Adversarial-review fixes for the #93057 snapshot-compaction PR:
- Fail-closed detachment: only re-enable compression after
bind_session_state successfully severs the engine's parent binding.
A failed rebind keeps the historical compression_enabled=False
behavior and warns, instead of running compaction against a
compressor still bound to the parent's SessionDB (#38727 re-open).
- Warm-cache parity: defer both compression gates (turn-prologue
preflight + pre-API pressure check) until the fork's first provider
response, so the first request replays the full snapshot as the
intended cached read and compaction applies from the second request
on — matching the documented budget mental model.
- Tests: regression for the rebind-failure fail-closed path (red on
pre-fix code) and the existing threshold-crossing test reworked to a
two-request review asserting the warm first request + compacted
second request. 116 tests green across all touched suites; ruff
clean.
Detach the review fork's compressor from the parent SessionDB/session_id
and re-enable in-memory-only compaction for oversized snapshots, instead
of the historical compression_enabled=False guard that left the fork's
replayed transcript unbounded (350k-384k input tokens per request, 1.49M
total across one 8-request review). Add an aggregate input-token budget
(auxiliary.background_review.max_input_tokens, default 600k) so repeated
tool calls cannot recreate an unbounded transcript; the tool loop stops
before the provider call that would cross it.
Closes#93057
Follow-up on top of the salvaged #93335:
- context_compressor._sanitize_tool_pairs now expands alias spellings on
the RESULT side too (tool_result_id_variants), so a composite
call|item-keyed result pairs with its split-field tool_call instead of
being dropped and its call stripped.
- The compressor's _tool_call_id_variants staticmethod and
agent_runtime_helpers' module-level _tool_call_id_variants are now thin
forwarders to agent.message_sanitization.tool_call_id_variants — one
policy owner for alias expansion, so the pre-call sanitizer, repair
pass, dedup pass, and compression sanitizer can never drift apart.
- Preserved the #91768 SDK-object tolerance in repair pass 1 (the
shared helper handles non-dict tool_calls via getattr; the salvaged
commit's isinstance-dict guard was dropped in the merge resolution).
New regression tests: composite-keyed results through
sanitize_api_messages (both directions) and _sanitize_tool_pairs, with
negative controls. Sabotage-verified: compressor test fails with raw
tool_call_id tracking.
Match Responses/Codex tool-call aliases across execution, repair, sanitization, replay, and duplicate handling so valid parallel results are not replaced by unavailable stubs.\n\nFixes #93251
The image-routing vision path calls detect_local_server_type without
the provider's API key. Against a remote API-keyed endpoint (sglang /
vLLM with --api-key) every leg of the 5-request probe waterfall came
back 401 — and because a failed verdict was never written to the
in-memory cache (only positive verdicts were), the waterfall re-ran on
EVERY image-bearing turn (#89863: 51 detail-less busy-acks observed in
one Slack channel while the probe sprayed the user's own server).
Two changes:
- image_routing._should_probe_ollama_vision now takes the API key and
forwards it; a new _resolve_inference_api_key mirrors
_resolve_inference_base_url's resolution order (runtime value,
model.api_key, providers blocks) so the key always matches the URL
being probed.
- detect_local_server_type caches a None verdict in memory with a short
failure TTL (5 min, vs 1h for positives) so the next turn is served
from the negative entry instead of re-running the waterfall — while
a transient failure (server starting, key being fixed) recovers in
minutes. Negative verdicts are deliberately not written to the
cross-process disk cache.
Fixes#89863. With a custom: provider pointing at a remote, API-keyed
endpoint (sglang/vLLM/OpenAI-compat), every image turn triggered a 5-request
probe waterfall without Authorization, spraying 401s at the backend.
Two fixes:
1. _should_probe_ollama_vision now takes api_key and forwards it to
detect_local_server_type so keyed local servers don't 401.
2. When provider != 'ollama', remote endpoints (per is_local_endpoint) are
rejected early — server-fingerprint probing is only valid for local
boxes. Non-Ollama remotes expose Ollama-compat endpoints that can
misidentify and trigger unnecessary /api/show probes.
