Per Teknium: the caps should bound a single agent loop, not accumulate
over the whole session. Rename SessionCapConfig -> LoopCapConfig and the
config section session_caps -> loop_caps; move the counters into
reset_for_turn (invoked per turn via turn_context) so each turn starts
with a fresh budget; retune defaults 200 -> 50 (a single turn issuing 50
web searches / spawning 50 subagents is already pathological). Block
codes session_*_cap -> loop_*_cap and messages updated to drop the
/new-resets-the-budget guidance (irrelevant now that it resets per turn).
Tests flipped: the old persists-across-turn-resets assertion becomes
resets-each-turn.
Add per-session lifetime caps on web_search calls and subagent spawns
(defaults 200/200, matching Claude Code v2.1.212). Unlike the existing
per-turn tool-loop guardrails, these count over the whole session and
reset only when a fresh agent is built (/new, /clear). Hitting a cap
blocks the offending call and halts the turn cleanly.
- agent/tool_guardrails.py: SessionCapConfig + session counters on the
controller (in __init__, not reset_for_turn, so they persist across
turns). before_call() enforces caps first, independent of
hard_stop_enabled. delegate_task batches count each task.
- hermes_cli/config.py: tool_loop_guardrails.session_caps defaults.
- docs + tests (unit + E2E validated against a real AIAgent).
After N consecutive guardian denials in a session the deny message escalates to a hard-stop instruction. Inspired by ChatGPT Work auto-review circuit breaker.
DEFAULT_CONFIG declared "kanban" twice. Python keeps only the last
literal for a duplicate key, so the first kanban block was silently
dropped and its "auto_subscribe_on_create": True default never made it
into DEFAULT_CONFIG. The consumer in tools/kanban_tools.py masks the
miss with cfg_get(..., default=True), but the documented default was
absent from DEFAULT_CONFIG (so config templates / 'hermes config show'
omit it), and the duplicate key is a standing hazard: any future key
added to the first block would also vanish.
Merge auto_subscribe_on_create into the single canonical kanban block.
Adds a regression test asserting both default sets survive and a guard
against any duplicate top-level DEFAULT_CONFIG key.
The default max_iterations/agent.max_turns budget was set when long
agentic runs were rare; complex tasks now routinely exceed 90 tool
calls. Raise the default to 500 across every surface that hardcodes
the fallback: AIAgent constructor, DEFAULT_CONFIG, CLI resolution
chain, gateway env bridge, cron scheduler, and TUI gateway. Explicit
user config values are unaffected (deep-merge preserves them; no
_config_version bump needed).
Docs (en + zh-Hans), CLI help text, tips, and pinned tests updated
to match.
Teknium review changes on the tiered policy:
1. threshold_pct default 10 -> 5 (listing budget = min(5% of context,
listing_max_tokens)); unknown-context fallback 20K -> 10K.
2. Tier 2 no longer leaves the model blind: when even names-only doesn't
fit, the bridge description now carries a one-line-per-server summary
('cloudflare (3320 tools)') plus an instruction to search FIRST rather
than substitute a generic tool or claim the capability is missing —
the measured tier-2 failure mode (core-tool substitution) at zero
meaningful token cost (~50 tokens/server).
3. Listing degradation is now PER SERVER, largest first: one oversized
server (Cloudflare) collapses to its summary line while small
co-attached servers (Linear) keep their full per-tool listings
('mixed' form). Previously global: attaching Cloudflare next to
Linear silently cost Linear its listing. Greedy fit is deterministic
(size then label) so the rendered block stays byte-stable per catalog
— prompt-prefix cache safe.
E2E on real captures (defaults, 200K ctx): linear alone -> tier 1 full;
unreal alone -> tier 2 groups (5% budget) / tier 1 names at 1M;
cloudflare alone -> tier 2 groups; linear+cloudflare -> tier 1 MIXED
(linear fully listed, cloudflare summarized). 48/48 tests.
Tier 0: no MCP/plugin tools -> everything eager (pass-through).
Tier 1: deferred tools whose catalog listing fits min(threshold_pct%
of context, listing_max_tokens) -> bridge + skills-style
listing, degrading to names-only over budget.
Tier 2: listing over budget even names-only (Cloudflare's flat API
surface: 3,320 tools, names alone ~32K tokens) -> bare bridge,
discovery through tool_search only.
The old activation threshold (defer only when schemas > threshold_pct
of context) let mid-size catalogs ride eager and pay full schema cost;
with servers like Cloudflare (~597K tokens of schema, would not even
fit a 200K window) the binary gate is the wrong shape. Activation is
now driven purely by deferrable-tool presence; threshold_pct is
repurposed as the listing budget's context-relative leg.
