Clamp session DB pagination limits for search and listing APIs

[aydnOktay]

hardens hermes_state.py by clamping limit and offset for search_messages, search_sessions, and list_sessions_rich so invalid or excessive pagination values cannot overload SQLite FTS/list queries, and adds focused regression coverage in tests/test_hermes_state.py for helper clamping behavior and end-to-end session search pagination bounds.

[GottZ]

This was generated by AI during triage.

Summary

Two open PRs address the same pagination-input problem by adding a shared clamp to three SessionDB methods, but neither has a recorded pipeline verify verdict. #32564 also removes unrelated query behavior, while #35308 preserves it but still applies the clamp at the wrong abstraction boundary and would truncate legitimate full-history callers.

Related pull requests

• #32564 related — (+60/-32) — supersede: The diff clamps search_messages, search_sessions, and list_sessions_rich, but it also deletes the existing message-sort normalization/ORDER BY construction and the enriched last_active session-query construction, changes unrelated to the reported pagination cause and potentially breaking their downstream use.
• #35308 duplicate — (+60/-0) — keep open for revision: This is the cleaner duplicate because it adds the same clamp without deleting adjacent query logic. The visible keep_open review on #35308 should be followed, but its blocking findings must be resolved before merge: clamping inside SessionDB changes the established LIMIT -1 unlimited contract and truncates export/cleanup callers requesting 100000 or 10000 rows, while the tests cover neither list_sessions_rich() nor datasets exceeding 200 rows.

Duplicates

#32564 and #35308 implement essentially the same shared SessionDB pagination clamp across the same three methods; #35308 is the cleaner current-main implementation despite the earlier contributor comment identifying it as a duplicate of #32564.

Suggested consolidation

Revise, then merge #35308 — move pagination validation to the untrusted HTTP/UI boundaries identified by the keep_open review, preserve SessionDB's unlimited and high-limit internal contracts, and add boundary plus full-store regression coverage. Despite the contributor duplicate comment favoring the earlier #32564, the actual #32564 diff removes unrelated sorting and enriched-session query construction; once the corrected #35308 is ready, close #32564 as superseded.

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
    subgraph Dup32564 ["PRs duplicating each other"]
        P32564["PR #32564 (open)"]
        P35308["PR #35308 (open)"]
    end
    class P32564 open
    class P35308 open
    class P32564 target
    click P32564 "https://github.com/NousResearch/hermes-agent/pull/32564"
    click P35308 "https://github.com/NousResearch/hermes-agent/pull/35308"

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 or no verify verdict yet (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 10 kB of PR diffs, 2 kB of issue/PR text, 1 kB of discussion (2 comments), 1 verify verdict. verdicts reflect diff content, not PR titles. Part of an automated triage batch.

[aydnOktay]

Closing to concentrate review on a single linked product bug: #89979.

[aydnOktay]

Closing to concentrate review on a single linked product bug: #89979.