_lookup_supports_vision resolves the runtime api_key via
_runtime_main_value and forwards it to both helpers. New test class
TestShouldProbeOllamaVision covers the contract in both directions.
Route the max-iterations diagnostic through logging when quiet_mode is active so automation wrappers keep stdout machine-readable.
Add a regression test covering quiet max-iteration summary handling.
/review takes the last 10 chat messages plus optional instructions,
spawns a full-privilege background subagent (the async delegation
rail) that investigates the referenced work (PR, code, docs), and its
complete review re-enters the spawning session as a normal
async-delegation completion the primary agent can act on.
- agent/review_engine.py: shared engine (snapshot, briefing,
auxiliary.review credential resolution, dispatch, note formatting)
- tools/delegate_tool.py: internal credentials_cfg per-call override
(never model-facing) resolved through the same credential system as
delegation.provider pins
- auxiliary.review config block (provider/model/base_url/api_key/
api_mode); provider auto + empty model = inherit the main model
- Surfaces: CLI process_command, gateway run.py dispatch +
slash_commands handler (binds the approval session key so the
completion routes back), TUI/Desktop live dispatch in
tui_gateway/server.py, CommandDef registry (+Slack /hermes-only cap)
- Docs: delegation.md section + slash-commands.md (both tables)
- Tests: 15 engine tests (sabotage-verified: credentials_cfg and
dispatch tests fail without the fix), 4 gateway handler tests
through the real async rail
Follow-up on top of the salvaged cluster: sanitize_api_messages step 3
(duplicate tool_call_id dedup) still tracked only the coalesced
(call_id||id) value in outstanding_call_ids, so after step 2's
variant-aware matching preserved a result keyed on the OTHER id variant,
step 3 deleted it as answering no outstanding call — whole parallel
batches of real results vanished with no stub at all (#93251's total-loss
mode). Track the full variant set per call and consume all siblings when
answered, preserving #58327 duplicate protection and llama.cpp
constant-id re-arm semantics.
Also aligns the #58287 compressor test with the in-flight tool chain
protection (#79278) that landed after that PR was opened: a trailing
user turn keeps the negative-control assistant message out of the
protected trailing window.
New regression tests: divergent-id batch survival through the dedup
pass, sibling-id replay still dropped, constant-id re-arm preserved.
Sabotage-verified: tests fail with the old single-id tracking.
The tool_call id-matching pass in repair_message_sequence only read
`.get("id"/"call_id")` on plain dicts, skipping non-dict tool_calls
entirely (`if not isinstance(tc, dict): continue`). Host-fed and
pre-serialization histories can carry unserialized SDK tool_call
objects (e.g. `ChatCompletionMessageToolCall`) instead of dicts, which
left `known_tool_ids` empty for that assistant turn. The following
`tool` message — a legitimate result already produced by executing the
tool — was then misclassified as an orphan and silently dropped,
corrupting the persisted conversation history and leaving the
assistant's tool_calls unanswered (itself a trigger for HTTP 400 on
strict providers).
Fix: extract id/call_id via getattr() for non-dict entries too,
mirroring AIAgent._get_tool_call_id_static's existing dict-or-object
tolerance, instead of skipping them.
`repair_message_sequence` registers BOTH `id` and `call_id` for each
assistant tool_call, because a matching tool result may be keyed on either
depending on which path built it (#58168). The duplicate guard added for
dropped rather than replayed.
Those two behaviours don't compose: a Codex/Responses tool_call registers
two DIFFERENT ids (`fc_...` and `call_...`), but only the id the first
result referenced is discarded. Its sibling stays in `known_tool_ids`, so a
duplicate result keyed on that sibling still matches and is kept — two tool
messages replayed for one call, which is exactly the HTTP 400 on strict
providers the consume step exists to prevent.
Duplicates of this kind come from the retry / crash / session-resume glitch
the guard was written for; the id-variant split just lets them slip past it.
