Wire wake and identity into crossbar; v2 smoke and README
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
This commit is contained in:
@@ -5,6 +5,7 @@ poll_interval = "1s" # 60s in production; 1s makes the smoke run
|
||||
lease_idle = "30m" # a conversation idle this long loses its host
|
||||
retention = "180d" # per-request rows older than this are rolled up daily
|
||||
queue_max = 1 # waiting places per (host, model) beyond `parallel`; 503 past that
|
||||
identity = "off" # "tailscale" gates routes with `peers` by `tailscale whois`; "header" trusts X-Crossbar-Peer (TEST ONLY)
|
||||
|
||||
[hosts.alpha]
|
||||
base_url = "http://127.0.0.1:18081" # e.g. http://straylight.<tailnet>:11434
|
||||
@@ -15,6 +16,10 @@ models = { "ornith-1.5-35b-a3b" = { parallel = 1 }, "small-9b" = { parallel =
|
||||
base_url = "http://127.0.0.1:18082" # e.g. http://titan.<tailnet>:8081
|
||||
weight = 2.0
|
||||
models = { "ornith-1.5-35b-a3b" = { parallel = 2 } }
|
||||
[hosts.beta.wake] # v2: wake a sleeping host when nothing else can take a new lease
|
||||
mac = "aa:bb:cc:dd:ee:02"
|
||||
broadcast = "127.0.0.1:19082" # the LAN broadcast address, port 9, in production
|
||||
wait = "20s"
|
||||
|
||||
# v1: a route is a set of candidate hosts; each conversation gets a sticky lease on the host with
|
||||
# the most free slots × weight at the time it starts. Pins and drains come from the admin API.
|
||||
@@ -24,3 +29,4 @@ default_model = "ornith-1.5-35b-a3b"
|
||||
|
||||
[routes.hermes-x]
|
||||
hosts = ["beta", "alpha"]
|
||||
# peers = ["talos"] # v2: with identity = "tailscale", only these tailnet nodes may use the route
|
||||
|
||||
Reference in New Issue
Block a user