Review follow-up (egilewski): the previous commit only hedged the guidance
text; the exact Anthropic 400 was still classified, persisted, and surfaced
as confirmed billing exhaustion. Carry the ambiguity all the way through:
- agent/error_classifier.py: 'out of extra usage' matches on the 400 and
status-less paths now attach error_context {billing_unverified,
possible_content_filter}. Reason stays FailoverReason.billing (rotation +
fallback remain the right recovery either way); ClassifiedError grows a
billing_unverified property.
- agent/credential_pool.py: new FAILURE_REASON_BILLING_UNVERIFIED. An
unverified billing exhaustion gets the short transient cooldown instead of
the one-hour bench, regardless of pool size: a content-filter rejection
leaves the credential healthy and fails identically on every key, and the
hour-long sole-credential latch is what replayed the stored error and made
real fixes look ineffective. A true 402 keeps the full bench. The marker
persists with the entry so a restart cannot upgrade it back to a bench.
- agent/agent_runtime_helpers.py + run_agent.py: recover_with_credential_pool
threads billing_unverified and hands the pool 'billing_unverified' as the
persisted failure_reason.
- agent/conversation_loop.py: the fallback-switch status, max-retries status,
terminal label, and both structured terminal results hedge when the verdict
is unverified. New _billing_terminal_label + _billing_failure_result build
the returned terminal response in one place; the result dict now carries
billing_unverified and the billing_block gains 'unverified': true. The
confirmed-billing path (a real 402 or an API-key credit depletion) keeps
the original assertive wording, so the caveat no longer dilutes it.
Regression tests: classifier marking (400 + status-less + unambiguous-body
negative), pool cooldown TTLs + persistence round-trip, pool failure_reason
plumbing, and the returned terminal response for both unverified and
confirmed verdicts.
Note: tests/agent/test_credential_pool_routing.py::TestFailureAttribution::
test_unmatched_key_does_not_retry_only_pool_entry fails identically on
current main without this change (pre-existing, unrelated).
On an Anthropic subscription OAuth credential, every request failed with
HTTP 400 "You're out of extra usage. Add more at claude.ai/settings/usage".
That is not a billing condition: Anthropic's server-side content filter rejects
the first sentence of Hermes' own built-in SKILLS_GUIDANCE prompt, and the
rejection is surfaced with a billing-shaped message. Because the message points
at the usage settings page, it reliably sends people to buy quota they do not
need — the reporter lost three debugging sessions to it.
Bisected against the live API with the real 71,721-char assembled prompt: the
first SKILLS_GUIDANCE sentence alone reproduces the 400 and removing it alone
clears it. Size was ruled out (20 KB of unrelated filler returns 200) and so was
the system[0] identity gate (that returns 429, a different failure).
Three changes, all serving the same outcome — a subscription user can no longer
be misdirected by this 400:
- agent/prompt_builder.py: reword the triggering sentence to the phrasing the
reporter verified returns 200. Meaning, the skill_manage reference, and the
## Skill Safety Rule block are all preserved. The reword is empirically
validated rather than understood, so a comment records the bisect and warns
that any rewrite must be re-verified against an OAuth token, not an API key.
- agent/conversation_loop.py: the Anthropic branch of the billing guidance no
longer asserts exhaustion as fact. It hedges the opening line, names the
content-filter alternative, and gives the operator a way to tell the two apart
(if the usage page still shows quota, suspect a content rejection). It also
points at `hermes auth reset anthropic`, because the credential exhaustion
latch replays the stored error for ~60 min without issuing a request — which
makes a real fix look like it did not work.
- hermes_cli/auth.py: document that CLAUDE_CODE_OAUTH_TOKEN is an OAuth token,
not an API key, despite auth_type="api_key". It stays in api_key_env_vars
because that tuple doubles as the credential-discovery list; removing it would
stop Hermes finding a `claude setup-token` credential at all.
Docs updated to match the reworded prompt.
Fixes#82154
The gateway rotation guard (#83339) only protects the rotate path, but
in-place compaction commits inside compress_context() via
archive_and_compact — before the gateway can inspect the result. Add the
anti-growth check at the commit site so both paths are covered: a
compression whose rough output exceeds its input is a strict no-op
(original transcript kept durable, session identity untouched).
Covers the observed failure where session hygiene persisted 426 -> 426
messages and ~379K -> ~688K tokens.
SessionDB could leave native SQLite handles open when construction failed
partway through schema/pragma/FTS/repair/lock/interrupt handling. Other
short-lived callers (MCP reads/polling, session search, reactions, trace
upload, insights, shutdown recovery) opened temporary SessionDB handles
without a complete ownership boundary. API-server profile caches and
RetainDB shutdown had similar late-close races. Under sustained load this
exhausted file descriptors (EMFILE).
- Close partially initialized SessionDB connections on every constructor
exception path via a finally block guarded by an initialization-complete
flag.
- Close temporary/cross-profile SessionDB handles in finally blocks across
CLI, MCP, search, trace, reactions, insights, and recovery paths.
- Add API-server per-profile cache ownership and disconnect cleanup.
- Make RetainDB writer-queue shutdown exception-safe: track connections per
thread, close on worker exit, reject new enqueues after shutdown starts,
and sweep any connections left by short-lived threads.
- Add regression coverage for constructor failures, worker-thread readers,
API disconnect failures, shutdown recovery, RetainDB late enqueue, and
foreign-loop async clients.
Salvage notes: the original PR's per-thread WAL-reader ownership changes
were superseded by main's read-connection pool (permits + checkout/return);
its cron timeout-abandon fix is credited separately to #72822's earlier
identical fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
A turn writing against a session already closed by compression died with
session_persistence_failed and a misleading "this is often a full disk"
dialog, even though the store was healthy and a live continuation existed
(#82001). Depth-1 recovery (find_live_compression_child) could not resolve
lineages with >=2 compression hops (root -> mid -> tip), reproduced
independently on two- and three-hop chains.
- run_agent.py flush chokepoint: on CompressionSessionClosedError, resolve
tip = db.get_compression_tip(old_id) (canonical bounded transitive walk),
adopt only when tip != old_id AND the tip row is live, retry the flush
exactly once (adoption budget); otherwise fail closed.
- gateway/session.py append_to_transcript: replace the depth-1 live-child
lookup with the same tip + liveness contract, so gateway transcript
reroutes follow full chains.
- agent/conversation_compression.py _adopt_live_compression_child: turn-start
recovery preflight now resolves via get_compression_tip with the same
liveness check, closing the last depth-1 consumer in this family.
- classify_persistence_error: new "compression_closed" bucket; the turn-end
explanation names compression rotation and tells the client to refresh the
session id instead of blaming a full disk.
Tests: depth-1 adoption, multi-hop chain adoption (agent + gateway), fail
closed with no continuation / stale-closed (ws_orphan_reap) tip, exactly-once
adoption budget, and error-wording guards (compression-closed never mentions
disk; real disk failures keep disk guidance).
Closes#82001
Co-authored-by: Al3xand3r1987 <125030427+Al3xand3r1987@users.noreply.github.com>
Co-authored-by: yuzilongleif-collab <235949691+yuzilongleif-collab@users.noreply.github.com>
The unconditional pop from the previous commit also stripped the key when
agent.client was still the real MoA facade, forcing the facade to re-prepare
from scratch — a duplicate reference fan-out per turn (caught by
test_moa_virtual_provider_aggregator_is_actor). Gate the defensive strip on
the same prepare()-capability probe used at the injection site so the
handshake survives on the facade while a swapped-in native client is still
protected.
After a client replacement (credential rotation, dead-connection cleanup,
or fallback+restore), agent.client may become a native OpenAI client
while agent.provider stays "moa". The _moa_prepared_request key was
passed through to the native SDK, causing TypeError on every turn.
Pop the key at the dispatch point (chat_completion_helpers.py:509).
The MoAClient facade already handles a missing key by falling through
to its normal resolution path.
Closes#78382
`_moa_prepared_request` is a private handshake between the conversation
loop and MoAChatCompletions.create. It is attached whenever
agent.provider == "moa", on the assumption that agent.client is still the
in-process MoA facade.
Credential rotation, provider fallback and dead-connection cleanup all
rebuild agent.client from _client_kwargs between attempts, and
pending_moa_prepared_request deliberately carries a prepared request
across exactly that boundary. The rebuilt client is a native OpenAI
client while provider stays "moa", so the key reaches an SDK that has
never heard of it:
TypeError: Completions.create() got an unexpected keyword argument
'_moa_prepared_request'
That error is non-retryable, so every remaining turn on the session
fails. Both dispatch paths are affected: the non-streaming one calls
agent.client directly, and _create_request_openai_client returns
agent.client unchanged for provider "moa".
Re-check the live client at the point the key is attached, which covers
both paths at once. When the facade is gone, send the prepared prompt
without the handshake and log the downgrade.
When agent.provider == "moa", the MoAClient facade *is* the client - there is
no real OpenAI wire endpoint behind the moa://local placeholder. Client rebuilds
(_replace_primary_openai_client: stream-retry pool cleanup, credential rotation,
dead-connection cleanup, fallback+restore) go through create_openai_client and
produce a native OpenAI client while provider stays "moa". The next primary
call then either raises a `_moa_prepared_request` TypeError (#78382) or, when
_client_kwargs carry an unrelated relay base_url, leaks the request to a foreign
gateway (observed as HTTP 503 "group ... no available channel" from an
unrelated new-api relay right after an aggregator empty-stream retry).
Fix: in create_openai_client, when provider is "moa", return
build_moa_facade(agent, model) instead of a native client. This covers every
rebuild entry point. The three already-fixed call sites
(restore_primary_runtime, try_recover_primary_transport, switch_model) assign
the facade directly and do not go through create_openai_client, so they are
unaffected.
Closes#78382
Residual from PR #70910 after #77509 landed the message-list scrub: a
whitespace-only system content block carrying a cache_control marker
still reached the wire and 400'd the whole request ("text content
blocks must contain non-whitespace text"), wedging the session on every
retry. The block cannot be dropped (it carries the cache breakpoint),
so coerce its text to the shared non-whitespace placeholder when
extracting the system param, copying the block so caller message dicts
are never mutated.
Adds SHL0MS's request-level regression suite from #70910; four of its
five cases already pass on main via #77509 — the system-block case
fails without this fix.
Strict OpenAI-compatible providers (onerouter / Qwen, DeepSeek v4) reject
an assistant message carrying tool_calls: [] (or null) with HTTP 400
'Empty tool_calls is not supported in message.'
The pre-API sanitizer in agent_runtime_helpers.sanitize_api_messages already
drops these on the conversation_loop path, but auxiliary / custom-provider
routes that bypass that sanitizer can still reach the wire with an invalid
empty array and abort the whole session (non-retryable 400).
Normalize at the transport layer too: detect an empty-list / null
tool_calls on assistant messages, strip the key on the per-call copy (never
mutate the stored history), and keep real tool_calls untouched. Includes
unit tests covering empty-list, null, real-call preservation, mixed batches,
user-role non-mutation, copy-on-write, and cross-provider parity.
Follow-up to #58755.
The dedup pass in sanitize_api_messages (introduced by #58327) can
produce an empty tool_calls array when all tool_call_ids in a message
are duplicates of earlier messages in a long conversation history.
DeepSeek v4 and newer OpenAI reject empty tool_calls with HTTP 400:
'Invalid messages[N].tool_calls: empty array'.
When kept_tcs is empty after dedup, drop the tool_calls key entirely
instead of writing tool_calls: [].
Fixes#64335
The gateway turn-inactivity watchdog (gateway/run.py::_watch_gateway_turn_inactivity)
abandons a turn once seconds_since_activity exceeds the inactivity timeout
(default 30 min). Activity was only stamped when a tool started and when it
completed, so a tool call that runs silently for 30+ minutes (quiet builds,
long pytest suites, large downloads, network waits with no output) froze the
clock and the watchdog hard-abandoned a turn that was still making progress,
reaping the tool's processes mid-execution (issue #84491).
Add a daemon-thread heartbeat inside _run_agent_tool_execution_middleware that
touches agent._touch_activity every 30s while the tool is in flight, until the
call returns. Both the sequential and concurrent execution paths funnel through
this single middleware, so one heartbeat covers every tool. The thread is
stopped in a try/finally so it always tears down even if execute() raises, and
is never started when a guardrail/authorization block short-circuits before
dispatch. A genuinely hung tool remains bounded by the tool layer's own
timeouts (terminal default 180s, concurrent batch deadline ~420s), so the
heartbeat only extends the turn's life while the call is legitimately running.
Verified by an independent reviewer (no security/logic defects); 5 unit/integration
tests pass on Python 3.12 (upstream CI). The 30-min gateway backstop remains
for turns whose agent loop itself stalls.
The keepalive httpx client uses read=None, and stranger-thread abort cannot close FDs, so a DeepSeek stall on the cron inline path waited until TCP died — hours past the 600s watchdog. Inject a per-call read timeout matching the stale budget, walk in-flight pool requests, and clear the socket timeout before shutdown without releasing the FD.
Port of the bug class from earendil-works/pi#7933 (DeepSeek base-URL
detection matched by raw substring, missing case variants and matching
lookalike URLs). Hermes had the same class at five sites:
- cli_agent_setup_mixin.py: keyless-custom-endpoint detection treated any
URL containing the OpenRouter host substring (path segment, lookalike
domain) as OpenRouter, and missed case variants of the real host.
- models.py validate_requested_model: same substring check for routing an
openrouter provider with a custom base_url to the custom catalog.
- runtime_provider.py: local-endpoint autodetect matched the string
localhost anywhere in the URL, including remote hostnames containing it.
- gateway/run.py: /status endpoint display, same local-host substring.
- agent_runtime_helpers.py: Nous Portal cache-layout detection matched
the nousresearch substring anywhere in the URL.
All sites now use the existing base_url_host_matches / base_url_hostname
helpers (exact host or subdomain, case-insensitive). Regression tests
proven to fail against the old predicates.
Some OpenAI-compatible providers (DeepInfra, etc.) return validation
errors as in-stream SSE chunks: HTTP 200 with choices=None and
error_type/error_message in model_extra. The streaming loop silently
dropped these chunks, causing EmptyStreamError ("empty stream") and
pointless retries on the same bad request.
Fix: check for error_type/error_message on chunks with no choices
before skipping. When found, raise RuntimeError with the provider's
error message so the error classifier can properly handle it.
Fixes#65631
Per-provider ssl_ca_cert / ssl_verify reached the httpx chat client and the
auxiliary clients (#56681), but the endpoint discovery and pricing probes did
not. Both probe families resolved TLS from process-wide env vars only:
- the requests-based metadata/pricing probe
(agent/model_metadata.py::_resolve_requests_verify)
- the urllib-based /models catalog probe
(hermes_cli/models.py::probe_api_models)
A custom endpoint whose chain verifies against the provider's configured
bundle, but not the process SSL_CERT_FILE, then logged a spurious
CERTIFICATE_VERIFY_FAILED on every probe even though the chat path worked.
Pointing a global CA env var at the bundle fixes it but changes verification
for every provider, defeating the point of a per-provider setting.
This threads the selected provider's TLS settings into both probe paths,
reusing get_custom_provider_tls_settings so there is no second precedence
chain:
- _resolve_requests_verify(base_url) looks up the provider's ssl_verify /
ssl_ca_cert before falling back to the env vars. Callers with no base_url
keep the exact env-only behavior.
- probe_api_models builds an ssl.SSLContext from the provider settings and
passes it through open_credentialed_url, which gains an ssl_context seam on
the cloned secure opener. Unmatched or public endpoints pass None and keep
urllib's default policy.
Tests: tests/agent/test_custom_provider_ca_probes.py covers both probe
families (provider CA, ssl_verify:false, unmatched, missing file, config
lookup failure) plus end-to-end assertions that the resolved verify value and
SSLContext actually reach the request seam. Verified against the neighboring
metadata, pricing, TLS, and urllib-security suites (266 tests) with no
regressions.
Refresh-loss interrupt is cooperative, so a stalled writer could still flush after another process reclaimed the conversation. Carry the holder into append_message / append_messages_batch and reject the write in the same SQLite transaction when the lease row is missing, expired, or owned by someone else.
The messaging gateway multiplexes profiles over ONE shared launch-home
state.db, binding the profile per turn via the HERMES_HOME ContextVar
(copy_context into the worker thread). _agent_home derived the home from
db_path unconditionally, so on that lane the launch home stomped the
correctly-bound profile — deterministically inverting the leak #86313
fixed (found by @kshitijk4poor's post-merge probe; @helix4u flagged the
plugin-metadata half).
