#54609 moves anthropic into the _slot_runtime name-preservation set (it must
NOT forward base_url/api_key — OAuth sk-ant-oat* needs the provider branch's
Bearer + anthropic-beta header). The pre-existing parametrized
test_moa_provider_backed_slot_survives_aux_resolution still listed anthropic
asserting the forward path, contradicting the new behavior. anthropic is now
covered by test_slot_runtime_anthropic_oauth_routes_through_provider_branch;
drop it from the forward-path parametrize (minimax-oauth/qwen-oauth remain).
MoA's _slot_runtime() whitelists providers that must keep their provider
identity (so call_llm runs their provider branch) instead of being treated
as a plain custom endpoint via forwarded base_url/api_key. Native anthropic
was missing from this set.
Native anthropic subscription OAuth setup-tokens (sk-ant-oat*) require Bearer
auth plus the 'anthropic-beta: oauth-*' header, which only the anthropic
provider branch adds. Without the whitelist entry, the slot's base_url/api_key
were forwarded and call_llm sent the OAuth token as x-api-key, which Anthropic
rejects with a bare 429 (rate_limit_error with no quota details). This made
anthropic references in MoA presets fail every time.
Add 'anthropic' to the whitelist so native anthropic reference/aggregator
slots route through the provider branch. Extends upstream 9229d0db1 which
added 'nous' for the same reason.
Buffered text/photo/media-group flushes and the polling-error recovery
task sit behind an asyncio.sleep(). On disconnect they kept running and
dispatched handle_message() into a torn-down session, producing stale or
duplicate deliveries. disconnect() only cancelled media-group and photo
batch tasks — text batches and the polling-error task leaked.
Set a _drop_delayed_deliveries flag from _mark_disconnected/_set_fatal_error
(cleared by _mark_connected) and check it in all enqueue+flush paths so a
flush that wins the race against teardown drops instead of dispatching.
_cancel_pending_delivery_tasks() now cancels+clears all four task maps,
skipping the current task. Media-group flush finally-block guarded so a
cancelled stale flush cannot erase a replacement task handle.
/queue rebuilt the queued MessageEvent with only text/type/source/
message_id/channel_prompt, silently dropping any photo, document, voice,
or reply context attached to the command. The deferred turn then ran with
the attachment lost. Carry the full payload through, and accept a /queue
that has media but no prompt text (e.g. "/queue" as an image caption).
Salvaged from #13913 by @ypwcharles — the gateway busy-session/queue
infrastructure was rewritten since that PR (Telegram moved to
plugins/platforms/, /queue now uses the FIFO chain), so the media fix is
reimplemented against the current handler; the PR's batching and
busy-bypass changes targeted code paths that no longer exist.
Co-authored-by: ypwcharles <92324143+ypwcharles@users.noreply.github.com>
GNU tools accept unique long-option prefix abbreviations at runtime, so
`chown --recurs root` and `git push --forc` evaded the approval gate's
exact-match `--recursive`/`--force` patterns. Switch those two entries
to prefix matches (--recur[a-z]*, --forc[a-z]*).
The rm/chmod/sed long-flag patterns were left unchanged: every abbreviation
of those is already caught by the sibling short-flag and target patterns
(rm -[^s]*r, base chmod 777, sed -[^s]*i), so prefix-matching them is a
no-op. Only chown (beyond the coincidental case-insensitive r->R catch) and
git push had genuine gaps.
Co-authored-by: Subway2023 <subw3@mail2.sysu.edu.cn>
The --nous flag was only wired into the argparse `hermes debug share`
subcommand. The /debug slash command (classic CLI + TUI, both via
process_command -> _handle_debug_command) built a hardcoded args
namespace with no `nous` attribute, so it always took the default
paste.rs path.
Pass cmd_original through to _handle_debug_command and parse an optional
destination word:
/debug -> public paste (default, unchanged)
/debug nous -> Nous-internal S3
/debug local -> stdout, no upload
local wins over nous (never touches the network); unknown words fall
back to the default. Add args_hint="[nous|local]" so help/autocomplete
surface it. New TestDebugSlashCommand covers the parsing + dispatch.