- AssemblyResult gains tier + listing_form for observability
- listing_max_tokens default 4000 -> 20000 (cap 60000) so an 830-tool
catalog keeps a names-only listing while Cloudflare-scale drops to
bare bridge
- E2E verified against real captures: Linear 24 tools -> tier 1 full,
Epic UE 5.8 830 tools -> tier 1 names-only, Cloudflare 3,320 tools
-> tier 2 (both 200K and 1M context)
Deferred MCP/plugin tools become invisible once the tool_search bridge
activates — live benchmarking (48 runs, Claude Haiku 4.5) showed models
substituting visible core tools (terminal/web_search/browser) for deferred
capabilities or declaring them nonexistent instead of searching: 16/24
task success vs 24/24 with eager loading.
Skills never had this failure mode because every skill keeps a ~21-token
name+description listing line in the system prompt. This ports that exact
pattern to the tool bridge: when tool_search activates, a grouped manifest
of every deferred tool (name + first sentence of description, clipped to
60 chars, grouped per MCP server / toolset) is embedded in the tool_search
bridge description.
- tools/tool_search.py: build_catalog_listing() with deterministic
ordering (byte-stable across assemblies -> prompt prefix stays
cacheable); token-budget fallbacks full -> names-only -> legacy bare
count; bridge_tool_schemas(listing=...) embeds it and instructs the
model to skip tool_search when the exact name is visible (one fewer
round-trip per use)
- config: tools.tool_search.listing auto|on|off (default auto),
listing_max_tokens (default 4000, clamped 200..20000); legacy bool
shapes keep working
- tests: 8 new tests (config parsing/clamps, short-desc clipping,
deterministic rendering, budget fallbacks, bridge embedding, assembly
on/off paths); full file green (47 passed)
- docs: tool-search.md config table + rationale
- scripts/tool_search_livetest2.py: benchmark harness v2 with real
per-call token accounting (normalize_usage spy) and a third 'listing'
mode for A/B/C comparison
2107b86024 flipped compression.in_place to True but left both explanatory
comments reading "Default False during rollout". The contradiction is
load-bearing: it is why two recent PRs (#71747, #48951) were built on the
premise that agent.session_id still rotates at every compaction.
Comment-only; no behavior change.
Some OpenAI-compatible endpoints — notably Tencent Copilot
(copilot.tencent.com) — only accept streaming chat requests; any
non-streaming call returns HTTP 400 (code 11101, 'Non-stream chat
request is currently not supported'). The main conversation loop already
streams, so interactive chat works, but every auxiliary task (title
generation, compression, web extraction) used the non-streaming path and
failed on each call.
_provider_requires_stream() detects stream-only endpoints
(copilot.tencent.com built in, plus user-configurable
auxiliary.stream_only_base_urls substring markers in config.yaml).
Matching sync auxiliary calls route through _create_with_progress
(force_stream=True) and async calls through the new
_acreate_with_stream, aggregating the chunk stream — including tool-call
deltas and reasoning deltas — into a complete response via the shared
_ChatStreamAccumulator.
Salvaged from PR #60686 by @kudi88 onto the progress-aware streaming
machinery from #71508, addressing both sweeper-review gaps: the async
path now consumes the stream with 'async for' (awaiting create() and
iterating synchronously raised on AsyncOpenAI streams), and tool-call
deltas are reassembled instead of dropped (MCP passes tools= through
call_llm). Under force_stream there is no silent non-streaming retry —
a stream-only provider rejects those by definition, so the original
error surfaces to the normal recovery chains.
The gateway's pre-agent session-hygiene compression killed the summary
call at a fixed 30s wall-clock deadline (compression.hygiene_timeout_seconds),
regardless of whether the summary model was hung or merely slow. A reasoning
model happily streaming a large summary was cut off mid-generation, the user
got '⚠️ Context compression timed out after 30.0s', and a 300s failure
cooldown left the session oversized — a doom loop for slow-but-healthy
auxiliary models.
Timeouts are now liveness-based instead of wall-clock-based:
- agent/auxiliary_client.py: new thread-local aux_progress_hook. When
installed (only by context compression today), the primary call_llm
attempt streams (stream=True) and aggregates chunks back into a complete
response, ticking the hook per chunk. The configured timeout then acts
per stream read (idle) instead of as a total budget. Providers that
reject streaming fall back to the plain non-streaming call; auth/payment/
rate-limit/transport errors propagate unchanged into the existing
recovery chains. Codex Responses (per SSE event) and Anthropic Messages
(per stream event, via the new create_anthropic_message on_stream_event
callback) tick the same hook from inside their wire adapters.