- _agent_home: bound override wins; session_db home is the unbound-thread
fallback
- _plugin_session_info: profile_name derives from _agent_home too
- full-prompt wiring regression (SOUL + skills + profile line on a bare
thread with the bot's DB) — reverting any call-site wire fails it;
multiplex, bare-thread, and plugin-metadata cases each pinned;
sabotage-verified both new tests fail against the merged behavior
- skills LRU cap 8 -> 32 (key is now per-profile x platform)
load_soul_md resolved the home ambiently, so a build thread that lost the
HERMES_HOME ContextVar read the launch profile's SOUL.md into another
profile's prompt — same class as the skills-index leak fixed in #86313.
load_soul_md and build_context_files_prompt now accept home_override, and
build_system_prompt_parts passes the agent's own home (from session_db)
at both SOUL call sites. Ambient behavior unchanged when no override.
The named-profile branch of the "Active Hermes profile" hint built its
paths by appending `/profiles/{active_profile}` to `get_hermes_home()`.
But `_resolve_active_profile_name()` returns a non-default name *only*
when `get_hermes_home()` has already resolved under `<root>/profiles/`
— that is how it derives the name in the first place. Both scoping
mechanisms (a `HERMES_HOME=<root>/profiles/<name>` env var and the
multiplexer's `set_hermes_home_override` contextvar) satisfy that, so
the suffix always doubled.
The same branch used `get_hermes_home()` for the *default* profile's
data pointers, where the root was intended — placing them inside the
active profile.
On a real 4-profile install the hint rendered:
reads and writes ~/.hermes/profiles/via/profiles/via/
default profile's data lives at ~/.hermes/profiles/via/skills/
against actual paths of `~/.hermes/profiles/via/` and `~/.hermes/skills/`.
Use the session home directly as the profile home, and
`get_default_hermes_root()` for the root pointers. Every named-profile
session was shipping a prompt that named nonexistent directories and
mislabeled this profile's own skills/plugins/cron/memories as the
default profile's — the exact cross-profile confusion the hint and
`classify_cross_profile_target` exist to prevent.
The default-profile branch is unchanged.
Fixes#72894
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The root-derivation for the profile-hint text now only engages when the
agent's own home is known; otherwise the line resolves exactly as before
(via this module's get_hermes_home), fixing
test_coding_prompt_preserves_legacy_workspace_order.
On a correctly bound profile session get_hermes_home() returns the profile
dir itself, so relative_to(home/'profiles') never matched and every profile
misreported as 'default' (with wrong paths in the profile hint text). Use
get_default_hermes_root() for both the name derivation and the default-data
root string. Adds regression tests for the bound-profile and root-home cases;
sabotage-verified the bound-profile test fails against the old resolution.
A bot profile's system prompt could list the DEFAULT profile's ~80
skills and print 'Active Hermes profile: default' — while the live
skills_list() correctly showed the bot's real (often empty) set. The
agent plans against that index, so a false inventory makes it claim
capabilities it doesn't have, waste context tokens, and lose trust.
Root cause (confirmed empirically): the skills-prompt builder and the
active-profile line resolve the home through get_hermes_home(), which
reads a HERMES_HOME ContextVar. ContextVars do NOT propagate into
threading.Thread, so an agent build running on a thread that didn't
bind the profile's home falls back to the launch (default) home and
builds default's index. A bare no-override thread builds default's
full 7621-char block; the same thread with the fix builds empty.
Fix: resolve the agent's OWN home from its dedicated _session_db.db_path
(ground truth, ContextVar-independent) and pass it explicitly:
- build_skills_system_prompt(skills_dir_override=...) scopes the index,
the disk snapshot, and external-dir resolution to that home
- the active-profile line derives the profile name from the same home
Both fall back to ambient resolution when no db is present, so the CLI
and default-profile paths are unchanged.
Regression tests: an empty bot profile yields an empty skills block on
a bare thread even with ambient HERMES_HOME bound to a skills-rich
default; agent-home resolution from session_db.db_path. 3/3.