NAS PR #349 (merged) ships a stateless presigned-PUT endpoint: the only
route is POST /api/diagnostics/upload-url, and the object's existence in S3
is the only state. There is no /api/diagnostics/confirm route — confirming
live against the merged preview returns 404.
The client's confirm_upload() therefore fired a guaranteed-404 request on
every --nous upload (harmless, since errors were swallowed, but dead).
Remove it and simplify share_to_nous() to the 2-step mint + PUT flow that
matches the shipped contract. Drop the corresponding TestConfirmUpload class
and confirm assertions; add a test that the share succeeds even when the
response carries no id (we no longer depend on it).
The separately-flagged cross-repo requirement from #349's review --
sizeBytes is now REQUIRED and signed into the presigned URL's ContentLength
-- was already satisfied: share_to_nous() sends len(bundle) as sizeBytes and
urllib sets a matching Content-Length on the PUT. Verified against the live
merged preview (missing sizeBytes -> 400 invalid_body; present -> 503 dark).
Tested: pytest tests/hermes_cli/test_diagnostics_upload.py tests/hermes_cli/test_debug.py -> 95 passed.
`hermes debug share --nous` uploads the (force-redacted) debug bundle to
Nous-internal S3 storage via a presigned URL minted by the Nous account
service, instead of a public paste. The bundle is private — viewable only
by Nous staff / allowlisted mods through a Google-OAuth-gated viewer — and
auto-deletes after 14 days. The paste.rs path is unchanged and remains the
default.
- hermes_cli/diagnostics_upload.py (new): stdlib-urllib NAS client —
request_upload_url(), put_bundle(), confirm_upload() (best-effort),
share_to_nous() orchestrator. Base URL via HERMES_DIAGNOSTICS_BASE_URL
(default https://portal.nousresearch.com).
- hermes_cli/debug.py: extract collect_share_bundle() from build_debug_share()
so the Nous path reuses the exact same redaction/collection (paste.rs
behaviour unchanged); add build_nous_bundle() producing the gzipped
{"format":"hermes-debug-share/1","redacted":...,"files":...} envelope the
discord-support viewer parses; add the --nous run path with a privacy
notice and a clean fallback (suggest --local) on failure.
- hermes_cli/main.py: add the --nous flag + help/epilog entry on
`debug share`.
- tests: test_diagnostics_upload.py (new) mocks urllib; test_debug.py adds
bundle/Nous coverage. 97 passing.
The dangerous-command approval prompt renders the flagged command so the
user can decide whether to approve. If the agent constructed it with a
credential (curl -H 'Authorization: Bearer sk-...', psql postgres://user:pw@host,
an execute_code script with api_key = 'sk-...'), that secret hit stdout and,
via the gateway notify payload, Discord/Slack messages — which are
screenshottable and forwardable.
Apply the existing agent.redact.redact_sensitive_text() to every user-facing
approval surface. Redaction is display-only: the raw command still executes
after approval, and approval persistence keys off pattern_key (not the command
text), so the allowlist is unaffected. Decision context (URL, flags, command
structure) is preserved; only the secret value masks.
Covers all surfaces, including the execute_code path the original PR missed:
- prompt_dangerous_approval(): callback + stdout fallback
- check_all_command_guards(): gateway approval_data + cron/batch pending fallback
- check_execute_code_guard(): gateway approval_data + no-notifier pending fallback
(script body can embed credentials)
Adds TestApprovalPromptRedaction covering callback redaction, no-over-redaction
of clean commands, and the execute_code pending fallback.
Salvaged from PR #13139 by @sgabel; extended to the execute_code surface.
Add regression tests for the sk-ant-oat OAuth heuristic and shorten the
inline comment. Verifies admin keys (sk-ant-admin-*) and standard API keys
classify as api_key, only sk-ant-oat- tokens flow into the OAuth refresh path.