- agent/conversation_compression.py: CompressionCommitFence gains
touch_progress()/seconds_since_progress(); compress_context() installs
fence.touch_progress as the progress hook around the compress call.
- gateway/run.py: the hygiene wait loop treats hygiene_timeout_seconds as
an INACTIVITY budget — while the fence reports fresh progress the wait
extends, bounded by the new compression.hygiene_total_ceiling_seconds
(default 600s, clamped >= the idle budget) so a degenerate trickle
stream still dies. The timeout warning now says the summary model
produced no output, which is the only case that still triggers it.
- config/docs: hygiene_total_ceiling_seconds added to DEFAULT_CONFIG and
configuration.md; hygiene_timeout_seconds documented as inactivity-based.
Tests: tests/agent/test_aux_progress_streaming.py (hook plumbing, stream
aggregation incl. tool-call deltas and reasoning deltas, rejection
fallback, ceiling kill, fence progress surface); two new gateway tests
prove a slow-but-streaming worker survives past the fixed timeout
(sabotage-verified: fails with the old fixed deadline) and a
forever-trickling worker is still cut off at the ceiling.
Leg 2 of #63078: prompt.submit returns {"status":"streaming"} immediately and
runs _start_agent_build + _wait_agent(timeout=30s) behind it. The deferred
build (MCP discovery with per-server retry backoff, synchronous model-metadata
HTTP, skills scanning) routinely outlives 30s on cold starts; on timeout
run_after_agent_ready emitted an error EVENT and returned without ever calling
_run_prompt_submit — the user's first message was permanently discarded while
the build finished successfully in the background. The desktop's optimistic
row eventually cleared with no visible error: the blank first session.
New _wait_agent_for_prompt replaces the flat cliff for the deferred prompt
path only (_sess()'s RPC-blocking _wait_agent keeps its 30s contract):
- The pending prompt stays attached to the (already off-RPC) run thread and
is delivered the moment the still-running build completes — a slow build is
no longer message loss.
- The wait runs in 5s slices so a cancel (session.interrupt / churn) is
honored promptly; the cancelled path returns None and defers to the
caller's cancel branch (the #65567 emit) for user-visible messaging.
- Past 30s the client gets ONE keyed notification.show ('Still starting the
agent…', key=agent-build-slow, desktop toast / TUI status bar), cleared on
delivery — patient, never silent.
- Permanent failure only when the build itself fails: agent_error set at
ready, the build thread died without signalling ready (fail fast via the
new _agent_build_thread handle instead of sitting out the cap on a corpse),
or the bounded cap expired on a genuinely hung build. The cap defaults to
600s and is tunable via agent.build_wait_timeout in config.yaml (no new
env vars); the error message states the message was not sent.
Tests: slow-build delivery with zero error events; the keyed progress notice
shown once and cleared; build-failure surfacing exactly one error event with
the real reason; dead-thread fail-fast; cancel honored mid-wait; config
override + fallback semantics; cap expiry message. The compute-host fallback
test stubs the new waiter alongside _wait_agent.
The renderer's session-state cache is memory-only and the backend's
inflight snapshot dies with the backend process, so nothing survived a
full app or machine death mid-turn: reopening the session showed the
transcript up to the last committed turn and silently dropped everything
the crashed turn had streamed.
While a turn runs, the visible tail (user prompt + streamed assistant
rows, tool calls included) is now journaled to localStorage — throttled
off the delta-flush hot path, bounded (24 entries / 7 days), cleared the
moment the turn settles. Session resume folds the journaled tail back
onto the restored transcript. When the backend also has a live text-only
inflight projection for the same turn, the journal overlays its richer
structure onto that row (longer text wins, base row id kept so live
deltas keep landing) instead of treating it as caught up — the ordering
defect that dropped locally recorded tool progress in the original PR.
Co-authored-by: Omar Baradei <omar@kostudios.io>
Both the Desktop panel and the CLI setup flow need somewhere in .env to put
a custom endpoint's API key. Deriving the name from the endpoint's hostname
collapses two servers on one machine onto a single slot, and every IP-based
local endpoint slugs to a digit-leading name that save_env_value rejects
outright. Key off the endpoint's own identity and keep a fixed prefix.
