
2a4d34c gglucass committed 1 month ago
fix(backends/litellm): keep the upstream 4xx status on send_message errors (#3944)
e0e41cd Younes Beriane committed 1 month ago
fix(parser): stop counting HTML comments twice in waste signals (#3942)
d0e9c4c Zhibo Zhang (Amur) committed 1 month ago
ci: move Apple Silicon jobs off the retiring macos-14 runner (#3953)
0d841fe gglucass committed 1 month ago
fix(proxy): run injected memory tools server-side on streaming turns (#3947)
1cf4966 Kenneth Wong committed 1 month ago
fix(proxy): inject memory context past a trailing system message (#3948)
3948dbe Kenneth Wong committed 1 month ago
deps: bump the npm-minor-patch group across 2 directories with 11 updates (#3903)
afe4f4e dependabot[bot] committed 1 month ago
feat(dashboard): add CO₂ Saved card to dashboard and /stats API (#1369)
2bc9419 Peter Lodri committed 1 month ago
feat(wrap): extend reduce-at-source quiet defaults to telemetry/nag banners (#2550)
3a3a465 Abhay Singh committed 1 month ago
fix(proxy): log response_content_length in proxy_inbound_response (#2701)
cfa479a im47cn committed 1 month ago
fix(install/apply): add --no-rate-limit flag, persist in proxy_args (#1350) (#1365)
31d0344 Peter Lodri committed 1 month ago