The background memory/skill review thread wrapped its whole body in
process-global contextlib.redirect_stdout/stderr(devnull). Those rebind
sys.stdout/sys.stderr for the ENTIRE process, so for the full duration of
the review (tens of seconds) every other thread — including a gateway
event-loop thread driving a Telegram long-poll — also wrote to devnull.
Any bare print/sys.stderr.write from those threads during the window was
silently lost (#55769 / #55925).
Replace the global redirect with thread_scoped_silence(): a per-thread
routing proxy installed once as sys.stdout/sys.stderr that sends only the
registered (bg-review) thread's writes to devnull and passes every other
thread through to the real stream. Depth-counted so nested use composes.
Verified: a concurrent thread writing while the bg-review thread is inside
the silence window keeps its output on the real stream.
Two independent fixes salvaged from #12811 (closing it; one of its three
bundled fixes — Discord free_response — is already on main).
Anthropic max_tokens (#12790): the chat-completions max_tokens fallback only
fired for OpenRouter/Nous URLs, so any other proxy serving a Claude model
(AWS Bedrock, NVIDIA, LiteLLM, vLLM, corporate gateways) shipped requests
with no max_tokens and inherited the proxy's low default (Bedrock: 4096),
exhausting on thinking + large tool calls. Changed the gate in
chat_completion_helpers.build_api_kwargs from URL-gated to model-gated:
fires whenever the model matches an _ANTHROPIC_OUTPUT_LIMITS key. This also
fixes a latent miss — the old 'claude' substring gate skipped MiniMax and
Qwen3 even on OpenRouter. Remains a last-resort fallback (build_kwargs only
applies it after ephemeral/user/profile max_tokens), so it never overrides
an explicit value, and only touches the chat-completions transport (native
Anthropic Messages API is a separate path).
Feishu channel_prompt (#12805): the Feishu adapter never resolved
channel_prompts config, unlike Discord/Slack, so per-channel role prompts
were silently ignored. Added _resolve_channel_prompt() (delegating to the
shared gateway.platforms.base.resolve_channel_prompt) and wired it into all
three MessageEvent construction sites — inbound message, reaction routing,
and card-action routing.
Tests: tests/gateway/test_feishu_channel_prompts.py (6 cases) covering exact
match, parent-thread fallback, no-match, missing-config safety, and event
propagation.
The discard done-callback added via task.add_done_callback runs on a later
event-loop iteration (call_soon) than the one that resolves `pending` and lets
handle_401 return. Both inflight-task tests asserted the live set was empty
immediately after the await returned, racing the callback. Add a single
`await asyncio.sleep(0)` before the cleanup assertions.
`handle_401` spawned a dedup'd recovery coroutine via
`asyncio.create_task(_do_handle())` and discarded the returned task
reference. Python's event loop only keeps weak references to tasks, so
the coroutine could be garbage-collected before it called
`pending.set_result(...)`. Every concurrent caller awaiting that future
then hangs forever, and the `finally: entry.pending_401.pop(...)`
cleanup never runs — so subsequent 401s for the same key latch onto the
dead future too. Same pattern the adapter-side fixes address (#11997,
#11998, #12000, #12001, #12006).
Hold the task in a process-wide set on the manager and discard it via
`add_done_callback` once it completes. Regression test covers both the
structural invariant (task tracked, then removed on completion) and a
concurrent dedup path with a forced `gc.collect()` between the handler's
await points.
A /learn request can mix the source(s) to gather (paths, URLs, "what we
just did") with requirements that shape the skill (focus, scope, what to
omit). When a request led with a path or link, the agent fetched it and
treated the trailing prose as incidental, dropping the user's stated
focus — the symptom @GrenFX reported.