Co-authored-by: asorry75 <33794789+asorry75@users.noreply.github.com>
Co-authored-by: webtecnica <75556242+webtecnica@users.noreply.github.com>
Follow-up to the salvaged #69164 commits: policy forbids introducing new
HERMES_* environment variables, so the four watchdog env knobs
(HERMES_GATEWAY_LOOP_WATCHDOG / _INTERVAL / _TIMEOUT / _STRIKES) are
replaced with a single config.yaml boolean:
gateway:
loop_watchdog: true # default; false disables both guards
- gateway/config.py: new GatewayConfig.loop_watchdog field (default True),
parsed from top-level or nested gateway: form, round-trips via
to_dict/from_dict.
- gateway/run.py: _start_loop_liveness_guards() checks config.loop_watchdog
before arming the floor timer + watchdog (getattr-guarded for bare
object.__new__ runners).
- gateway/shutdown_watchdog.py: start_loop_liveness_watchdog() no longer
reads the environment; probe interval/timeout/strikes are module
constants (30s/10s/3 — ~90s to restart, matching the systemd watchdog
layer's posture).
- hermes_cli/config.py: documented gateway.loop_watchdog default so
'hermes config set gateway.loop_watchdog false' validates.
- tests: env-knob tests replaced with config-gate + round-trip tests;
the final-strike boundary test injects its probe via max_strikes
directly instead of patching the removed env helper.
The .env sanitizer inferred missing newlines from known KEY= substrings
inside existing values. Plain secrets containing those bytes could therefore
be split into synthetic assignments and rewritten to disk.
Treat each physical line as the only assignment boundary and keep bytes after
the first equals sign opaque for boundary discovery. Preserve safe formatting,
null-byte removal, BOM handling, and normal one-assignment-per-line parsing.
Cover direct loading, dotenv loading, sanitization, writers, and migration
with behavioral regressions.
Fixes#29155
Builds on this PR's diagnosis by @Frowtek: a missing workdir is
ambiguous (deleted project vs. an unmounted external volume / network
share / VPN not yet up), so it's not safe evidence for a destructive
GC sweep — especially one that runs unattended at startup.
- cli.py / gateway/run.py: the startup auto-maintenance sweep now
always passes delete_orphans=False to maybe_auto_prune_checkpoints().
It still prunes by retention_days, size cap, and legacy archives —
none of which require guessing whether a project was deleted or is
just temporarily unreachable.
- hermes_cli/config.py: drop the now-unused delete_orphans default.
- hermes_cli/checkpoints.py: `hermes checkpoints prune` (the explicit,
human-invoked path) now previews the orphan project list and asks
for confirmation before deleting, unless -f/--force is passed.
- Docs updated (EN + zh-Hans) to match.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
New sessions.auto_archive / auto_archive_days config: soft-hide (never
delete) sessions with no activity for N days, aging on last activity
rather than creation so an old-but-active chat is spared. Sweeps are
throttled through state_meta and fire from CLI startup, gateway startup
+ hourly housekeeping, and the serve/dashboard backend (opportunistic
on session list + an hourly lifespan ticker), so every surface honours
one setting.
A new pinned column (declaratively migrated) exempts sessions from the
sweep; PATCH /api/sessions/{id} accepts pinned and flips the whole
compression lineage as a unit, mirroring set_session_archived.
Routine automatic compression stays silent-by-design on chat platforms
(default unchanged, byte-identical). New opt-in config key
compression.progress_notices (bool, default false) opens a gate on the
gateway noise filter (_prepare_gateway_status_message) that lets ROUTINE
compression progress statuses through to chat surfaces.
- Membership is derived from the #69550 status template constants in
agent/conversation_compression.py (compiled to a literal-escaped regex,
never re-inlined wording), so unrelated noisy statuses (aux failures,
provider retry/rate-limit chatter) stay suppressed even when enabled.
- The compaction completion notice (COMPACTION_DONE_STATUS, #69546
lifecycle 'compacted' edge) already flows through the status path and
passes the filter — no new emit site needed.
- Config wired everywhere: hermes_cli/config.py DEFAULT_CONFIG,
cli-config.yaml.example, gateway raw-YAML read (live, mtime-cached),
gateway hot-reload cache-busting key list, website configuration docs.
- Failure notices and manual /compress feedback remain always-visible;
VISIBLE_COMPRESSION_MESSAGES and emit sites untouched.
Design by @havok-training (issue #52995).
Adds moa.privacy_filter ('' | display | full, default off — issue #59959):
- display: redact user-visible surfaces only (reference blocks emitted to
the UI + saved MoA trace records, including per-advisor full input/output
and the aggregator-input copy); the aggregator sees raw advisor text so
synthesis quality is unaffected.
- full: additionally redact the advisor text injected into the aggregator
prompt, on both the persistent facade path and the one-shot /moa
synthesis path (the issue's literal ask). Legacy boolean true maps here.
