Add adaptive backoff to the daemon scheduler by maxkle1nz · Pull Request #68 · maxkle1nz/m1nd

maxkle1nz · 2026-04-05T20:41:45Z

Summary

add idle-streak tracking and adaptive poll backoff to the daemon scheduler
expose effective_poll_interval_ms and idle_streak in daemon_status
keep next_tick_due_ms / overdue_ms aligned with the effective scheduler interval

Validation

cargo fmt --check
cargo check -p m1nd-mcp -p m1nd-ingest
cargo test -p m1nd-mcp daemon_ -- --nocapture
cargo clippy -p m1nd-mcp -p m1nd-ingest -- -D warnings
MCP smoke for effective_poll_interval_ms / idle_streak / next_tick_due_ms / overdue_ms

Why this matters

This makes the daemon scheduler materially smarter before native watcher dependencies arrive. The daemon now backs off when the workspace is quiet, while still exposing exactly when it plans to wake up next.

The daemon scheduler now reports when the next tick is due and whether it is currently overdue, and the server loop uses remaining-time scheduling instead of sleeping the full poll interval every time. Constraint: Improve scheduler visibility and efficiency without introducing new watcher dependencies Rejected: Waiting for notify/watchman before adding scheduler timing signals | left the current polling scheduler opaque and less efficient than necessary Rejected: A new scheduler tool | the timing state belongs in daemon_status and the existing server loop Confidence: high Scope-risk: narrow Reversibility: clean Directive: Keep next-tick timing derived from persisted daemon state so future watcher backends can share the same status contract Tested: cargo fmt --check; cargo check -p m1nd-mcp -p m1nd-ingest; cargo test -p m1nd-mcp daemon_ -- --nocapture; cargo clippy -p m1nd-mcp -p m1nd-ingest -- -D warnings; MCP smoke for next_tick_due_ms / overdue_ms Not-tested: Interaction with future native watcher backends and long-running background scheduler drift over many hours

The daemon scheduler now tracks idle streaks, scales its effective poll interval under quiet conditions, and reports that effective interval through daemon_status. This reduces wasted polling while keeping the daemon fully observable through the existing status contract. Constraint: Improve daemon efficiency without adding watcher dependencies or changing the existing daemon tool surface Rejected: Fixed-interval polling until native watchers land | wastes cycles and hides useful scheduler state during the interim phase Rejected: A separate scheduler configuration tool | unnecessary surface growth before watcher backends exist Confidence: high Scope-risk: narrow Reversibility: clean Directive: Keep scheduler backoff state derived from daemon runtime state so native watcher backends can eventually bypass polling without breaking status semantics Tested: cargo fmt --check; cargo check -p m1nd-mcp -p m1nd-ingest; cargo test -p m1nd-mcp daemon_ -- --nocapture; cargo clippy -p m1nd-mcp -p m1nd-ingest -- -D warnings; MCP smoke for effective_poll_interval_ms / idle_streak / next_tick_due_ms / overdue_ms Not-tested: Very long idle periods across multiple days and interaction with future watcher-native backends

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 28d7059167

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-04-05T20:45:44Z

+    pub idle_streak: u32,
+    pub max_backoff_multiplier: u32,


Add defaults for new daemon-state fields

The new idle_streak and max_backoff_multiplier fields are required during deserialization, so pre-existing daemon_state.json files (written before this commit) will fail to parse in load_daemon_state and silently fall back to DaemonRuntimeState::default(). On upgrade, that drops persisted daemon runtime state (including active status, watch paths, tracked files, and counters), which breaks cross-session daemon persistence.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-04-05T20:45:44Z

+        state
+            .daemon_state
+            .last_tick_ms
+            .map(|last| last.saturating_add(state.daemon_state.poll_interval_ms))


Base daemon_status due time on effective interval

next_tick_due_ms is computed with last_tick_ms + poll_interval_ms, but the scheduler now waits on a backoff-expanded interval. When idle_streak > 0, daemon_status reports an earlier due time and premature overdue_ms, so monitoring/automation that trusts these fields will see false overdue conditions.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-04-05T20:45:44Z

+            } else {
+                effective_poll_interval_ms
+                    .saturating_sub(elapsed)
+                    .clamp(25, 1000)


Remove 1s cap from adaptive daemon wait

Capping the remaining wait with .clamp(25, 1000) truncates backoff intervals above 1 second. After a few idle ticks (for example, 200ms base with idle_streak >= 3), the loop wakes every second and background_tick_if_due can tick immediately because the base poll interval has elapsed, so adaptive backoff never reaches its intended multi-second range.

Useful? React with 👍 / 👎.

Copilot

Pull request overview

This PR adds idle-streak tracking and an adaptive (exponential) backoff mechanism to the daemon scheduler, and exposes additional scheduling telemetry (effective_poll_interval_ms, idle_streak, plus next/overdue tick timing) via daemon_status.

Changes:

Extend DaemonRuntimeState with idle_streak and max_backoff_multiplier for adaptive scheduling.
Add daemon_wait_duration_ms() and use it to drive the server loop’s recv_timeout wait duration.
Enhance daemon_status output and update tick logic to increment/reset idle_streak based on whether work happened.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 6 comments.

File	Description
`m1nd-mcp/src/session.rs`	Adds persisted daemon runtime fields needed for backoff state tracking.
`m1nd-mcp/src/server.rs`	Introduces a helper to compute adaptive wait duration and applies it to the main server loop; adds tests.
`m1nd-mcp/src/daemon_handlers.rs`	Initializes/updates idle streak, and exposes scheduling/backoff fields in `daemon_status`.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-05T20:46:07Z

    pub last_tick_changed_files: usize,
    pub last_tick_deleted_files: usize,
    pub last_tick_alerts_emitted: usize,
+    pub idle_streak: u32,


DaemonRuntimeState is persisted to disk; adding non-optional fields without #[serde(default)] will make older daemon_state.json files fail to deserialize, causing load_daemon_state() to fall back to a full default state (potentially disabling an active daemon and losing counters). Add #[serde(default)] (or #[serde(default = "...")]) for the new fields to keep backward compatibility.

