Conversation
…columns Three improvements to session browsing and resume: 1. /sessions search <query> Search session titles/IDs inline using existing query_session_listing infrastructure. Previously the CLI handler never branched on 'search' even though the parser and gateway already supported it. 2. /resume --limit N Show more than 10 sessions. /resume --limit 30 shows 30, /resume without --limit still defaults to 10. Stores _pending_resume_limit so the bare-number flow re-fetches the same count. 3. Richer listing columns (Model + Msgs) The inline /resume and /sessions table now shows: - Model (abbreviated, provider prefix stripped, max 12 chars) - Msgs (total message count, right-aligned) Also: query_session_listing gains an offset parameter for future pagination work (e.g. /resume --more). Closes the loop: users no longer need to start a new conversation to search for past sessions.
…h cards
Previously /sessions search used LIKE title/id matching which only
found sessions whose title happened to contain the query term. Now
it uses SessionDB.search_messages() — the same FTS5 engine the AI's
session_search tool uses — to search actual conversation content.
Display format changed from compact table to per-session cards:
# Date Source Model Title
FTS5 snippet with matched context...
This shows date, source, model, title, an 80-char match snippet,
and message count per session — much closer to what the AI produces
with session_search.
Filters: user/assistant roles only, excludes tool/subagent/cron
sources, deduplicates by session_id, skips current session.
Updates the CLI-level (non-interactive) hermes sessions list rendering to match the interactive /sessions format with Model (12-char, provider prefix stripped) and Msgs (message count) columns. All 4 rendering paths updated: - with titles + workspace - without titles + workspace - with titles, no workspace - without titles, no workspace
Adds a search subcommand to the non-interactive 'hermes sessions' CLI that mirrors /sessions search inside the interactive session. Uses FTS5 content search via SessionDB.search_messages() — searches actual message text, not just session titles. Usage: hermes sessions search <query> Output: rich session cards with date, source, model, title, and an 80-char FTS5 match snippet grouped by session (up to 10 results). Allows the AI agent to discover relevant sessions programmatically.
…nt('')
_cprint() signature is def _cprint(text: str) — calling it with no
argument raises TypeError, silently killing the /sessions search
rendering loop after it finds results.
…dren, card format Three fixes: 1. Show session ID (@session:default/<id>) in each result so users can /resume <id> directly from search output. Format matches the AI's session_search output style. 2. Deduplicate compression children via parent_session_id chain walk: if a parent session and its compressed child (NousResearch#2, NousResearch#3, etc.) both match, keep only the latest descendant. 3. Remove diagnostic debug prints that were added to find the _cprint() TypeError bug.
…ession dedup Back to the native /resume table style the user wanted: # Title Model Msgs Preview Last Active ID ── ───────────────────────────── ──────────── ──── ────────────────────────────── ──────────── ──────────────────────── 1 Fetching X impressions in 30… deepseek-v4… 133 Right, let me check what pla… Jul 04 20260704_210455_b9279f Changes vs the first table version: - ID column always shown (was missing before) - Compression children deduplicated via parent_session_id chain walk - Preview uses FTS5 snippet (actual matched text) instead of first-user-message Both interactive (/sessions search) and CLI (hermes sessions search) updated.
… Source column
Changes:
- Remove compression lineage dedup (was too aggressive — removing
valid sessions that happened to share a parent_session_id)
- Preview now uses session.preview (first user message) instead of
raw FTS5 snippet (was displaying JSON/tool-call noise)
- Add 'Source' column (cli, telegram, etc.) matching AI output
- Fix bug: header string wasn't an f-string, showing literal
'{search_query}' instead of the actual query
get_session() doesn't return the 'preview' field — it's computed by list_sessions_rich via a subquery on the first user message. Added a batch SQL query to fetch previews before the render loop so all 10 results show meaningful first-user-message previews instead of blanks.
Re-add parent_session_id chain walk to deduplicate compression children. If A → A NousResearch#2 → A NousResearch#3 all match the search, keep only the latest descendant (A NousResearch#3). This removes duplicate entries for the same logical conversation.