Secret/credential shapes (API-key prefixes, JWTs, private keys, DB
connection strings) are delegated to the central redactor
(agent.redact.redact_sensitive_text, force=True + code_file=True); the MoA
filter adds only email and clearly delimited phone-number patterns. No
bare 10-digit matching: line numbers, timestamps, epoch values, git SHAs,
IPs, versions, and source-code assignments in code-review-shaped advisory
text pass through byte-identical. The reference cache always holds raw
text — redaction happens at each consuming surface, so a mid-session mode
change never leaks or double-redacts.
Reworked from PR #60463: replaced its hand-rolled pattern list (which
matched bare digit runs and re-implemented key shapes) with central-
redactor reuse + safe patterns, and split the single boolean into
display/full modes. Credited for the feature framing.
Co-authored-by: webtecnica <75556242+webtecnica@users.noreply.github.com>
Follow-up fixes on top of the salvaged #22566 mechanism:
- N-collector now counts only REAL actionable user turns via
_is_actionable_user_turn + _is_synthetic_compression_user_turn —
the same filter pair _find_last_user_message_idx uses post-#69291.
The contributor's bare role=='user' + _is_context_summary_content
check let blank platform echoes and continuation/todo rows consume
N slots, silently degrading the guarantee.
- Default flipped 3 -> 1 (behavior-preserving): a default of 3 was
measured to change the tail cut on transcripts whose budget covers
only the last turn. min_tail_user_messages=1 delegates to the
existing single-user anchor; N>1 is opt-in, and the call site is
gated so the default path is byte-identical to main.
- Hardened config parse in agent_init (bool rejected, fractional
floats rejected, floor 1) matching the max_attempts parser shape.
- Wired the recurring external-PR config gaps: hermes_cli/config.py
DEFAULT_CONFIG + cli-config.yaml.example (PR only had cli.py).
- Regression tests: blank echoes / synthetic rows don't count toward
N; tool-call/result pairs never split by the N-boundary (no-orphan
both directions); N-guarantee wins over tail_token_budget and the
_MAX_TAIL_MESSAGE_FLOOR (floor is a minimum, not a cap); default
parity pin; DEFAULT_CONFIG pin.
The phase-1 tool-result prune only runs inside compress(), which fires
near 50% of the context window, so it never triggers on large-window
models; old tool outputs then ride in history and are re-sent every turn.
Add prune_tool_results_only(): the same no-LLM prune on a separate, low
proactive_prune_tokens trigger, run as an elif to the compression branch.
Opt-in (default 0), protects the recent tail by message count.
Add the method to the ContextEngine base as a no-op default so pluggable
engines inherit it safely (the post-tool-call path never AttributeErrors on
a non-built-in engine); the built-in compressor supplies the real prune.
Register both keys under the top-level compression config with defaults and
document them.
PR #37303 added per-platform streaming defaults and the commit message
explicitly called out "Discord/Slack/etc. only have edit-based streaming
(repeated editMessage), which flickers and is noticeably jankier" — but
only discord.streaming=false was shipped. Slack uses the same edit-based
streaming mechanism and has the same flicker problem, yet it was left to
follow the global switch (default true when streaming is enabled).
Add "slack": {"streaming": False} to DEFAULT_CONFIG["display"]["platforms"]
alongside the Discord default. The same deep-merge semantics apply: a user
who explicitly sets display.platforms.slack.streaming: true keeps their
value unchanged. The dashboard schema gains a slack.streaming toggle
automatically since it is generated from DEFAULT_CONFIG.
Update test_per_platform_streaming_defaults.py to cover slack in all
existing assertions and rename the resolver test to reflect both platforms.
Cut steady-state Computer Use latency without changing default behavior
or waiting on cua-driver:
- Cap screenshots via set_config(max_image_dimension) on session start
(config: computer_use.max_image_dimension, default 1456)
- Cache aux-vision routing per (provider, model) so captures skip
repeated load_config()
- Add computer_use.capture_after_mode (default som) so users can opt
follow-ups down to ax (elements only) for speed
Auto-detect now disables the cursor overlay on darwin as well as
headless/WSL2 Linux. After start_session, also call
set_agent_cursor_enabled(false) when the policy is on so older drivers
without --no-overlay still tear the overlay down.
Co-authored-by: David Metcalfe <80915+DavidMetcalfe@users.noreply.github.com>
cua-driver's cursor overlay rendering loop can consume CPU indefinitely
when idle (#28152, #47032). On Linux/WSL2, the overlay serves no visual
purpose and the rendering path is the primary source of idle CPU usage.
