Watch
0
0
Fork
You've already forked hyperhive
0
hyperhive/swarm-controller/src
Repository files (latest commit first)
Filename Latest commit message Latest commit date
atlas 2c5eef73cd review classifier: model, effort and review concurrency as swarm-wide settings
The classifier ran every classification on a hard-coded `haiku` with no
`--effort`, and the controller ran at most two requested reviews at once,
also hard-coded, so changing any of them took a code change and a
redeploy.

swarm-controller now keeps one settings document,
`{model, effort, max_concurrent_reviews}`, in its own
`review-classifier-settings` KV bucket under `all`, the way it keeps the
schedule definitions. No document means today's values: haiku, no
`--effort`, two reviews. `GET /api/review-classifier/settings` reads it;
`PUT` replaces it whole, needs the gateway's `Remote-User` like the
try-out POST, and refuses a bad model (the inference route's check), an
effort outside the five levels or a limit outside 1..=8 with
problem+json 400.

Every run request (try-out and requested review) now carries the model
and effort, and the classifier passes them to claude instead of its
constant, which is gone. A try-out reads the settings when it is posted;
a review reads them for the limit before it waits for a place and again
for the model and effort once it has one. The review semaphore is
replaced by a running count under a resizable limit, so a PUT, or the
next review's read, changes how many run without a restart, and a lower
limit lets the runs already going finish. MAX_WAITING_REVIEWS is
unchanged.

Closes #5151, part of #5096.
2026-10-13 01:45:19 +02:00
..
forge swarm-controller: record the review classifier's runs and list them 2026-10-12 21:48:51 +02:00
matrix_account swarm-controller: revoke matrix admin from former operators 2026-10-05 12:39:02 +02:00
review_classifier review classifier: model, effort and review concurrency as swarm-wide settings 2026-10-13 01:45:19 +02:00
agent_icon.rs swarm: let an agent publish its own icon 2026-09-28 13:47:37 +02:00
agent_identity.rs swarm-secret-client: agents may list their own subtree; controller rewrites agent policies 2026-10-01 17:43:28 +02:00
agent_input.rs swarm UI: send a message to an agent, without its hive 2026-10-04 03:02:14 +02:00
agent_limits.rs swarm-controller: record the review classifier's runs and list them 2026-10-12 21:48:51 +02:00
agent_renewal.rs agents: re-pull the agent's own bao mTLS leaf on a timer 2026-10-11 10:00:43 +02:00
agent_state_stream.rs swarm-controller: end a dead replay, clamp stale resume ids, test on a real server 2026-10-11 23:26:24 +02:00
agent_status.rs swarm-ui: add agent start/stop, backed by the wanted-state route 2026-09-02 19:57:04 +02:00
auth.rs swarm options: no swarm.* value depends on a deploy option or this host's hive 2026-10-04 18:13:47 +02:00
config_pr.rs swarm-controller: bound forge calls in job executors 2026-10-09 21:00:59 +02:00
forge.rs swarm-ui, swarm-controller: the review try-out's repo and PR fields offer the repos and PRs to pick 2026-10-12 19:05:37 +02:00
forge_account.rs swarm-controller: record the review classifier's runs and list them 2026-10-12 21:48:51 +02:00
github_account.rs swarm-controller: record the review classifier's runs and list them 2026-10-12 21:48:51 +02:00
inference.rs review classifier: model, effort and review concurrency as swarm-wide settings 2026-10-13 01:45:19 +02:00
issue_report.rs swarm-ui, swarm-controller: the review try-out's repo and PR fields offer the repos and PRs to pick 2026-10-12 19:05:37 +02:00
linked_accounts.rs swarm-controller: record the review classifier's runs and list them 2026-10-12 21:48:51 +02:00
main.rs review classifier: model, effort and review concurrency as swarm-wide settings 2026-10-13 01:45:19 +02:00
matrix_account.rs swarm-controller: record the review classifier's runs and list them 2026-10-12 21:48:51 +02:00
otel_http_client.rs swarm-queue-client: request the bearer-authz scope when minting an agent token 2026-09-17 09:51:44 +02:00
pulls.rs swarm-ui, swarm-controller: the review try-out's repo and PR fields offer the repos and PRs to pick 2026-10-12 19:05:37 +02:00
queue_identity.rs swarm-controller gains forge access when the token arrives, without a restart 2026-10-08 13:40:24 +02:00
read_policy.rs swarm-secret-client: agents may list their own subtree; controller rewrites agent policies 2026-10-01 17:43:28 +02:00
review_classifier.rs review classifier: model, effort and review concurrency as swarm-wide settings 2026-10-13 01:45:19 +02:00
schedules.rs swarm: per-agent inference settings (model, effort, subagent_models) in KV 2026-10-09 23:13:13 +02:00
status.rs swarm-controller: make status::render/row pub(crate) so agent_status.rs's doc links resolve 2026-09-02 10:20:29 +02:00
store.rs fix doc-comment pointers broken by the module-eval split 2026-09-20 04:31:08 +02:00
subagent_term.rs swarm-controller: end a dead replay, clamp stale resume ids, test on a real server 2026-10-11 23:26:24 +02:00
term_stream.rs swarm-controller: end a dead replay, clamp stale resume ids, test on a real server 2026-10-11 23:26:24 +02:00
test_nats.rs swarm-controller: end a dead replay, clamp stale resume ids, test on a real server 2026-10-11 23:26:24 +02:00
vcs_metrics.rs swarm CA: describe the store's pki root as the only anchor in docs and comments 2026-10-05 21:17:10 +02:00
wanted.rs swarm-controller: first credential-renewal pass waits for the queue connection 2026-09-30 15:46:56 +02:00
webhook.rs swarm-controller: post the review classifier's verdict on a review request 2026-10-12 16:11:50 +02:00