Follow-up on the #84795 salvage: the sequential deadline gets its own
resolver key. Unset, it inherits the concurrent batch deadline (same
value, same HERMES_CONCURRENT_TOOL_TIMEOUT_S bridge) so the two executor
paths cannot drift by default; set, it can be tuned or disabled
independently. Documented in cli-config.yaml.example; 5 contract tests.
Deliberately NOT on run_bounded_sync: the executors extend deadlines
dynamically during human approval waits (authorization-gate excluded
seconds) — the shared primitive is fixed-deadline. Noted in the docstring.
Clarify waits on a human for up to 3600s or unlimited. The generic sequential timeout was aborting that wait at 420s and leaving the prompt and worker active.
Ports Claude Code's /loop (and its /proactive alias) across every Hermes
surface. /loop [interval] <prompt> re-runs a prompt or slash command on a
recurring cadence inside the live session; omitting the interval enables
self-paced mode (starts at the floor, backs off exponentially while the
agent's replies stop changing, snaps back on change — local digest
comparison, zero extra LLM cost).
Stop conditions: agent-emitted LOOP_COMPLETE marker, --times N,
--until <condition> (judged by the existing goal_judge aux task,
fail-open), /loop stop, and a loops.max_ticks backstop budget.
Core: hermes_cli/loops.py (LoopState + LoopManager + shared
dispatch_loop_command), persisted per session in SessionDB state_meta
(loop:<sid>) so /resume picks it up; migrates across compression
boundaries like /goal. New SessionDB.list_meta_prefix() powers the
gateway's cross-session scan.
Surfaces:
- CLI: /loop handler + idle-fire and post-turn-complete hooks in
process_loop (mirrors the /goal hook shape; Ctrl+C pauses the loop)
- Gateway: /loop handler with route capture, mid-run control-verb guard,
post-turn tick completion, and a supervised loop_wakeup_watcher that
injects due wakeups into idle chats via the synthetic-message path
- TUI/dashboard/desktop: command.dispatch handler + per-session
notification-poller wakeup driver + post-turn completion in the turn
dispatcher; /loop added to the desktop slash palette
- /goal mixing: an active non-parked goal owns the idle boundary — loop
ticks defer until it finishes, pauses, or parks; real user input always
wins over both
Config: loops.{min_interval_seconds,max_ticks,self_paced_floor_seconds,
self_paced_ceiling_seconds}. Docs page + sidebar entry. 77 new tests.
Slack's 50-slash cap: /version moves to /hermes version to free the
native slot for /loop.
A captured native-compaction checkpoint lives in the persisted
codex_reasoning_items sidecar, but the wire restructure that follows it
(prune_pre_checkpoint_items) ran unconditionally: the native gate only
decided whether context_management went into the request, and no signal
from it ever reached _chat_messages_to_responses_input.
So a single checkpoint kept deleting every pre-checkpoint item from all
later requests — after a mid-session swap out of the gpt-5.6 family,
after compression.enabled: false, after the rejection kill switch, and
after a session resume that reloads the sidecar from state.db. The model
receiving the opaque blob was no longer the one able to decode it, and
nothing was logged.
Thread a single native_compaction_eligible boolean, derived from the same
value that gates the context_management field, into the converter. When
ineligible: do not replay type: "compaction" items and do not prune. Safe
because native compaction never truncates Hermes' local history, so the
fallback still carries the full conversation.
All Responses call sites are covered: build_kwargs and convert_messages
derive the flag via _native_compaction_active, the auxiliary/compression
client is explicitly ineligible, and the converter defaults to False
(pre-feature wire) so future call sites are safe by construction.
Fixes#85914
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Run the expensive-model warning for explicit startup `-m` / `--provider`
overrides before the chat loop starts, and fail closed for non-interactive
invocations that select an expensive or known-confusing model.
Also classify Nous paid-model 404s that say credits are required as billing
exhaustion so they fail fast with billing guidance.
Tested:
- scripts/run_tests.sh tests/hermes_cli/test_cli_startup_model_cost_guard.py tests/hermes_cli/test_model_cost_guard.py tests/agent/test_error_classifier.py -- --tb=short -q
A provider-confirmed rearm (#85846) resets the shared attempt budget, but
an earlier insufficient-progress verdict left _preflight_compression_blocked
armed, keeping the pre-API gate dark for the rest of the turn — a later
pressure spike could still grow unchecked until the provider overflow
handler fired. Clear the blocker and the stale pressure reading inside the
provider-confirmed rearm branch: the prompt is proven back below the
threshold, so the old verdict describes a request shape that no longer
exists.