Add computer_use.no_overlay config option (default: auto-detect) that
passes --no-overlay to cua-driver when enabled. Auto-detection disables
the overlay on Linux (covers WSL2, headless, containers) where it has no
benefit, and keeps it enabled on macOS/Windows where it is visually
useful.
Refs: #28152, #47032
Port of the Slack half of #13855 (by @kshitijk4poor), reimplemented against
the plugin adapter (the original PR targets the deleted
gateway/platforms/slack.py and six other legacy adapters).
Channels listed in slack.require_mention_channels (config.yaml) or
SLACK_REQUIRE_MENTION_CHANNELS ALWAYS require an explicit @mention, even
when require_mention is false globally or the channel is in
free_response_channels — the opposite direction of free_response_channels.
Instead of duplicating the PR's inline reply-to-bot-thread/mentioned-thread/
session checks, the forced channel falls through to the SAME decision chain
as normal mention gating, so all five wake checks in
_should_wake_on_unmentioned_message keep applying (single decision path).
Credit: adapted from #13855 by @kshitijk4poor (Slack half only; the other
platform halves target deleted legacy adapters and are out of scope for
this cluster).
Once the bot is @mentioned in a Slack thread it auto-follows and replies
to every later message in that thread, including ones a human addresses to
another human (e.g. "@rasha check this out"). The bot butts in.
Add slack.ignore_other_user_mentions (env SLACK_IGNORE_OTHER_USER_MENTIONS,
default off). When on, a channel/thread message whose first token @mentions
someone other than the bot is treated as addressed to that person and the
bot stays silent unless it is also mentioned. This is Slack parity for the
Discord option of the same name (#33501), adapted to Slack's thread model:
the trigger is a leading mention ("addressed to"), so a message that merely
references another user mid-sentence still reaches the bot.
The gate sits ahead of the free-response / require_mention ladder so it also
overrides the mentioned-thread auto-follow. DMs are never filtered.
The streaming gate broadens from ElevenLabs-only to any provider that
passes check_tts_requirements(). In continuous voice mode a mic monitor
runs during playback: talking over the agent cuts TTS at detection
while the monitor keeps recording, then the captured interruption is
transcribed and queued as the next turn (process_loop's auto-restart
stands down while the capture owns the mic). New voice.barge_in config
key, default true.
Removes Homebrew and PyPI wheel/sdist as Hermes distribution paths while
preserving the supported source, Docker, and Nix workflows.
Changes:
- Removes the Homebrew formula, PyPI publish workflow, sdist manifest
(MANIFEST.in), and wheel/sdist release-attachment logic from scripts/release.py.
- Keeps setuptools metadata and entry points required by editable installs
and Docker/Nix builds, but adds a setup.py guard that rejects wheel/sdist
builds outside a sealed Nix derivation (HERMES_NIX_BUILD=1).
- Removes pip/Homebrew install detection, PyPI update checks, the pip
self-update path, the deprecation-banner state, the postinstall subcommand,
wheel data-directory fallbacks in agent/i18n.py and hermes_constants.py,
and the ACP Registry manifest/version-lockstep release logic.
- Adds /nix/store/ path detection so `nix run` / `nix profile install`
installs (which don't set HERMES_MANAGED) are correctly identified as
"nix" rather than falling through to "git"/"unknown".
- Retired install-method values ("pip", "homebrew") in existing
.install_method stamps (both code-scoped and home-scoped) are ignored by
the allowlist reader and fall through to "unknown" instead of resurrecting
a retired enum value.
- Updates Nix packaging to ship bare runtime data (locales, optional-mcps)
through store symlinks and wrapper env vars instead of wheel data-files.
- Removes the ACP Registry manifest/icon and their version-lockstep tests.
- Deletes or rewrites packaging, pip-update, Homebrew, and ACP Registry
tests; adds parametrized coverage for the packaging build guard covering
BOTH sdist and wheel paths (the guards live in separate cmdclass entries
— a passing sdist test proves nothing about the wheel path).
- Updates installation/platform documentation and related user-facing copy.
- Adjusts the supply-chain scan so deleted install-hook files do not trigger
a finding, while additions or modifications still require the existing
ci-reviewed label gate.
Supported installation paths (unchanged):
- git installer (install.sh)
- Docker
- Nix/NixOS
- editable development installs (uv sync, uv pip install -e ., pip install -e .)
_emit and _broadcast_global_event were each building the JSON-RPC event
envelope — extract _event_frame and use it from both. Type the registry as
set[Transport] (protocol already imported), and cut comment bloat at the
call sites. No behavior change; suites stay green.
