fix(cli): add timeout to state.db health probe in hermes doctor (#72441)

[JonthanaHanh]

Summary

Fixes #72441.

hermes doctor runs _db_opens_cleanly() which executes PRAGMA integrity_check on the state.db database. On profiles with 2000+ sessions, this can take minutes, making the doctor command appear to hang indefinitely.

Changes

• hermes_cli/doctor.py: Wrap _db_opens_cleanly in a ThreadPoolExecutor with a 30-second timeout. On timeout, emit a warning with actionable suggestions.

Testing

• py_compile: OK
• ruff check: All checks passed
• The timeout is generous (30s) to avoid false positives on slow but healthy databases.

[isak-ialogics]

The 30s timeout currently does not bound hermes doctor: after _fut.result(timeout=30) raises, execution leaves the with ThreadPoolExecutor(...) block, whose implicit shutdown(wait=True) waits for _db_opens_cleanly() to finish before the warning can be observed. So the reported multi-minute hang is preserved (then a timeout warning appears afterward). A focused regression can use a blocking probe and assert elapsed time; the implementation needs a truly cancellable boundary (for example a subprocess that is terminated on timeout, or SQLite-level interruption/progress handling), since a running Python thread cannot be cancelled reliably.

[rdxhemadri]

@alt-glitch @teknium1 merge this pr,

[rdxhemadri]

hey @OutThisLife if you can pls merge this pr
@izumi0uu @helix4u @teknium1 @knoal @OutThisLife @benbarclay @alt-glitch

[rdxhemadri]

some one pls merge this pr @kshitijk4poor @OutThisLife
@izumi0uu @helix4u @teknium1 @knoal @benbarclay @alt-glitch

[rdxhemadri]

some one pls merge this pr @kshitijk4poor @OutThisLife
@izumi0uu @helix4u @teknium1 @knoal @benbarclay @alt-glitch
sry for spamming

[teknium1]

Thanks for targeting the state.db health probe; the current main path still invokes it immediately after the session-count row (hermes_cli/doctor.py:1501-1510), and _db_opens_cleanly() runs PRAGMA integrity_check (hermes_state.py:1091-1113).

Problems

• hermes_cli/doctor.py:1505 in bf49560cc0eb uses a with ThreadPoolExecutor(...) block. Exiting that block waits for the running worker, so a Future.result(timeout=30) exception does not bound the command. The repository documents this exact behavior at tools/skills_hub.py:4078-4085.
• bf49560cc0eb changes only hermes_cli/doctor.py; it does not add a regression test showing that doctor returns while the probe remains blocked.

Suggested changes

• Use a genuinely cancellable SQLite-level deadline, such as a progress handler that interrupts the integrity scan, rather than a thread timeout alone.
• Add an elapsed-time regression test for a blocking/over-deadline probe.

Automated hermes-sweeper review.

[teknium1]

This context manager calls shutdown(wait=True) on exit. After result(timeout=...) raises, it still waits for _db_opens_cleanly() to finish, so the command remains blocked and the timeout warning is not observable at 30 seconds. Use a genuinely cancellable SQLite boundary and add an elapsed-time regression.

[GottZ]

This was generated by AI during triage.

Summary

Two PRs address #72441 by bounding the state.db health probe that follows the session-count check. #72527 adds a thread-pool timeout that still waits for the worker on context-manager shutdown, while #76003 replaces that ineffective boundary with SQLite-level cancellation, an abandonable fallback, and elapsed-time and repair-safety regressions.

Related pull requests

• fix(cli): add timeout to state.db health probe in hermes doctor (#72441) #72527 best fix — (+20/-1) — superseded: The diff correctly identifies _db_opens_cleanly() as the source of the apparent hang, but with ThreadPoolExecutor(...) invokes shutdown(wait=True) after Future.result(timeout=30) times out, so doctor remains blocked. Despite the keep_open review on fix(cli): add timeout to state.db health probe in hermes doctor (#72441) #72527, the diff in fix(cli): bound the doctor state.db health probe with a cancellable SQLite deadline (supersedes #72527) #76003 implements the requested cancellable boundary and regression coverage, leaving fix(cli): add timeout to state.db health probe in hermes doctor (#72441) #72527 no distinct salvageable fix.
• fix(cli): bound the doctor state.db health probe with a cancellable SQLite deadline (supersedes #72527) #76003 best fix — (+447/-12) — keep open with a salvage path: The diff adds a SQLite progress-handler deadline, an abandonable daemon-worker fallback for uninterruptible I/O, explicit timeout reporting that is not treated as corruption, and six focused tests covering elapsed-time bounding and prevention of repair on timeout. It directly addresses the contributor review on fix(cli): add timeout to state.db health probe in hermes doctor (#72441) #72527 and preserves the original unbounded behavior for callers that do not request a deadline.

Duplicates

#72527 and #76003 target the same #72441 failure path; #76003 supersedes #72527 with a boundary that actually returns on timeout and with regression coverage.

Suggested consolidation

Keep #76003 open with the concrete salvage path already present in its diff: retain the cancellable SQLite deadline, abandonable fallback, non-corruption timeout semantics, and focused regression tests while completing normal review. Close #72527 as a duplicate of #76003; despite the keep_open review on #72527, its context-managed executor still waits for the blocked probe, whereas #76003 implements that review's requested SQLite-level cancellation and elapsed-time regression.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    I72441(["issue #72441 (open)"])
    subgraph Dup72527 ["PRs duplicating each other"]
        P72527["PR #72527 (open)"]
        P76003["PR #76003 (open)"]
    end
    P72527 -->|best fix| I72441
    class I72441 open
    class P72527 open
    class P76003 open
    class P72527 best
    class P76003 best
    class P72527 target
    click I72441 "https://github.com/NousResearch/hermes-agent/issues/72441"
    click P72527 "https://github.com/NousResearch/hermes-agent/pull/72527"
    click P76003 "https://github.com/NousResearch/hermes-agent/pull/76003"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 1 issue in this complex. Each diff was read against this issue; Assessment working set: 23 kB of PR diffs, 19 kB of issue/PR text, 8 kB of discussion (13 comments), 4 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.