The input layer was never the cause: both CLI (split(None, 1)) and
gateway (get_command_args()) capture the full free-text argument. The
gap was in build_learn_prompt, which dumped the request as one
undifferentiated source blob.
build_learn_prompt now tells the agent the request may mix sources and
requirements in any order, that prose after a path/link is authoring
guidance to honor (not noise), and to never fetch the first source and
ignore the rest. Adds step 1b: apply every requirement to what the
SKILL.md covers, not just which sources get read. Both surfaces inherit
it; no parser change, zero tool footprint.
Tool call ids are used to name persisted large-result files. Treating that id as a raw path segment allowed traversal-like ids to resolve outside hermes-results even though the shell command quoted metacharacters.
Convert ids to single filename stems, preserve normal ids, and add a short hash when normalization is needed so unsafe ids do not collide silently.
Constraint: Avoid new dependencies and preserve existing tool-result paths for normal tool call ids
Rejected: Quote only the path | shell quoting does not prevent ../ path traversal
Confidence: high
Scope-risk: narrow
Reversibility: clean
Tested: source /Users/peter/hermes-agent/venv/bin/activate && pytest tests/tools/test_tool_result_storage.py -q
Tested: source /Users/peter/hermes-agent/venv/bin/activate && python -m compileall tools/tool_result_storage.py tests/tools/test_tool_result_storage.py
Tested: git diff --check
Some legitimate @bot pings were dropped because the mention gates relied on
message.mentions alone, which does not always populate raw <@ID> / <@!ID>
forms (mobile, edited, relayed messages). A bare @bot with no other text
could also spawn a fake empty-text turn.
- add _self_is_explicitly_mentioned() / _raw_mentioned_user_ids() helpers that
treat the bot as mentioned via resolved mentions OR raw content forms
- use them at the allow_bots=mentions gate, multi-agent bot filtering, the
mention-strip/mention_prefix step, and the require_mention gate
- drop bare mention-only pings (no text, no media, no injection, no backfill
context) instead of injecting a placeholder empty turn
Co-authored-by: Teknium <teknium1@gmail.com>
Registers PowerShell (.ps1/.psm1/.psd1) in the LSP server registry,
spawning PowerShellEditorServices over stdio via a pwsh/powershell
host. PSES ships as a GitHub release zip (no npm/go/pip recipe), so it
sits in the manual install tier alongside rust-analyzer and clangd.
The spawn builder resolves the module bundle from (in order) the
lsp.servers.powershell.command override, init bundlePath, the
PSES_BUNDLE_PATH env var, or <HERMES_HOME>/lsp/PowerShellEditorServices,
then launches Start-EditorServices.ps1 -Stdio with a non-interactive,
no-profile host. hermes lsp status/list report it as manual-only until
pwsh is present.
Docs and tests included.
When the summary LLM hits a 429/transient failure, _generate_summary() sets
a cooldown and returns None; compress() inserts a static fallback marker and
returns. Tokens stay above threshold, so should_compress() kept returning
True and every subsequent agent turn re-fired _compress_context() — the CLI
appeared frozen until the cooldown expired.
Add a cooldown guard to should_compress(): return False while
_summary_failure_cooldown_until is in the future. Reuses the existing float;
no new state. Manual /compress (force=True) still clears the cooldown first.
Fixes#11529
Generic provider:custom relays were force-routed to the OpenAI Responses
API whenever the model matched gpt-5*, and a stale persisted
model.api_mode=codex_responses survived /reset and upgrades. Some
OpenAI-compatible relays do not implement Responses semantics, which
surfaced as malformed function_call.name replay errors in gateway sessions.
- runtime_provider: route custom-provider api_mode through
_resolve_plain_custom_api_mode(), which drops a stale codex_responses
unless the URL is direct OpenAI/xAI
- run_agent: _provider_model_requires_responses_api returns False for
custom; direct api.openai.com / api.x.ai URLs still upgrade via
_is_direct_openai_url() / URL detection
- regression coverage for plain relays vs direct OpenAI/xAI URLs
Co-authored-by: HiddenPuppy <HiddenPuppy@users.noreply.github.com>
* fix(agent): drop tool_calls with empty function.name to prevent orphan 400
Salvage of #12807 by @melonboy312 — rebased onto current main (sanitizer
moved to agent_runtime_helpers), scoped to the sanitizer fix, with a
regression test that fails without it.