Track each registered id back to its tool_call's full variant set and
discard all of them on a match. Results keyed on either variant are still
accepted (no false orphaning), and two parallel Codex calls answered via
different variants both survive.
Adds regression tests for the sibling-keyed duplicate and for the
two-calls/mixed-keys case that must NOT be affected.
_sanitize_tool_pairs() matched tool_call/tool_result pairs using a
single-value call_id||id precedence per tool_call (_get_tool_call_id).
In the Codex Responses API format an assistant tool_call carries both a
distinct id (fc_...) and call_id (call_...); a tool result's
tool_call_id may be keyed on either depending on which code path built
it. Whenever a genuinely matching pair used the field the precedence
didn't pick, the sanitizer misclassified it as orphaned on BOTH sides:
it dropped the valid tool result AND stripped the tool_call from the
assistant message, even though neither was orphaned.
Live-verified before the fix: {"id": "fc_777", "call_id": "call_777"}
+ a tool result with tool_call_id="fc_777" (a valid pair) was fully
removed by current main.
Register both id and call_id as valid match keys via a new
_tool_call_id_variants() helper (a set per tool_call, not a single
value), matching #58168's fix for repair_message_sequence's known-id
set today. A tool_call now survives if ANY of its id variants has a
matching result, which is not vulnerable to precedence order at all
(unlike swapping which field is checked first, which only trades which
sub-case is broken).
Note on #56425 (open, unreviewed): that PR touches this same function
for the same underlying issue (#55626) by swapping the call_id||id
precedence to id||call_id. That fixes the specific case where a result
matches `id` but not the reverse case (a result matching `call_id`
while `id` is also present) -- the precedence-swap approach cannot fix
the class, only relocate which sub-case is broken. This fix instead
mirrors the already-merged #58168 pattern (register the superset of
both ids as valid matches), which has no such blind spot. Adds 2
regression tests: the previously-mismatched case, and a negative
control confirming genuine orphans are still stripped alongside a
valid dual-id pair in the same window.
Register every id variant (call_id AND id) of each assistant tool_call in
sanitize_api_messages so a tool result keyed on either variant is treated
as paired. Previously only the coalesced (call_id||id) value was
registered, so Responses-style tool_calls carrying divergent id (fc_...)
and call_id (call_...) had their real results dropped as orphans and
replaced with '[Result unavailable]' stubs.
Cherry-picked from PR #56148 (unrelated busy_ack_templates files dropped
per the author's own follow-up commit).
Follow-up to @BrunoBza's #93062:
1. Set failed=True only for the new repeated_outer_errors exit reason.
Previously the error exit left failed=False, so finalize_turn reported
completed=True for a turn that actually failed — incorrect.
2. Don't append_message the assistant response at the break. A thinking-
prefill or interim assistant may already be the tail, and appending
would create assistant→assistant role-alternation violation.
finalize_turn (lines 341-353) handles this safely by checking
_tail_role != 'assistant' before appending.
3. Update test to assert failed=True and completed=False for the
repeated_outer_errors exit.
The outer conversation-loop except handler only left the loop on a
local-processing error or when api_call_count >= max_iterations - 1.
With the turn budget now unlimited by default (sys.maxsize), a
permanent failure that escaped the inner retry/fallback machinery
retried forever: ~64 retries/s, one core pegged, and the rotated
agent.log history overwritten within minutes.
Bound the loop with a small per-turn cap on total escaping exceptions
(_MAX_OUTER_LOOP_ERRORS = 8, scaled down by a tiny explicit
max_iterations so a manually bounded budget still governs). The legacy
local-processing and near-limit exits are byte-identical; a new
'repeated_outer_errors' exit reason gets a user-facing explanation.
The inner retry/fallback layer owns transient API recovery and
terminates on its own, so only exceptions that escape it reach this
cap - a successful turn is unaffected.
Fixes#92450
PR #93269 (kshitijk4poor) landed the outer-handler break for the same
symptom while this branch was in flight. Keep his guard (it covers
shutdown errors from local post-processing and does the resume-hint +
best-effort persist) and keep this branch's inner-retry-handler return
(it fires BEFORE the ⚠️ retry trace, credential rotation, and fallback
attempts that the outer handler never sees). Point his
_is_interpreter_shutdown_error at tools/interpreter_shutdown.py so the
class has exactly one text-matching site, preserving his RuntimeError
type gate and all 7 of his tests.
When the TUI exits while the post-turn background review fork is still
mid-request, every further API attempt raises 'cannot schedule new
futures after interpreter shutdown'. The conversation loop treated this
as a retryable API error: un-gated ❌ prints leaked onto the user's
shell AFTER the TUI exited (call #4, #5, #6...) and the loop retried a
doomed request until the interpreter froze the thread.
Fix the class, not the site:
- tools/interpreter_shutdown.py: single shared shutdown predicate
(matches both CPython message variants + sys.is_finalizing()).
- cron/scheduler.py, agent/tool_executor.py: existing per-site
predicates now delegate to the shared home (tool_executor previously
matched only the fuller variant).
- agent/conversation_loop.py: inner retry handler recognizes the
shutdown signal and abandons the turn — one log warning, no print,
no traceback, no debug dump, no retry; outer handler gets the same
guard for shutdown errors raised outside the API call.
- The outer handler's bare print() now honors suppress_status_output
(set by the background-review fork) instead of bypassing it.
Refs #55924#58720 (same class in cron delivery), adjacent to #90683.
When the Python interpreter begins teardown (user closes hermes, SIGTERM,
OOM-kill), every executor-backed operation raises 'cannot schedule new
futures after interpreter shutdown'. The outer except handler in
run_conversation caught this error but did not recognize it as fatal —
it kept retrying (API calls #4, #5, #6) until max_iterations, each time
hitting the same dead executor and printing another traceback.
The fix adds an early check: if sys.is_finalizing() or the error matches
the 'cannot schedule new futures' pattern, break immediately with a clean
interpreter_shutdown exit reason instead of retrying. The codebase already
had this pattern in cron/scheduler.py and agent/tool_executor.py — the
conversation loop just wasn't using it.
The 85% compaction autoraise exists to stop wasting the small advertised
272K Codex window. -900k large-context picker variants (#92797) run at
~900K, where the global compression.threshold (default 50%, ~450K) is the
right behavior — autoraising them to 85% (~765K) would delay compaction
far past what the user configured.
- _is_codex_gpt54_or_gpt55() excludes valid -900k variants, so both the
85% override and the one-time autoraise notice skip those sessions.
- Base slugs are unchanged: 272K window + 85% autoraise.
- Tests: variant/base threshold pairs incl. namespaced ids; docs note in
the -900k section.
The demote pass (pass 2) and the retire pass (3.5, #92783) each carried
their own copy of the two image-strip branches. The copies had already
diverged: the retire pass dropped the stale api_content sidecar on
rewrite, the demote pass did not — leaving an exact-wire sidecar that
replay could use to resend the pre-strip image bytes.
Extract _strip_images_from_tool_msg as the single policy owner; both
passes now use it, closing the sidecar gap in the demote path.
Review findings on #92797 (@100yenadmin):
- is_codex_900k_base() is now the single source of truth used by picker
synthesis, context resolution, /model validation, and wire stripping.
Eligibility is an exact table (sol/terra/luna, gpt-5.4, daybreak alias)
plus date-shaped 5.6 snapshots — family-prefix matching removed, so
non-routable -pro slugs and unknown descendants never gain variants.
- strip_codex_context_variant_suffix() strips conditionally: ineligible
aliases (gpt-5.5-900k) are returned unchanged and fail honestly at the
API instead of silently running as the base model at 272K.
- validate_requested_model() rejects ineligible *-900k aliases before the
hidden-slug soft-accept, and accepts valid variants missing from a
stale catalog without letting the typo auto-corrector eat the suffix.
- Codex context resolver drops vendor/ namespaces, so
openai/gpt-5.6-sol-900k resolves to 900K like the bare id.
- Table-driven regression covering eligible bases/snapshots/namespaced
ids and rejected -pro/-mini/5.5/unknown aliases, asserting context AND
wire model.
The Aug 16 change that auto-raised gpt-5.4/5.6 Codex OAuth context to the
live-verified 900K burned through subscription usage for users who never
asked for the larger window (bigger window = more input tokens per request).
- Base Codex slugs (gpt-5.6-sol/terra/luna, gpt-5.4) now resolve to the
advertised 272K again — the cheaper limit is the default.
- The model picker synthesizes explicit <slug>-900k variants (e.g.
gpt-5.6-sol-900k) for every live-verified slug; selecting one opts into
the 900K window. Slugs that genuinely enforce 272K (gpt-5.5,
gpt-5.4-mini) get no variant.
- The -900k suffix is Hermes-side only: stripped before the model id hits
the wire (main transport + auxiliary Responses adapter), and pricing
aliases the variants onto the base entries.
- Docs: new opt-in section in context-compression-and-caching.md.
Review-pass follow-up on the load-time durability stamp:
- hermes_state.py: import the marker from agent.context_compressor instead
of a third synced literal (hermes_state already imports agent.* at module
level; only run_agent is circular). Old comment claimed otherwise.
- agent/turn_finalizer.py: replace the raw "_db_persisted" string at the
fill-empty-tail pop site with the shared constant (was outside the drift
guard).
- agent/conversation_compression.py: the no-op progress check now falls back
to a marker-insensitive comparison (_strip_marker_for_comparison). Loaded
rows are stamped at materialization time while compress() output is
marker-swept, so a semantically-identical no-op copy on a cold-resumed
session would previously compare unequal and take the progress branch.
Raw == still runs first so engine-returned list subclasses keep their
__eq__ semantics.
- test_marker_constant_in_sync extended to turn_finalizer + identity
assertions; new test_noop_progress_check_is_marker_insensitive
(mutation-checked: fails when the helper is neutered).
browser_vision's native fast path base64-encoded screenshots at full
resolution and baked them into the tool result uncapped — the exact
sibling of the vision_analyze path #92699 fixed. Apply the same
proactive 256KB/1568px resize before the embed enters reusable history.
Fail-open by design: without Pillow the resize helper falls back to raw
bytes and the compressor's keep-newest pass still retires stale embeds.
Sibling-gap follow-up for the #92725 salvage; the shared-cap approach
mirrors the policy-owner idea from #92748.
Co-authored-by: joaomarcos <joaomarcosdias444@gmail.com>
Images locked in protect_last_n never shrank, so compression savings
stayed under 10% and anti-thrash disabled further compaction. Keep the
newest three tool-result screenshots live for follow-up QA and replace
older native embeds with placeholders.
Address review feedback: the gate reused has_vertex_credentials(), which also
returns True for an ambient GOOGLE_APPLICATION_CREDENTIALS path. That var is
commonly set globally for unrelated GCP work, so a user who never configured
Hermes for Vertex would see it in the explicit-only picker and could spend
against those credentials — weakening the explicit-configuration guarantee the
gate documents (mirrors the existing _IMPLICIT_ENV_VARS carve-out).
Add has_explicit_vertex_config() in agent/vertex_adapter.py that checks only
Hermes-scoped signals — VERTEX_PROJECT_ID / vertex.project_id (project
override) or a resolvable VERTEX_CREDENTIALS_PATH — and NOT
GOOGLE_APPLICATION_CREDENTIALS. Route the auth gate through it.
Adds a regression test asserting an ambient GOOGLE_APPLICATION_CREDENTIALS
path alone does not mark Vertex explicit, and updates the existing test to
drive the real config signal instead of mocking has_vertex_credentials().
- Drop the 'aborted before its tail' no-op sentence: early aborts are
intercepted by the aborted/no-progress branches and never reach the
would-grow check, so the framing overstated its relevance (2c finding).
- Test now also asserts the durable model_config copy still holds the
armed runway after the refusal — locking in the memory==disk half of
the contract, not just the in-memory value.
compress()'s successful tail zeroes _proactive_prune_rearm_tokens in
memory — correct for a committed compaction, whose boundary already broke
the prompt-cache prefix. But compress_context's anti-growth guard can then
REFUSE the result and keep the original transcript, whose cached prefix is
intact. The refusal returned with the in-memory runway still at 0 while the
durable model_config copy kept the old value, so:
- the next eligible iteration's proactive prune fired without the regrowth
interval #79640 introduced — an immediate, unthrottled cache-breaking
rewrite (#91830's bug class), and
- memory and disk disagreed until a restart silently re-armed the throttle
from the stale durable row.
The refusal branch now restores the runway from the attempt snapshot — the
same targeted restore the rotation-failure rollback already performs.
Sibling non-commit branches audited: aborted (returns before the tail
zero), no-progress (tail zero only runs after a real boundary rewrite,
which no-progress by definition lacks), empty-transcript (built-in tail
never returns []), fence-denied (full snapshot restore already covers the
runway), in-place DB failure (in-memory transcript keeps the compacted
form, so the zeroed runway is consistent with it).
Fixes the reachable half of the structural asymmetry flagged in #91830.
Mirror the credits_tracker caveat in _is_free_model (a paid stealth/
model would bypass the free_only gate and paid-lane warning) and fix
the stale _warn_paid_lane_once docstring.
- credits_tracker: trim inline comment block (duplicated docstring) and
correct its safety claim - a paid model under stealth/ would fail
closed (suppressed banner), not open; state the trade-off honestly.
- run_agent: update stale call-site comment to mention stealth/ prefix.
- auxiliary_client: widen sibling free-SKU detector _is_free_model to
recognize stealth/ prefix (same bug class as #91843: free_only=true
wrongly skipped the OpenRouter fallback and the paid-lane warning
fired spuriously for stealth models).
- tests: bind the new sibling behavior (stealth/ox-alpha free,
my-stealth/model not).
Stealth-preview SKUs (e.g. stealth/ox-alpha) are free-tier but carry no
:free suffix, so is_free_tier_model() returned False for them. On gateway
sessions (which never run the model picker's pricing fetch), the free-model
suppression of the credits.depleted banner never engaged, and any response
carrying paid_access:false triggered a false "Credit access paused" notice.
Add stealth/ prefix detection to is_free_tier_model() as a zero-network
signal, same design as the existing :free suffix check. Fail-open to
False (banner still shows) if the prefix changes — recoverable noise,
never a masked depletion on a paid model.
Closes#91843
Addresses @helix4u's review on #91493:
- conversation_loop now stamps failure_retryable (the real ClassifiedError
verdict) next to failure_reason; error_surface prefers it and only falls
back to the reason set for older results. Fallback set corrected to match
classify_api_error (auth, format_error, billing_unverified now
non-retryable).
- The descriptor carries the failing session's provider/model captured at
classification time; Copy error details prefers them over the foreground
composer atoms.
- Open logs is labeled 'Open Desktop logs' on remote/cloud connections —
the local folder holds transport logs, not the remote runtime's.
- API-exception module allowlist widened to botocore/boto3/google/grpc/
requests/aiohttp so other adapter SDKs don't misclassify as gateway.
Turn errors now carry a structured {layer, code, retryable} descriptor
(agent/error_surface.py) built from the same classifier the retry loop
uses. The tui_gateway stamps it on terminal error frames, retained
failed-turn snapshots, and resume replay; the Desktop error card renders
the layer title (provider / endpoint / streaming / auth / billing /
gateway / runtime / disk) plus matched actions: Retry, Switch provider,
Open logs, Copy diagnostics.
Older backends that omit the descriptor keep today's behavior (generic
title, string-sniff fallbacks) — the field is advisory on both sides.
Bot Mode agents now DM teammates through a real tool instead of
hand-assembled shell commands. message_agent(target, message) validates
the target against the live roster, applies the sender's attribution
prefix server-side, and delivers over the existing proven transports
(hermes -p ... --query-file for local teammates, hermes peer dm for
peer gateways) as a tracked background process with notify-on-complete
— fire-and-forget, the reply wakes the sender on a later turn.
Containment: the schema is injected per-turn ONLY into a bot's
canonical 'Bot Chat' session on Bot-Mode-managed installs (same gate as
the protocol section); it is never registered in the tool registry or
any toolset, and dispatch re-gates on the session title so a forged
call from any other session refuses. The gate is session-stable, so the
tool list stays byte-identical across turns (prompt-cache safe).
The protocol section is rewritten to teach the tool and now carries the
teammate roster WITH ROLES (Bot Mode title + profile description), so
bots know who does what before picking a recipient. Roles and a
protocol version salt join the capability fingerprint: existing eternal
Bot Chats adopt the v2 protocol + tool with one epoch refresh, and a
rename/description edit refreshes the roster on the next message.
Follow-up structural pass on the review fix:
- Runtime provider, auxiliary resolution, model validation
(hermes_cli/models.py), live discovery (bedrock_model_ids_or_none),
and the Mantle URL/SigV4 fallbacks all resolve their region through
resolve_bedrock_runtime_region() — one canonical implementation of the
config-first priority instead of three hand-rolled copies.
- agent_init: drop the 'if "client_kwargs" in locals()' guard by
initializing client_kwargs unconditionally at the top of the else
branch; the Mantle kwargs hook is a documented no-op for non-Mantle
base URLs.
Address review feedback on #65076:
- Add resolve_bedrock_runtime_region() to agent/bedrock_adapter.py: the
config-first region resolution (bedrock.region in config.yaml, then
AWS_REGION/AWS_DEFAULT_REGION/botocore profile/us-east-1) that the main
runtime resolver uses, exposed as a shared helper.
- Switch auxiliary client resolution (agent/auxiliary_client.py aws_sdk
branch) to the new helper. Previously it derived its region with bare
resolve_bedrock_region() (env-first), so when config.yaml pinned
bedrock.region to a different region than the ambient AWS env, auxiliary
calls (compression, memory, vision) left the primary runtime's region.
Both the AnthropicBedrock/Converse path and the new Mantle OpenAI
Responses path now resolve identically to the main runtime.
- Add regression tests covering the bedrock.region-vs-AWS_REGION mismatch
for both the Claude auxiliary path and the Mantle auxiliary path.
- Update website/docs/guides/aws-bedrock.md: the guide claimed Hermes never
uses the OpenAI-compatible endpoint, which the Mantle route made stale.
Document the triple routing (AnthropicBedrock / Mantle OpenAI Responses /
Converse), the Mantle auth model (bearer token or SigV4), and add the
GPT-5.5/5.6 model IDs to the models table.
GPT-5.6 Sol, Terra, and Luna went GA on Amazon Bedrock on 2026-07-13.
Like GPT-5.5, they are served exclusively from the Bedrock Mantle
OpenAI-compatible Responses endpoint (the model cards list
bedrock-runtime/Converse as unsupported), so they ride the allowlist
routing introduced for GPT-5.5:
- Add openai.gpt-5.6-{sol,terra,luna} to BEDROCK_OPENAI_RESPONSES_MODEL_IDS
so runtime resolution, auxiliary calls, and MoA slots all take the
SigV4/bearer Mantle Responses path.
- Surface the family in the curated Bedrock picker list.
- Record the 272K context window from the AWS model cards for all four
Mantle OpenAI models (previously fell back to the 128K default).
- Generalize picker tests from the hardcoded single-model checks to the
BEDROCK_OPENAI_RESPONSES_MODEL_IDS allowlist so future Mantle model
additions do not require test surgery; add routing, picker, and
context-length coverage for the 5.6 family.