Real-world failure from dogfooding the live-theme flow: display.skin was
already 'synthwave' in config, but the desktop never visibly applied it (the
activation event predated the WS transport fix / the connect). The desktop's
gateway.ready seed records the baseline WITHOUT painting (by design — never
stomp the persisted desktop theme on connect), so it believed it was synced.
Re-running 'hermes config set display.skin synthwave' then did nothing twice
over: the watcher signature (name, skin-file mtime) hadn't moved, so no
skin.changed fired; and even on an event, the desktop's name-equality guard
blocked the apply against the seeded baseline.
Two halves:
- hermes_cli: setting display.skin touches the named skin file so the
watcher signature always moves on an explicit set — a same-name re-affirm
now broadcasts skin.changed like any real move. Built-ins (no file) are
unaffected; a name switch already moves their signature.
- desktop: track whether the synced baseline was actually APPLIED vs merely
seeded at connect. A skin.changed matching a seed-only baseline is an
intentional apply and repaints; once applied, repeat same-name events stay
no-ops (protects a manual desktop-side theme switch from snap-back, incl.
across a reconnect re-seed).
Follow-up for the salvaged #55800 idle-compaction commit:
- turn_context.py: treat a skipped _compress_context (per-session
compression lock held by another path, failure cooldown, anti-thrash
breaker, codex-native routing) as a strict no-op — only re-baseline
conversation_history and re-anchor current_turn_user_idx after a REAL
compaction. Also re-anchor the user-message index after idle compaction
(the PR predates the reanchor helper).
- hermes_cli/config.py: add idle_compact_after_seconds: 0 to
DEFAULT_CONFIG's compression block (the PR only had
cli-config.yaml.example).
- gateway/run.py: add the idle-compaction status wording to
_TELEGRAM_NOISY_STATUS_RE so the new 💤 message stays out of
human-facing chat surfaces (routine compaction is silent by design);
pin it in tests/gateway/test_telegram_noise_filter.py.
- docs: idle_compact_after_seconds in user-guide/configuration.md and
developer-guide/context-compression-and-caching.md parameter table.
- tests/agent/test_idle_compaction_lock_and_guards.py: end-to-end
coverage with a real AIAgent + SessionDB proving the idle path honors
the per-session compression lock (added after the PR), the persisted
failure cooldown, and the anti-thrash breaker, and that the lock is
released after an idle-triggered compaction.
Salvaged from #55800 by @iso2kx. Implements #27579.
Add compression.threshold_tokens config option that sets an absolute
token cap for auto-compaction. When configured alongside the existing
ratio-based threshold, the effective trigger point is the lower of the
two, so compression never fires later than the user's preferred token
count regardless of which model is active.
This solves the problem where switching between models with different
context windows (e.g. 1M → 400K) shifts the absolute trigger point,
causing premature or delayed compression.
Rework from PR #24279 addressing sweeper feedback:
- The cap is now a first-class compressor configuration value
(threshold_tokens_cap parameter on ContextCompressor.__init__),
not a post-construction patch on the live instance.
- Applied in both __init__ and update_model() so it survives model
switches and fallback activations (the old approach was undone by
update_model() restoring _configured_threshold_percent).
- Clamped to the model's context length so a cap above the window is
a no-op (ratio-based threshold wins).
- Works with max_tokens output-token reservations.
- Added 9 tests covering cap-vs-ratio selection, model switch survival,
context-length clamping, max_tokens interaction, and invalid values.
- Updated user-facing configuration docs.
- Removed unrelated background-review/curator/Honcho changes (main
already contains background-review memory isolation in 973f27e95).
Config example:
compression:
threshold: 0.50
threshold_tokens: 200000 # never compress later than 200K tokens
Integration layer for the cjk_unicode61 tokenizer, rebuilt on the v23
schema (the contributed integration in PR #65544 predated it):
- messages_fts_cjk: external-content FTS5 over a tool-row-excluding view
(same v23 storage discipline as the trigram index it supersedes — zero
inline text copies). Serves EVERY CJK query shape the legacy routing
split between trigram (>=3 chars/token) and LIKE full scans (1-2 char
tokens). Lone 1-char CJK runs and role_filter=['tool'] queries keep
their legacy routes.
- Dedicated marker pair (fts_cjk_rebuild_high_water/progress) gates the
id-scoped triggers, so a cjk-only backfill never gates the complete
messages_fts/trigram triggers.