Suggested change

pub idle_streak: u32,

pub idle_streak: u32,

#[serde(default)]

Copilot · 2026-04-05T20:46:07Z

+    let effective_poll_interval_ms = state.daemon_state.poll_interval_ms.saturating_mul(
+        2u64.pow(
+            state
+                .daemon_state
+                .idle_streak
+                .min(state.daemon_state.max_backoff_multiplier.saturating_sub(1)),
+        ),
+    );


Using 2u64.pow(exponent) risks overflow/wrap (and can panic in debug) if idle_streak/max_backoff_multiplier allow exponent >= 64. Prefer u64::saturating_pow, checked_pow with a fallback, or cap exponent to a safe maximum (e.g. 0..=63) before exponentiation.

Suggested change

let effective_poll_interval_ms = state.daemon_state.poll_interval_ms.saturating_mul(

2u64.pow(

state

.daemon_state

.idle_streak

.min(state.daemon_state.max_backoff_multiplier.saturating_sub(1)),

),

);

let backoff_exponent = state

.daemon_state

.idle_streak

.min(state.daemon_state.max_backoff_multiplier.saturating_sub(1))

.min(63);

let effective_poll_interval_ms = state

.daemon_state

.poll_interval_ms

.saturating_mul(2u64.pow(backoff_exponent));

Copilot · 2026-04-05T20:46:07Z

+        "idle_streak": state.daemon_state.idle_streak,
+        "max_backoff_multiplier": state.daemon_state.max_backoff_multiplier,


max_backoff_multiplier is used as an exponent cap (2^exponent), so a value of 8 actually allows up to 128× backoff. Either rename this field to reflect exponent semantics (e.g. max_backoff_exponent) or change the math to cap the multiplier itself (e.g. multiplier = min(2^idle_streak, max_backoff_multiplier)).

Copilot · 2026-04-05T20:46:08Z

+                effective_poll_interval_ms
+                    .saturating_sub(elapsed)
+                    .clamp(25, 1000)
+            }


The remaining-wait clamp to <= 1000ms prevents adaptive backoff from ever sleeping longer than 1s; once the effective interval exceeds 1s, the loop will wake every second and background_tick_if_due() will see the base poll_interval_ms as due, effectively negating the backoff. Consider allowing the remainder up to the effective interval (or the same 10_000ms cap) and ensure the tick-due checks use the effective interval as well.

Copilot · 2026-04-05T20:46:08Z

+    let effective_poll_interval_ms = state
+        .daemon_state
+        .poll_interval_ms
+        .saturating_mul(2u64.pow(exponent))


Same overflow risk here: 2u64.pow(exponent) can overflow/wrap (and panic in debug) if exponent grows. Use a saturating/checked exponentiation approach or cap exponent before computing the multiplier.

Suggested change

let effective_poll_interval_ms = state

.daemon_state

.poll_interval_ms

.saturating_mul(2u64.pow(exponent))

let backoff_multiplier = 1u64.checked_shl(exponent).unwrap_or(u64::MAX);

let effective_poll_interval_ms = state

.daemon_state

.poll_interval_ms

.saturating_mul(backoff_multiplier)

Copilot · 2026-04-05T20:46:08Z

+    let now = now_ms();
+    let next_tick_due_ms = if state.daemon_state.active && state.daemon_state.poll_interval_ms > 0 {
+        state
+            .daemon_state
+            .last_tick_ms
+            .map(|last| last.saturating_add(state.daemon_state.poll_interval_ms))
+    } else {
+        None
+    };
+    let overdue_ms = next_tick_due_ms.map(|due| now.saturating_sub(due));
+    let effective_poll_interval_ms = state.daemon_state.poll_interval_ms.saturating_mul(
+        2u64.pow(
+            state
+                .daemon_state


next_tick_due_ms/overdue_ms are computed using the base poll_interval_ms, but the scheduler uses an adaptive effective interval (idle backoff + clamping). This will make daemon_status diverge from actual scheduling; compute these based on the same effective interval logic used by the scheduler (and apply the same clamps).

max kle1nz added 2 commits April 5, 2026 22:27

Copilot AI review requested due to automatic review settings April 5, 2026 20:41

maxkle1nz enabled auto-merge (squash) April 5, 2026 20:42

Copilot started reviewing on behalf of maxkle1nz April 5, 2026 20:42 View session

chatgpt-codex-connector Bot reviewed Apr 5, 2026

View reviewed changes

Copilot AI reviewed Apr 5, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add adaptive backoff to the daemon scheduler#68

Add adaptive backoff to the daemon scheduler#68
maxkle1nz wants to merge 2 commits into
mainfrom
codex/m1nd-daemon-scheduler-v1

maxkle1nz commented Apr 5, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 5, 2026

Uh oh!

chatgpt-codex-connector Bot Apr 5, 2026

Uh oh!

chatgpt-codex-connector Bot Apr 5, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 5, 2026

Uh oh!

Copilot AI Apr 5, 2026

Uh oh!

Copilot AI Apr 5, 2026

Uh oh!

Copilot AI Apr 5, 2026

Uh oh!

Copilot AI Apr 5, 2026

Uh oh!

Copilot AI Apr 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		"idle_streak": state.daemon_state.idle_streak,
		"max_backoff_multiplier": state.daemon_state.max_backoff_multiplier,

Conversation

maxkle1nz commented Apr 5, 2026

Summary

Validation

Why this matters

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 5, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants