vinzenz/hyperhive
vinzenz/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Revert "recv mcp tool: default max to a small batch of 5 (wire default stays 1)"

Browse source 
This reverts commit b8d608d4b6a7ed95faffdec8d8b3167ae8f3b39e.
This commit is contained in:
damocles 2026-07-02 22:09:51 +02:00 • committed by mara
commit ee45db6323
3 changed files with 23 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

42
hive-ag3nt/src/mcp.rs
View file

@ -611,12 +611,13 @@ pub struct RecvArgs {
#[serde(default)]
pub wait_seconds: Option<u64>,
/// Maximum number of messages to pop in this round-trip. Default
/// (None) is 5 — a small batch, so a burst of related messages
/// lands in one call without an extra round-trip. Pass `max: 1`
/// for strict single-message behaviour, or a higher value (capped
/// at 32 server-side) when the wake prompt mentions a bigger
/// pending count. Once the long-poll wakes up, the call drains up
/// to `max` in total before returning.
/// (None) is 1 (single-message behaviour — exactly what you want
/// when you're called to drive a turn off the first wake). Pass
/// a higher value (capped at 32 server-side) when you've been
/// told the inbox has more queued (the wake prompt mentions
/// pending count) and want to drain everything in one tool call.
/// Once the long-poll wakes up, the call drains up to `max` in
/// total before returning — no extra round-trip needed.
#[serde(default)]
pub max: Option<u32>,
}
@ -799,18 +800,17 @@ impl AgentServer {
#[tool(
description = "Pop messages from this agent's inbox. Returns one or more messages, or \
an empty marker if nothing is waiting. \n\n\
**Small-batch default**: with no args you get up to 5 messages — a burst of \
related messages lands in one call. Pass `max: 1` for strict single-message \
behaviour. Without `wait_seconds` (or with 0) the call returns immediately — a \
cheap 'anything pending?' peek. Pass a positive `wait_seconds` (capped at 180) to \
park the turn waiting for new work — incoming messages wake you instantly, \
otherwise the call returns empty at the timeout. That's strictly better than a \
fixed shell `sleep`. \n\n\
**Bigger drains**: pass `max: N` (capped at 32) when the wake prompt says the \
inbox has more queued than the default batch — one tool call beats N consecutive \
recvs. `wait_seconds` still applies to the FIRST message; once one arrives the \
call drains up to `max` in total. Empty result reported the same way regardless \
of `max`. \n\n\
**Single-message default**: with no args (or `max: 1`) you get the next message — \
same behaviour the harness uses to drive a turn. Without `wait_seconds` (or with 0) \
the call returns immediately — a cheap 'anything pending?' peek. Pass a positive \
`wait_seconds` (capped at 180) to park the turn waiting for new work — incoming \
messages wake you instantly, otherwise the call returns empty at the timeout. \
That's strictly better than a fixed shell `sleep`. \n\n\
**Batch drain**: pass `max: N` (capped at 32) to drain up to N messages in one \
round-trip. Use this when the wake prompt told you the inbox has more queued, or \
any time you expect a burst — one tool call beats N consecutive single recvs. \
`wait_seconds` still applies to the FIRST message; once one arrives the call drains \
up to `max` in total. Empty result reported the same way regardless of `max`. \n\n\
Typical pattern: when you have nothing else useful to do, call \
`recv(wait_seconds: 180)` to park until something arrives."
)]
@ -818,14 +818,10 @@ impl AgentServer {
let log = format!("{args:?}");
run_tool_envelope("recv", log, async move {
let waited = args.wait_seconds.is_some_and(|w| w > 0);
// Tool-layer default of a small batch (5). The wire default
// stays 1: the harness turn loop also sends `max: None` and
// consumes only the first message, so bumping the server-side
// default would silently drop the rest of a popped batch.
let (resp, retries) = self
.dispatch(hive_sh4re::Request::Recv {
wait_seconds: args.wait_seconds,
max: args.max.or(Some(5)),
max: args.max,
})
.await;
annotate_retries(format_recv(resp, waited), retries)