- Transitions ride (the existing
throttled/resumable chunk engine): fresh DBs are born with the index;
legacy v22 DBs land on v23+cjk in one run; already-optimized v23 DBs
gaining the tokenizer get a marker-gated backfill; live writes are
indexed immediately in every case.
- Tokenizer-loss self-heal: a process that can't load the extension drops
the cjk triggers (writes keep working), leaves a stale breadcrumb, and
the index is rebuilt from scratch on the next optimize run — triggers
are never reinstalled over a gap (external-content 'delete' on an
unindexed rowid is the FTS5 corruption hazard the marker gating exists
to prevent).
- Capability classification: 'no such tokenizer: cjk_unicode61' joins the
degraded-runtime error class everywhere (read probe, write probe,
repair) so tokenizer absence is never misclassified as corruption.
- Config: sessions.cjk_fts (default on, inert without the .so) and
sessions.search_slow_ms in config.yaml, bridged to env by CLI + gateway
(startup + per-turn reload). build.sh falls back to vendored SQLite
headers so no libsqlite3-dev is needed.
Slow-query log path attribution updated: fts_cjk / fts5 / trigram /
like_scan. Tests: 14 lifecycle tests (fresh/legacy/stale/backfill paths,
tokenizer-loss round-trip) + 5 config-bridge tests + slow-log suite.
Follow-up to the salvaged contributor commit, closing the three gaps
flagged in the sweeper review:
1. Init ordering: assign compression.model_thresholds to a selected
plugin context engine BEFORE the initial update_model() call in
agent_init.py, so the initial model's override applies from init
(previously it only took effect after the first /model switch).
Base-class ContextEngine.update_model() now snapshots the
pre-override percent once so repeated switches fall back to the
engine's configured threshold, not a previous model's override.
2. DEFAULT_CONFIG: add compression.model_thresholds (empty map) to
hermes_cli/config.py — additive key, no _config_version bump.
3. Docs: document the key in
website/docs/developer-guide/context-compression-and-caching.md
(yaml example, parameter table, dedicated section) and update the
plugin-boundary note in context-engine-plugin.md to state the
explicit context-engine contract for model_thresholds.
Adds tests/run_agent/test_per_model_threshold_init_ordering.py:
plugin-engine AIAgent init regression (override applies at init,
empty map unchanged), DEFAULT_CONFIG key presence, floor interaction
on the model-switch path (override below the small-context floor is
raised to the floor; above the floor wins), and base-class config
snapshot across repeated switches. Also maps @bennybuoy in
contributors/emails/.
The conversation loop hardcodes max_compression_attempts = 3. Sessions
that legitimately need more rounds are stranded: on a restart history
reload, incompressible tool schemas can keep the per-request estimate
above the compressor threshold even though the message floor compresses
correctly, so three rounds cannot clear it and the turn dies with
"Context length exceeded: max compression attempts (3) reached" — the
same failure class as #62605, where the rough estimate similarly leaves
3 retries short.
Make the cap a config key, compression.max_attempts:
- default 3 = identical to today, so an unset key is behavior-neutral;
- parsed and validated in agent_init alongside the other compression.*
keys (>= 1, hard-capped at 10, non-integer values fall back to 3),
attached as agent.max_compression_attempts;
- the loop reads it via getattr(agent, "max_compression_attempts", 3),
so objects without the attribute keep the prior behavior;
- documented in the DEFAULT_CONFIG compression block.
Tests pin the parse/validate/attach seam: default preserved, custom
value honored, floor and ceiling enforced, garbage tolerated, and the
loop-side getattr degradation.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Rework on top of the salvaged #58133 commits:
- Remove the compression.persist_in_response_store config key — this is
a bug fix (stored transcripts must reflect what the agent will actually
replay), not behavior that should be opt-out-able.
- Drop the per-request load_config() imports the handler-level persist
blocks added.
- Dedupe the two handler-level persist blocks: the compressed-transcript
substitution already lives in _build_response_conversation_history
(via result["_compressed"]), so the handlers only need to propagate
the effective (possibly rotation-changed) session_id. The streaming
path does this via a new session_id_snapshot arg on
_persist_response_snapshot; the non-streaming path picks up
result["session_id"] directly.
- Rotation propagation no longer gates on history-from-store: the first
request in a chain can also rotate, and its stored session_id must be
the child session or the next previous_response_id request resumes the
pre-rotation session and re-compresses every turn.
- Detect when history is loaded from response_store (via previous_response_id)
- Add history_from_store parameter to distinguish history source
- When compression occurs, persist compressed messages instead of original
- Add persist_in_response_store config option (default True)
- Update session_id and response headers to reflect session rotation
Cherry-picked from alidev 2eb816f6b
