Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
Commits on main
test: require a catalog sync to fill a NULL Codex display_name (#731)
0998e99 Alfredo Lopez committed 1 month ago
fix: read the Codex catalog display_name in normalizeModels (#731)
96212fb Alfredo Lopez committed 1 month ago
test: require the recorded Codex display_name as displayName (#731)
fef59ca Alfredo Lopez committed 1 month ago
fix: mark a rejected Codex refresh reauth_required (#753)
d5729ca Alfredo Lopez committed 1 month ago
test: require a rejected Codex refresh to mark the row reauth_required (#753)
3412896 Alfredo Lopez committed 1 month ago
feat(llm-proxy): log request-limit reasons and check the deadline before redeem
3e9b9ce Alfredo Lopez committed 1 month ago
docs: document the 8 MiB request cap and subscription context windows (#731)
ad29fb6 Alfredo Lopez committed 1 month ago
feat(llm-proxy): raise the per-attempt stream cap to 30 min and bound queue wait
6cd077f Alfredo Lopez committed 1 month ago
feat: carry the catalog context window to subscription tasks (#731)
edf6e08 Alfredo Lopez committed 1 month ago
test: require the resolved window on the manual compact path (#731)
a2b899e Alfredo Lopez committed 1 month ago
test: require the catalog context window end to end (#731)
1174a7b Alfredo Lopez committed 1 month ago
feat(llm-proxy): keep silent attempts open with SSE keepalive comments
5aaf89a Alfredo Lopez committed 1 month ago
docs(workflow-recipes): document applyGate read-failure and GFS spec compare
eeac475 Alfredo Lopez committed 1 month ago
fix: admit a cap-sized request on the LLM authorize route (#731)
59ce5d3 Alfredo Lopez committed 1 month ago
test: require an envelope allowance on the LLM authorize route (#731)
866bac7 Alfredo Lopez committed 1 month ago
fix: bound request-body memory in both LLM proxies before parsing (#731)
774e0ae Alfredo Lopez committed 1 month ago
Merge remote-tracking branch 'origin/dev' into fix/wrc-requeue-backoff-and-read-first-apply
d3e7e7a Alfredo Lopez committed 1 month ago
test: require memory-bounded body admission in both LLM proxies (#731)
9192026 Alfredo Lopez committed 1 month ago
feat(llm): surface the stream duration cap as a terminal error code
cba325f Alfredo Lopez committed 1 month ago
fix: raise the non-image request cap to 8 MiB in both contracts (#731)
0d39eec Alfredo Lopez committed 1 month ago
test: require an 8 MiB request cap derived from the contracts (#731)
2d12a19 Alfredo Lopez committed 1 month ago
docs: correct #731 comments on bounds, tokenizer scope and refusal labels
0333ddf Alfredo Lopez committed 1 month ago
fix(llm-proxy): bound upstream streams by a typed idle timeout and total cap
0ec093c Alfredo Lopez committed 1 month ago
fix(mcp-host): restore the ContextManager log component and drop the conversation id (#731)
adfd72e Alfredo Lopez committed 1 month ago
test(mcp-host): require the ContextManager log component and pin refusal labels (#731)
772d475 Alfredo Lopez committed 1 month ago
fix: classify proxy 413 as context length and derive the body limit (#731)
2b04bd6 Alfredo Lopez committed 1 month ago
test: require a size-refusal 413 and a proxy body limit above the contract cap (#731)
4f2b7d7 Alfredo Lopez committed 1 month ago