* fix(agent): repair (not drop) empty-name tool_calls to preserve anti-priming + prevent 400
Dropping empty-name tool_calls in the pre-call sanitizer collided with #47967,
which intentionally keeps an empty-name call paired with a synthesized
'tool name was empty' anti-priming result so weak models self-correct without
a full catalog dump. Dropping the call orphaned that result and stripped the
signal (breaking tests/agent/test_empty_tool_name_loop_dampening.py).
The actual HTTP 400 cause is an ORPHANED function_call_output (adapter drops
the empty-name function_call but keeps its output). Rename the blank name to a
non-empty sentinel instead: the call and its result stay paired, the adapter
no longer drops the function_call, no orphan, no 400 — and the anti-priming
result content the model needs is preserved.
---------
Co-authored-by: Bartok9 <danielrpike9@gmail.com>
The FactRetriever's _fts_candidates passed the raw query string directly
to FTS5's MATCH operator. FTS5 defaults to AND-between-tokens, which
means any multi-word prose query like 'what happened with the deployment
rollback' required every single token to co-occur in a fact — dropping
recall to zero on the kind of queries agents actually issue via prefetch().
Fix: add _sanitize_fts_query() that:
- tokenizes the query and drops English stopwords
- strips FTS5 operator characters per token
- OR-joins the remaining content tokens as phrase literals
For pathological inputs (all stopwords, empty), falls back to the raw
query so the caller sees zero results instead of a SQL error.
This is a pure-retrieval-quality fix — the HRR + Jaccard reranking
stages still keep precision high. Ships with 10 tests covering the
sanitizer and retrieval integration.
The init snapshot dumped functions with a line-based filter:
declare -f | grep -vE '^_[^_]'
That strips a function's *header* line (e.g. `_foo () `) but leaves the
orphaned `{ ... }` body behind, corrupting the snapshot that is sourced
before every command. Sourcing the torn snapshot runs leftover body code
and breaks subsequent commands (intermittent exit 127).
- Filter private (`_`-prefixed) functions by NAME via `declare -F` and
dump only the wanted whole definitions, so a body is never torn. Guard
against an empty name list (bare `declare -f` dumps everything).
- Treat a non-zero bootstrap exit code as snapshot-init failure, so
execution safely falls back to login-shell-per-command mode.
- Add a regression test asserting snapshot_ready stays false when
bootstrap exits non-zero.
Preserves the atomic-write ($BASHPID temp + mv -f) machinery from #38249.
The polling heartbeat's pending-update probe treated a stopped updater
(running=False) as "someone else's job" and silently reset its counter,
so a long-poll task that disappears with no reconnect in flight was never
recovered. get_me() on the general request path stays healthy, so neither
PTB's error_callback nor the connectivity probe ever fires — the gateway
keeps running but stops receiving messages indefinitely (#55769).
Detect the stopped-updater case directly in _probe_pending_updates and feed
it into the existing _handle_polling_network_error ladder, debounced over two
consecutive probes so a just-starting updater or the brief stop()->start_polling()
window of an in-flight reconnect never trips it.
background_review hardcoded enabled_toolsets=["memory", "skills"] in the
review fork's whitelist, so a skill-review fork on a profile with
memory_enabled: false still granted the LLM the built-in MEMORY.md read/write
tool — contaminating a profile that opted out of built-in memory. The flag was
already in scope (review_agent._memory_enabled). Include "memory" only when
_memory_enabled or _user_profile_enabled (USER.md also needs the tool).
Layer 1 of #54937 (the path leak) is fixed by this PR's thread-context
propagation: get_memory_dir() is already per-call on main, so once the
bg-review thread inherits the profile override its writes land in the right
profile (verified). This commit closes the remaining whitelist layer.
The inbound-media validator _is_allowed_bridge_path() checked against
IMAGE_CACHE_DIR / AUDIO_CACHE_DIR / VIDEO_CACHE_DIR / DOCUMENT_CACHE_DIR
value-imported at module load. After the base.py cache-dir getters became
per-call resolvers, the bridge writes media into the active profile's cache
while the validator still matched the frozen launch-profile constants — so
media was rejected under a profile override (multi-profile gateway).
Resolve the cache roots per-call via the get_*_cache_dir() getters and drop
the now-unused frozen value-imports. Caught by automated review on #55867.
The reachability claim that single-process multi-profile leakage is desktop-
only is incomplete. gateway/run.py:_profile_runtime_scope shows a SECOND such
runtime: the multiplexed gateway (gateway.multiplex_profiles) serves every
profile from one process, scoping each inbound turn with the same
set_hermes_home_override ContextVar the desktop uses (and the /p/<profile>/
URL prefix). The M1 (import-time path globals) and M2 (thread/executor
context) leaks are reachable there identically.
- tests/gateway/test_multiplex_credential_isolation.py: add a class driving the
skills-dir + cache-dir resolvers and a propagated worker thread under the
real _profile_runtime_scope, asserting each resolves the active profile. Sits
beside the existing credential-isolation proofs for the same topology.
- Correct the inline comments in model_tools/run_agent/async_delegation/
rich_sent_store to name both runtimes (desktop tui_gateway AND the
multiplexed gateway) instead of implying desktop is the only surface.
(ACP runs one agent per subprocess and the kanban dispatcher Popens
'hermes -p <profile>' children, so neither is an in-process multi-profile
surface; desktop + multiplexed gateway are the two confirmed ones.)
- tools/skills_hub.py: the per-call resolvers now honor a test-injected real
module attribute (patch.object(hub, 'SKILLS_DIR', ...) / monkeypatch.setattr)
before falling back to dynamic profile resolution. PEP 562 __getattr__ only
fires when no real attribute exists, so an unpatched module resolves the
active profile and a patched one respects the test's value — keeping the
existing skills_hub test seam intact (5 tests had broken).
- tests/test_profile_isolation_runtime.py: real two-profile (no-mock) suite
driving each previously-leaking site under override A then B and asserting
the active profile's path/identity is used: skills_hub paths + derived
constants + default-arg resolution, gateway cache getters (incl. the
monkeypatch-still-wins seam), rich_sent_store path, and thread/executor
context propagation (raw-thread hazard documented; primitive + _run_async
worker proven to preserve the override).
Adds gateway.platform_connect_timeout (default 30s) to DEFAULT_CONFIG and
bridges it to the internal HERMES_GATEWAY_PLATFORM_CONNECT_TIMEOUT env var
at gateway startup, following the existing gateway_timeout config->env
pattern. The env var remains the manual-override escape hatch and wins if
set explicitly; otherwise config.yaml supplies the value. This closes the
issue's documentation/config-surface request (#19776 suggestion 2) on top
of the adapter ready-wait fix, so users no longer need an undocumented env
var to raise the Discord connect timeout.
Refs #19776
The extension-less MEDIA delivery guards short-circuited on
"MEDIA: not in text and [[audio_as_voice]] not in text", so a
response carrying only [[as_document]] (an image-only reply requesting
unmodified document delivery) leaked the directive as visible text.
Add [[as_document]] to both guard conditions (_strip_media_tag_directives
and strip_media_directives_for_display) and cover it with a regression
test.
Follow-up to liuhao1024's #46924. Route plain-text approval replies
through the canonical /approve and /deny handlers (resolve thread, resume
typing, return localized confirmation) and deliver that confirmation back
to the user — previously a plain 'yes' resolved silently. Synthesize a
literal '/'-prefixed command so get_command_args() parses always/session
modifiers on every platform (is_command() only recognizes '/'). Add E2E
tests covering approve/deny/always/session vocab plus the no-pending and
unrelated-text fall-through cases.
