Block a user
/v1/models: re-expose
max_model_len (+ input/output) — Hermes Agent can't read context window from limit.context
feat(cortex): re-expose flat max_model_len / max_input_tokens / max_output_tokens on /v1/models (#78)
[epic] Multi-tenant governance: auth, federated token budgeting, concurrency control
Priority housekeeping: Stages 1–3 (identity, budget enforcement, admission control, fair-share, load-aware routing, fail-fast pre-validation) shipped and have been live since 3b60dd7 — the A0…
Evaluate speculative decoding for neuron (EAGLE-3 first, bench-gated) — batch-1 decode latency
Evaluate speculative decoding for neuron (EAGLE-3 first, bench-gated) — batch-1 decode latency
Closing as superseded by #96 (F4b in epic #84), which now carries the wire-up of speculative decoding into generation. The evaluation criteria captured here — EAGLE-3 self-speculative first,…
perf(neuron): speculative decoding with a small same-family drafter
perf(neuron): speculative decoding with a small same-family drafter
Closing as superseded by #96 (F4b — wire speculative decoding into generation, part of epic #84). The core speculative machinery already exists in harness/speculative.rs; #96 tracks connecting…
agent: heartbeat/event client accepts expired controller TLS cert
feat(cortex): re-expose flat max_model_len / max_input_tokens / max_output_tokens on /v1/models (#78)