… U/A counts, full title Improvements: - Sort results by last_active descending (most recent first) - Preview from root ancestor's first user message (not compressed child) - U/A columns showing user/assistant message counts - Created date (YYYY-MM-DD) column - Full title with dynamic width based on terminal size - Deduplicate compression children - Preview line below each result row
Column layout: #, Title, Model, Tok (input/output), Created, Last, Preview, ID - Tok column shows human-readable input/output tokens (e.g. 220k/16k) - Model column back - Preview column back inline (root ancestor's first message, 20 chars) - Last Active column shortened to compact format (49m, 26d) - Removed U/A and CWD columns
…ped at 50 Removed terminal-width-based sizing. Title column now sizes to fit the longest title in the result set (min 16, max 50). Clean table.
…mports Code review of sessions-resume-improvements found: CRITICAL: interactive /sessions search printed duplicate headers — the '⚙️ /sessions search' block AND the table column header block each executed twice (leftover from iterative patching; widths even differed 40/24 vs 20/12). CLI version was clean, which is why the bug was invisible in 'hermes sessions search' tests. MEDIUM: - Duplicate unreachable 'if not seen' blocks in both handlers - Unused: sid_order dict, role_counts GROUP BY query (ran a full query and discarded it), import re as _re, query_session_listing import, import os as _os (module-level os exists) Verified: both files compile; single header + single table header in both paths; CLI search output unchanged and correct.
/sessions, /sessions search, 'hermes sessions list' and 'hermes sessions search' now render the identical table: # Title Model Tok Created Last Preview ID. Extracted render_sessions_table() into hermes_cli/session_listing.py — the single source of truth for the table (previously the layout logic was duplicated ~3x and had already drifted apart: /sessions showed Msgs/Last Active columns, search showed Tok/Created). - render_sessions_table resolves previews: explicit lookup (search precomputes root-ancestor previews) > parent-chain walk to root ancestor > row preview. Walk is capped at 20 hops. - _show_recent_sessions (cli.py) delegates to the shared renderer. - hermes sessions list keeps the Workspace column only when --workspace is explicitly passed; otherwise canonical table. - Search handlers now build canonical row dicts and delegate, deleting ~90 lines of duplicated render logic. Tests: 21 resume tests + 58 session tests pass.
- hermes sessions search [--limit N] — cap results (default 10, max 100). Limit is applied AFTER dedup + recency sort, so --limit N always means 'N most recent matches' regardless of FTS5 rank order. - /sessions search <query> [--limit N] (interactive) — same semantics. - Bare /sessions and /resume listings now print a 'More:' footer pointing to /sessions search, all, full, --limit. - /sessions search with no query prints a fuller usage block. - hermes sessions --help and sessions search --help now describe search, dedup behavior, and --limit. - Bonus: dropping the mid-iteration cap lets the default search show sessions the old code missed (e.g. Kanban task summary) — the coverage gap vs the agent's session_search is closed. - Gateway /sessions search usage message now shows an example. Tests: 95 pass (resume, session listing, workspace, cli init).
- hermes sessions search/list accept -l N as shorthand for --limit. - Interactive /sessions accepts -l N (bare listing and search), and /sessions -l N now actually limits the bare listing (previously the flag was stripped but ignored on the non-search path). - Footers and usage blocks trimmed to only the options the user asked for: search and -l. Removed all/full mentions (pre-existing parser flags, but not requested).
…s to 20 - After /sessions search, the displayed # column is now actionable: the handler arms the one-shot pending-resume snapshot with the search results, so '/resume 2' (or just typing '2') resumes search match NousResearch#2. - _handle_resume_command now prefers the armed snapshot for explicit numeric targets (falling back to the recent list), so the numbers on screen are always the numbers that work — no drift if new sessions appeared since the list was shown. - Interactive listing defaults raised 10 -> 20 to match 'hermes sessions list' (so a number seen there, e.g. 13, resolves interactively too). Search default stays 10 (matches CLI search); -l overrides either. - Footers now say '/resume <number> (the # column above)' and the CLI search footer shows the real resume paths: /resume <number> in interactive, hermes --resume <id> from the shell. - Fixed a false help claim: 'hermes sessions <id>' does NOT resume; the CLI resume path is 'hermes --resume <id>'. Tests: 42 pass.