The judge gate added for kanban_complete (Issue #38367, PR #38388) only
covers one of the two exit paths out of run_kanban_goal_loop(). The loop
treats status == "blocked" as terminal identically to "done" (and any
other status outside running/ready/done/blocked also stops the loop —
see goals.py's status dispatch). A goal_mode worker that has learned
kanban_complete is gated can simply call kanban_block(reason="anything")
to escape the loop with zero judge involvement, fully defeating the
intent of #38367's fix.
This is Issue #38696, filed as the explicit follow-up by a reviewer on
PR #38388: "kanban_complete is one way out; kanban_block is another...
A worker that learns the complete path is gated can shift to calling
block to escape the loop with the same effect."
Implements the issue's "Option B" (deterministic allowlist, no extra
judge LLM call) using the kind taxonomy that already exists in
kb.VALID_BLOCK_KINDS, rather than inventing a new judge_goal() outcome
type (judge_goal only returns done/continue/wait/skipped — there's no
"is this block legitimate" verdict to hook the issue's "Option A"
pseudocode onto without expanding the judge's contract).
goal_mode tasks may only block with kind in {dependency, needs_input} —
the two kinds that represent a genuine external blocker the worker
cannot resolve itself. `capability`, `transient`, and an unset kind are
rejected with a message directing the worker to kanban_complete instead,
which the judge now gates. Non-goal_mode tasks are completely unaffected.
kitty fits an image to its cell rect preserving aspect, so a frame whose pixel
size isn't a whole multiple of the cell rounds up — clipping the bottom row
("clipped feet") and letterboxing a blank row. Trim each frame to its union
alpha bbox, then snap to an exact cell multiple before transmit so the sprite
hugs its box and renders full-body. (ratatui-image#57: render in multiples of
the font-size.)
Assert a journey edit leaves MEMORY.md byte-identical to MemoryStore's
own §-join (no trailing-newline drift) and round-trips through
MemoryStore._read_file, so the two surfaces can never diverge on format.
MoA sessions could not stream: the gateway streaming toggle was a no-op for
provider "moa", so users saw nothing until the entire response finished — minutes
of silence on long turns. The aggregator's reply was always fetched whole.
Root cause was twofold:
1. conversation_loop hard-disabled streaming for provider in {"copilot-acp",
"moa"} (MoA grouped with the ACP client, whose facade isn't a stream).
2. MoAChatCompletions.create() fetched the aggregator response whole via
call_llm(), which had no streaming mode.
For provider "moa", _create_request_openai_client() returns the MoAClient facade
itself, so the existing streaming consumer already calls
MoAChatCompletions.create(stream=True). We reuse that battle-tested consumer
(text-delta delivery, tool_call reassembly, stale-stream detection, non-streaming
fallback) instead of adding a parallel streaming path.
Changes:
- call_llm() gains stream/stream_options. When streaming it returns the raw SDK
stream iterator directly, bypassing _validate_llm_response and the
temperature/max_tokens/payment fallback chain (which assume a complete
response). The caller owns reassembly and fallback.
- MoAChatCompletions.create() runs the references first (unchanged), then when
stream=True returns the aggregator's raw stream, forwarding stream_options and
the consumer's per-request read timeout. stream=False is byte-identical to
before (no stream/stream_options/timeout forwarded).
- conversation_loop streams MoA only when a display/TTS consumer is present;
quiet/subagent/health-check paths keep the complete-response path.
Tests: tests/run_agent/test_moa_streaming.py — create() stream/non-stream
branches, stream_options + timeout forwarding, call_llm raw-stream return vs
validated non-stream. Existing MoA tests unchanged (20 passed).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
_build_gemini_contents emitted one contents entry per source message and
never merged adjacent same-role entries. Gemini's generateContent requires
strict user/model alternation and rejects consecutive same-role turns with
HTTP 400 ("Please ensure that multiturn requests alternate between user and
model"). A parallel tool call turns into two tool results in a row, which
become two consecutive user functionResponse contents, so every multi-tool
turn produced an unsendable history.