Builds on @h-mascot's #84995, whose commit is preserved on this branch;
his rearm condition was superseded by #85846's latch-verified variant, but
the blocker-clear half was correct and is kept.
A marathon tool turn burned all compression_attempts on *successful*
pre-API compactions; the gate then went permanently dark and the context
grew unchecked until the provider rejected the request terminally
("Context length exceeded: max compression attempts (3) reached", session
f087963205f9, 2026-08-01). The budget now refunds at loop top when the
assembled request is back under threshold * 0.8 AND the compressor's own
should_compress() agrees the pressure is gone.
Anti-thrash intent of the cap is preserved (#11529): no-progress passes
never reach the refund margin, divergent-signal cases (should_compress
still True) keep the budget burnt, and the insufficient-progress blocker
is untouched.
Validation: new behavioral suite (7) + all 130 compression/context tests
via scripts/run_tests_hermetic.py.
Rückbau: Commit revertieren; kein Zustand, keine Migration.
(cherry picked from commit 041b489d566bfb1d6816d53d9acd5ac50b8d5af0)
Salvaged from PR #85702 by @JoaoMarcos44, composed onto the mapping-safe
_usage_get reads (PR #74591 by @RelaxJonh) and the flat cached_tokens /
Anthropic-name fallbacks (PRs #66105, #52571):
- cache-write precedence in the chat_completions branch:
details.cache_write_tokens > details.cache_creation_input_tokens >
usage.cache_creation_input_tokens > usage.cache_write_tokens
- codex_responses branch reads details.cache_write_tokens (GPT-5.6+
documented name) with cache_creation_tokens fallback (from PR #70522)
- _usage_count(): clamp malformed negative counters to 0
- all reads in every branch are mapping-safe via _usage_get
When the Responses API returns usage as a plain dict (e.g. from a
middleware or proxy that deserialises JSON to dict instead of a typed
SDK object), normalize_usage() used getattr() exclusively, which
silently returned 0 for every field on a dict.
Add _usage_get() helper that reads via .get() for dicts and getattr()
for attribute-style objects. All accessor sites in normalize_usage()
now use this helper, so token counts and cost are correct regardless
of the usage object's type.
Regression tests: two new tests feed the same payload as both a dict
and a SimpleNamespace through the codex_responses and
chat_completions branches, asserting identical output and non-zero
values.
Kimi/Moonshot's native API (api.moonshot.cn / .ai) reports context-cache hits
as a top-level ``usage.cached_tokens``. The chat-completions branch of
normalize_usage() walks a fallback chain of
prompt_tokens_details.cached_tokens -> cache_read_input_tokens ->
prompt_cache_hit_tokens; none of those names match, so direct Kimi sessions
normalized to cache_read_tokens=0. The hits were invisible in accounting and
the cached prefix was billed at the full input rate.
Appended as the last link in that chain, so it only fills a genuine zero and
cannot override a provider that reports the nested OpenAI shape or DeepSeek's
prompt_cache_hit_tokens.
Rebuilt on current main rather than rebased — the branch was ~3400 commits
behind. The DeepSeek half of the original branch is dropped: 03c0b00f4
(#65678) landed prompt_cache_hit_tokens on main, so this is Kimi-only as the
review asked. The scripts/release.py addition to the frozen LEGACY_AUTHOR_MAP
is dropped too; contributors/emails/mehmet.kar@std.yildiz.edu.tr already
exists on main.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Local OpenAI-compatible servers like mlx_vlm.server emit
input_tokens/output_tokens in chat completion responses instead of
prompt_tokens/completion_tokens. The OpenAI Python client preserves
these as extra attributes, but normalize_usage() only looked at
OpenAI-style field names, causing input_tokens to always be 0.
This made the context progress bar stay at 0% forever and prevented
auto-compression from triggering.
Fix: add Anthropic-style fallback (input_tokens/output_tokens) in the
default else-branch of normalize_usage, with OpenAI-style names taking
priority.
Fixes#14686
Jack Lau