The # column in /sessions search and 'hermes sessions search' now shows each result's position in the canonical sessions list (the same list as 'hermes sessions list': all sources except tool, unnamed included, ordered by original start time, compression chains projected to their live tip) instead of a renumbered 1..N of the matches. So 'Add oh-my-openagent docs to qmd collection' shows 13, matching the listing the user sees, and /resume 13 resolves it. - session_rank_lookup() in session_listing.py: id -> rank via the exact list_sessions_rich query the CLI list uses (limit 500 window). - render_sessions_table: rows can carry 'rank'; renders it in the # column with a width sized to the largest number (fallback: sequential index when rank is absent, e.g. plain listings). - _list_recent_sessions now queries the same canonical list (all sources, current session included) so interactive /sessions, /resume, and /history show the same numbers and /resume <N> resolves the number on screen. Dropped the search-result snapshot arming from the previous commit — displayed numbers are global ranks now, so resolution always goes through the global list. - Test updated: /history lists the current session (matches the canonical list; previously asserted it was hidden). Tests: 95 pass.
The rank lookup only contained projected *tip* ids, but FTS5 search hits roots and mid-chain sessions just as often (old messages match). Those resolved to None and silently fell back to the sequential result index — producing the mixed garbage the user saw (4, 80, 3, 4, 262, 6, 7 with duplicate 4s and impossible small numbers). Now session_rank() walks forward along continuation children (latest started_at first, max 20 hops) until it finds an id in the projected rank map — every chain's live tip is in the map — and returns that chain's position. Result for 'search tmux' is now 4, 80, 44, 80, 262, 303, 552: all real positions in # Title Model Tok Created Last Preview ID ── ───────────────────────────────────────────────── ────────── ────────── ────────── ──────── ──────────────────────────────────────── ──────────────────────── 1 — deepseek-… 0/0 2026-07-31 4h 20260731_071846_9fa701 2 — deepseek-… 30k/8k 2026-07-31 4h can you fix copy selected to clipboar… 20260731_065709_d308d6 3 fintwit-daily-fetch · Jul 31 06:07 deepseek-… 28k/412 2026-07-31 5h [IMPORTANT: You are running as a sche… cron_9223b2202bd2_20260731_060057 4 Tmux config and scripts review summary deepseek-… 99k/69k 2026-07-31 4h review .tmux config and scripts in .l… 20260731_055644_5e1c37 5 — deepseek-… 30k/10k 2026-07-31 6h is there a way to open all folders in… 20260731_052757_d3bd37 6 Managing Personal Hermes Skills Organization deepseek-… 188k/54k 2026-07-31 7h I realiezd that I need to keep my per… 20260731_030112_fd9939 7 Search sessions for data collector deepseek-… 64k/1k 2026-07-31 9h sessions search data-collector 20260731_014020_1bec3c 8 Update Plan Skill with Discovery Phase deepseek-… 52k/24k 2026-07-31 9h [User attached file: /home/shiro/.her… 20260731_012537_aa571d 9 Search Hermes browser iMac socat session deepseek-… 117k/44k 2026-07-31 9h search session where we worked on /br… 20260731_004120_67edd7 10 Hermes CLI Sessions and Resume Commands NousResearch#13 deepseek-… 121k/56k 2026-07-31 2s [CONTEXT COMPACTION — REFERENCE ONLY]… 20260731_113524_142504 11 Linking iMac Chrome to opencode browser deepseek-… 277k/20k 2026-07-31 7h I want to give opencode the same brow… 20260731_000128_f44ad5 12 Improving Hermes Agent Codebase Access Beyond RAG deepseek-… 72k/18k 2026-07-31 11h Looking to improve hermes agent, trie… 20260730_234830_1a2cdd 13 Add oh-my-openagent docs to qmd collection deepseek-… 108k/40k 2026-07-30 11h can you add oh my openagents docs in … 20260730_232541_0a40bc 14 fintwit-daily-fetch · Jul 30 06:04 deepseek-… 28k/358 2026-07-30 1d [IMPORTANT: You are running as a sche… cron_9223b2202bd2_20260730_060042 15 Monid setup requires API key deepseek-… 35k/6k 2026-07-29 1d set up https://monid.ai/SKILL.md 20260729_230009_5f48fd 16 Quick storage cleanup with approval deepseek-… 45k/15k 2026-07-29 1d find quick way to free storage space,… 20260729_215349_39b397 17 Phone SSH Fix and Termux Venv Setup deepseek-… 95k/39k 2026-07-29 1d help me fix my phone tmux launcher. s… 20260729_205759_70586e 18 Config file fixes for opencode deepseek-… 71k/26k 2026-07-29 1d can you fix @file:opencode.json and @… 20260729_181012_522948 19 Setup FaceSwap Google Colab Notebook deepseek-… 55k/6k 2026-07-29 1d @file:README.md I want to test this t… 20260729_163129_b94fe9 20 Fix Discover Offline Cloudflare Repo Error deepseek-… 41k/5k 2026-07-29 1d trying to update my apps in Updates -… 20260729_065842_ee8947, with the root/tip pair of one chain correctly sharing its slot (80). Rank window raised 500 -> 2000 so the whole store is covered (569 projected rows today). Tests: 95 pass.