Fold adjacent same-role contents into one by concatenating their parts after
the per-message loop, matching the Anthropic and Bedrock converters. For a
parallel call this yields the grouped multi-functionResponse user turn Gemini
expects.
Memories are the only drillable rows, so give them the primary "clickable"
ink and demote skills (dead-ends) to the muted complement — previously the
non-openable skills wore the link-looking primary color. Flipped in both
the TUI and CLI palettes for parity.
A single 'hermes update' / 'hermes -p' could rewrite a hand-curated config.yaml
into a near-full DEFAULT_CONFIG dump (the 'you blow up my profile config on one
tweak' reports). Root cause: migrate_config() had ~16 independent save_config()
call sites, each author deciding ad hoc whether to materialise a value, and many
persisted pure schema defaults with strip_defaults=False. Defaults already merge
transparently at read time via load_config(), so writing them is pure bloat that
also shadows future default changes (see save_config's docstring).
Architectural fix (not a per-site patch): introduce a single _persist_migration()
chokepoint that enforces one invariant — a migration may persist only values that
DIFFER from the current schema default, plus explicit removals/renames of user
data; pure defaults are never written. Every migration write (all 17 sites incl.
the version-bump finalizer) now routes through it. The invariant is mechanically
correct for all cases and verified empirically:
- pure-default seeds (timezone='', curator/auxiliary.curator blocks, interim
flag, curator.consolidate=False, empty plugins.enabled) are stripped → merged
in at read time;
- non-default values (write_approval=True, model_catalog.ttl_hours=1) preserved
via explicit-raw-path preservation;
- behaviour flips (agent.verify_on_stop=False, schema default still 'auto')
preserved because False != 'auto';
- data transforms (custom_providers->providers, stt.model relocation,
write_mode->write_approval, compression.summary_* removal, MCP-disable)
persist their removals/renames.
An explicitly user-set non-default value (e.g. matrix.require_mention: false) is
preserved across the bump.
Guard tests lock the architecture: an AST check asserts migrate_config() makes no
direct save_config() call (all writes go through _persist_migration), and a
full-range v1->latest test asserts a lean config is never dumped. Two existing
change-detector tests that froze the on-disk representation of default-valued
keys are rewritten to assert the effective value via load_config() (behaviour
contract, not snapshot).
Validation: lean v1->latest migration drops from ~567 bytes to ~196 bytes;
148 config+setup and 196 profile/curator/migrate tests pass on scripts/run_tests.sh.
Builds on the zero-match feedback fix (previous commit) to close the silent-hang
symptom: when memory is at capacity, a failed `add`/`replace`/`remove`
consolidation could loop the whole turn to iteration-budget exhaustion and
deliver no user-facing reply.
#41755 turned the at-capacity overflow error into a *commanded* in-turn retry
("...then retry this add — all in this turn"); combined with the fragile
substring-only `replace`/`remove` matching (LLMs can't reliably re-quote a long
entry verbatim), the model loops add↔replace on inexact guesses until the turn
dies. The existing tool_guardrails halt would catch this, but hard_stop_enabled
is opt-in (off by default), so a default install still hangs.
This fixes it at the memory layer without changing global guardrail behavior:
- MemoryStore tracks per-turn consolidation failures; after a cap (3) it drops
the "retry in this turn" instruction and returns a terminal "leave memory
unchanged, continue your reply" result, so a failed memory side effect can
never block the turn's reply.
- The counter resets on any successful write (progress) and at each turn
boundary (turn_context.reset_consolidation_failures, guarded via getattr so
plugin memory stores without the method are a no-op).
Co-authored-by: liuhao1024 <sunsky.lau@gmail.com>
Regression for #54820: a desktop-only helper with a failing check_fn must
not mark the whole terminal toolset unavailable when terminal/process
still pass their per-tool gates.