Docs: https://docs.aws.amazon.com/bedrock/latest/userguide/model-cards-openai.html
Route Bedrock-hosted OpenAI GPT-5.5 through the Bedrock Mantle OpenAI Responses endpoint with SigV4 request signing. Keep native Bedrock Converse and Claude Bedrock routing unchanged, and add picker/runtime regression coverage.
GLM-5.3 accepts a graded low/medium/high/max reasoning_effort scale
(verified live in #91789: monotonic reasoning-token scaling, no 400s),
but the effort mapper reused GLM-5.2's two-level vocabulary, silently
rewriting low/medium to high. Adds GLM53_EFFORTS/GLM53_OVERRIDES and a
per-model vocabulary pick in the zai plugin; 5.2 keeps its high/max
clamp. Closes#91789. Also covers the gap noted when closing #86947
(credit @santhanakrishnan-d and @terje1965 for the graded-scale finding).
GLM-5.3 is live on api.z.ai (coding plan endpoint) but had no entries in
Hermes, so it silently fell back to the generic 202K GLM context —
triggering premature context compression on a 1M-window model.
- model_metadata: 'glm-5.3': 1_048_576 (same base model as 5.2; 1M
context / 128K max output per docs.z.ai/guides/llm/glm-5.3, verified
2026-08-14)
- auth: add glm-5.3 to coding-plan probe lists (global + CN)
- models: add glm-5.3 to picker/model lists (6 sites)
- zai provider: reasoning_effort mapping covers glm-5.3 (accepted live
by the endpoint, HTTP 200)
Identify a completed merged assistant handoff from the carrier's own stop state instead of an unrelated adjacent history row. Keep carriers with pending tool calls actionable so compaction cannot abort a live tool chain.
Treat a merged assistant-role summary carrier as the driving reference handoff when it immediately follows a completed assistant stop. Its preserved prose and stale tool_calls are assistant continuity, not a fresh live user request.
Keep legitimate in-flight behavior unchanged when there is no completed stop, a real user turn follows, or a distinct later assistant tool-call row continues the loop.
Extends the #80622 active-turn guard for the merged-carrier shape reported under #42768.
The Codex OAuth backend (chatgpt.com/backend-api/codex) intermittently
injects prompt_cache_retention into its own upstream call and then rejects
it, returning HTTP 400 invalid_parameter. Hermes never sends that field on
this route (see agent/transports/codex.py::_default_prompt_cache_retention_
for_request, which only sets it for api.meta.ai and bedrock-mantle hosts).
Reproduced live: a minimal 1-message request carrying no cache parameters
at all failed 4/20 (20%) with this error, so the rejection is not
deterministic and retrying the identical request is the correct recovery.
Previously the catch-all in _classify_400 returned format_error/
retryable=False, which tripped the is_client_error abort gate in
conversation_loop and killed the turn on the first attempt - burning an
entire large-context request (~550k tokens) per failure.
Classify these as retryable server_error (should_compress=False - the
request shape was never the problem). The same guard is applied to the
sibling 5xx request-validation branch, where a fronting proxy can surface
the identical rejection.
Deliberately narrow: keyed on parameters we only send on specific routes,
and skipped when the current provider is one that legitimately sends them,
so a genuine client-side bad parameter (max_tokens on GPT-5) still fails
fast as a format_error.
prune_pre_checkpoint_items() had a hardcoded role=='user' filter that
discarded all non-user messages before a checkpoint — including Hermes'
own compression summaries (role='assistant'), causing total context amnesia
about past conversation summaries.
The fix:
- _is_summary_item delegates to the canonical
agent.context_compressor.is_compaction_summary_message provenance check
(not an ad-hoc heuristic)
- Summaries are retained whole (never byte-sliced) within a 32k token budget
- Idempotent across repeated checkpoints (dedup by identical text)
- _chat_messages_to_responses_input threads item_sources (raw chat messages)
through to the pruner, so it can read summary content directly from the
source when the Responses conversion shape is lossy (tool-result carrier
becomes function_call_output, or stale codex_message_items replay shadows
merged content)
Fixes#90975.
Salvage of #90976 by @JoaoMarcos44.