…In/Out) The dedup only dropped ancestors that were directly in the FTS5 result set. When intermediate generations were missing (search matched only e.g. NousResearch#7 and NousResearch#13), the walk stopped and both children were listed twice. Now every result is grouped by its deepest compression ancestor (backward walk across compression edges only — branch children stop the walk so distinct conversations are never collapsed) and the newest descendant wins, so one conversation = one row regardless of which generations matched. Uses a backward-only walk instead of get_compression_lineage, whose forward walk assumes a linear chain and fragments on divergent children (seen in the 'Script unique pwd' chain). Also renames the token column header from 'Tok' to 'Tok(In/Out)' (input/output) per request. Tests: 95 pass.
Search rows were built from get_session(matched_generation), so Created showed when the matched compression child spawned (e.g. 'Script unique pwd NousResearch#8' = 07-28) instead of when the conversation began (root = 07-25). Now root_started_at() resolves the deepest compression ancestor and the row's started_at is overridden with the root's, matching what the plain listings already show on projected rows. Last column is unaffected (keeps the tip's last_active). Tests: 95 pass.
…it back to 10 - 'hermes sessions list [PAGE]' now takes an optional page number; page size = --limit (default 10, max 100). 'sessions list 2' shows rows 11..20, 'sessions list -l 5 3' shows rows 11..15. The # column renders the global position in the canonical list (offset + idx), so page 2 shows 11..20 and matches the ranks search results display. - '/resume list [page]' and '/sessions list [page]' do the same interactively and arm the one-shot numbered selection with the page's limit+offset, so a number on screen resolves immediately (bare number or /resume <N>) even on page 2. - '/resume <N>' resolution is offset-aware: it maps N against the stored page (global ranks), falling back to offset 0. Stale pages can't shift later resolution: any non-session command resets the stored limit/offset (mirrors the existing one-shot pending disarm). - Defaults restored to 10 everywhere (hermes sessions list --limit, /sessions, /resume bare, _list_recent_sessions, _show_recent_sessions). Tests: 95 pass.
hermes sessions list now prints a two-line tip after the table ([PAGE] [-l N] [--source SRC] [--workspace NEEDLE]) and defers the full parameter reference + examples to 'hermes sessions list --help' (added an argparse epilog with pagination/source/workspace examples, RawDescriptionHelpFormatter keeps the layout). Interactive /sessions and /resume listings get the same treatment: a two-line footer covering resume-by-number/id/title plus search and paginated listing, instead of the verbose usage dump. Tests updated to the new footer wording (95 pass).
_consume_pending_resume_selection bounds a typed number against len(pending) (page-relative) but _handle_resume_command converts it with the stored offset (global -> local). On any page with offset > 0 every valid selection failed one of the two checks: a visible rank (e.g. 15 on page 2) was rejected by the guard, and a number that slipped past it computed a negative local index. Guard against the displayed window [offset+1, offset+len(pending)] and forward the global rank, which the existing conversion resolves correctly. Adds page-2 bare-number regression tests.
…inition Search sorted and displayed COALESCE(ended_at, started_at, 0) while the listing uses MAX(messages.timestamp). A session that idled after its final message showed a search Last in the future relative to its own listing row, and the search sort inherited the same wrong key — a closed-2h-ago session could sort above a live one. Adds last_active_of() (latest message ts, falling back to ended_at/started_at for empty sessions) to session_listing.py and wires both the CLI and interactive search pipelines to it, so order and the Last column agree with list_sessions_rich. Real-DB tests assert the definitional equivalence.
The listing projection swapped id/title/last_active to the live tip but kept the root's input/output_tokens, so a projected chain row displayed the root's historical usage — for the 19-gen conversation in the real store that was 121k/56k while the tip alone had 109k/6k; neither is the conversation's usage. Adds SessionDB.chain_token_totals() + a recursive-CTE batch sum that walks compression-continuation edges (excluding branch/delegate/tool children, mirroring get_compression_tip) and applies it in the listing projection and both search pipelines. Header renamed to Tok(ΣIn/ΣOut) to make the semantics explicit. Real-DB tests: 3-gen chain sums, branch/ delegate/tool exclusion, standalone passthrough.
Addresses hermes-sweeper review on PR NousResearch#75496: - Search dedup now projects ancestor-only FTS hits to the live compression tip and synthesizes the tip's row, so the displayed row, rank, and Last column always describe the same generation. - The preview/root walk reuses _compression_root() (edge-aware) instead of following every parent_session_id, so branch children show their own opener instead of the parent conversation's. - Gateway /sessions search over-fetches the candidate pool (200) before the origin filter so cross-room matches can't starve the caller's results out of the visible page. Two regression tests cover the projection and branch-preview cases.
Author
|
Addressed all four points.
97/97 tests pass across the target suites. |
Esashiero
marked this pull request as ready for review
July 31, 2026 16:26
alt-glitch
added
type/bug
19 tasks
…d session commands) Resolves 5 conflict zones (cli.py, gateway/slash_commands.py, gateway/platforms/api_server.py, hermes_state.py, tests/hermes_cli/test_session_listing.py) by taking the branch's canonical session-listing logic (shared query_session_listing / search_session_listing across CLI + gateway) and preserving main's unrelated fixes (kanban exclusion, pin-windowing, admin cross-origin widening, _is_branch_child_row method that was dropped during the auto-merge of hermes_state.py). One main-only gateway test (test_sessions_busy_platform_..._lane) is skipped: it asserts pre-branch per-lane session scoping, which the branch intentionally replaced with per-source scoping to match CLI semantics. All 123 other affected tests pass; real-DB smoke test of list / list 2 / list -l 5 3 / search confirms correct pagination, global rank continuity, and chain-aware dedup.
…lists The sessions refactor dropped the SQL-level session_key filter in _list_titled_sessions and the /sessions non-search branch, replacing it with a source-wide fetch + Python per-origin visibility filter. On a busy platform the caller's own lane sessions starve out of the pre-filter limit window (repro: 60 foreign sessions -> 0 lane rows, 'No sessions found.'). Restore the pre-refactor semantics: pass session_key to query_session_listing in both call sites (widen/cross_origin only for an admin's explicit --all). The Python _resume_row_visible filter stays as a second defense for the admin-widen and Matrix paths. The three TestResumeListingSemantics fixtures created telegram sessions without session_key (NULL routing key — unroutable rows that real gateway flows never produce, since session_key is written at CREATE time); they now carry build_session_key(event.source), matching the handler's own key.
… decouple imports _chain_token_totals hardcoded its recursive-CTE depth bound (c.depth < 100) while the module also exported COMPRESSION_CHAIN_MAX_HOPS = 100 — a DRY violation that would silently skew totals if the constant ever changed. Interpolate the constant in the existing f-string (value identical, 100). hermes_cli/session_listing.py imported COMPRESSION_CHAIN_MAX_HOPS through the hermes_state re-export; import it from hermes_state_common directly so the CLI listing module does not couple to the giant state module's namespace. The hermes_state re-export stays for back-compat. Behavior-identical; verified 37 tests green, zero new diagnostics.
The /sessions search path fetches a global pre-filter pool (200 content + 200 title rows) and aggregates in Python, with no SQL-level lane filter. On a busy platform, foreign sessions with fresher matches crowd the caller's own lane out of that window entirely (repro: 210 foreign matches -> 'No sessions found.'), the same starvation class fixed for the list paths in 0ea1610. Thread a keyword-only session_key through search_messages, _search_messages_impl, _search_unindexed_gap and the LIKE fallback, and append s.session_key = ? in all seven WHERE builders (FTS5, CJK, trigram, LIKE x2, gap scan, unindexed LIKE). search_session_listing passes it into both the content pool and list_sessions_rich title pool; the gateway search branch passes session_key (None only for an admin cross-origin --all). CLI call sites keep the default None, so CLI behavior is unchanged.
Esashiero
force-pushed
the
sessions-resume-improvements
branch
2 times, most recently
from
August 20, 2026 12:18
8faed47 to
30bd45c
Compare
Author
|
Quick note on the branch history for reviewers: This PR head ( What that means for you:
If you need to compare against the original pre-rebase series, the old tip was Closes the stale-base issue; ready for review. |
This was referenced Aug 20, 2026
Closed
Closed
Closed
Contributor
|
Note from #91312 (compression-aware lineage reconstruction for the agent \session_search\ tool): our chain-identity semantics match — positive compression-continuation roots, same \COMPRESSION_CHAIN_MAX_HOPS\ family — but the surfaces are complementary. This PR owns the CLI /sessions\ listing/search UX (row-per-chain, pagination, lane-scoping); #91312 owns the agent tool's bounded/fail-closed lineage reconstruction (B=2000 budget, query-local memo, lineage summarization) which this PR does not implement. Classified as COEXIST. If this merges first, the #91312 branch will re-audit ownership of the lineage seams rather than mechanically rebasing. |
Contributor
|
#128 (fork) is being contributed upstream as a Draft PR: indexed session metadata search over title/id/display_name. This PR is a WATCH item for that contribution. This PR redesigns the CLI/gateway sessions list/search and touches Related: Skywind5487#128, #140 (benchmark), #141 (gate). |
This was referenced Aug 21, 2026
10 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this adds
New options and columns for the sessions CLI, plus a search subcommand. The interactive
/sessionsand gateway/sessionsshare the same logic.hermes sessions list-l/--limit— sessions per page (default 10, max 100)pagepositional —hermes sessions list 2shows page 2 (rows 11-20)#(rank),Tok(ΣIn/ΣOut)(token totals across the compression chain),Created,Preview(the chain root's first message instead of the COMPACTION banner)--workspacefilter works on any pagehermes sessions search <query>(new subcommand)Shared fixes (CLI +
/sessions+ gateway):/resume Nworks on any page/sessionsand/resumeuse the same row semantics as the CLI (tool sessions excluded, same ranking)list 2shows 11-20; search ranks match the list)Follow-up fixes (consolidated from the duplicate #90619)
This branch is the earlier #75496 head, rebased onto current
mainand carrying three cleanup fixes:fix(gateway): restore lane-scoping in /sessions and numbered /resume lists(0ea16106e) — the sessions refactor had dropped the SQL-levelsession_keyfilter, so a busy lane with many foreign sessions could starve the caller's own lane out of the pre-filter window (repro: 60 foreign sessions -> 0 lane rows, "No sessions found"). Restoredsession_key=None if widen/cross_origin else session_keyat both gateway call sites. Test fixtures now carry a realbuild_session_key(NULL routing keys were an unrealistic fixture).refactor: pin chain-token CTE depth to COMPRESSION_CHAIN_MAX_HOPS and decouple imports(70aa00924) — chain-token totals now use the shared hop cap constant instead of a magic number, and the import surface is decoupled.fix(sessions): lane-scope FTS5 search candidate pools via session_key(30bd45c43) — threads a keyword-onlysession_keythroughsearch_messagesand appendss.session_key = ?in all seven WHERE builders (FTS5, CJK, trigram, both the content pool and thelist_sessions_richtitle pool). Gateway search passessession_key(None only for an admin cross-origin--all).Files
16 files changed against current
main:cli.py,gateway/platforms/api_server.py,gateway/slash_commands.pyhermes_cli/cli_commands_mixin.py,hermes_cli/main.py,hermes_cli/session_listing.py,hermes_cli/sessions_cmd.pyhermes_state.py,hermes_state_common.py,hermes_state_search.pytests/cli/,tests/gateway/,tests/hermes_cli/Test plan
pytest tests/cli/test_cli_resume_command.py tests/cli/test_cli_init.py tests/hermes_cli/test_session_listing.py tests/test_session_workspace_binding.py tests/gateway/test_resume_command.py tests/hermes_cli/test_sessions_workspace_pagination.py -o asyncio_mode=auto -q→ 125 passed, 1 skipped.hermes sessions list,list 2,list -l 5 3,sessions search tmux— pagination, global-rank continuity between list pages and search, and no duplicate conversation rows all verified.Known limits (this started as a fast WIP)
main; rebased onto currentmainto give a clean diff. All 45 original feature commits are retained in the series, plus the 3 follow-up fixes.s.session_key = ?filter blocks into a shared WHERE-builder helper.Branch:
sessions-resume-improvements(tip8faed4730), forkEsashiero/hermes-agent. Supersedes #90619.