Files
hermes-agent/agent/moa_loop.py
Teknium 4e65d089c0 refactor(agent): simplify moa_loop — dataclass accounting, extracted create() helpers, relay dispatch table, compacted docstrings
- _RefAccounting is a slotted dataclass (trace redaction uses dataclasses.replace)
- create(): preset resolution, fan-out cadence key, reference joining and
  usage/cost summing extracted into named helpers; guidance header shared
- reference pricing and pending-accounting folds unified (one lock path)
- build_moa_facade display-event relay routed via _RELAY_EVENTS table
- dict/object tool-call field access collapsed into _field; dead branches removed
- comments/docstrings compacted to their invariants and WHY content
2026-09-02 13:29:35 -07:00

1769 lines
77 KiB
Python

"""Mixture-of-Agents runtime helpers for /moa turns.
The slash command marks one user turn as MoA-enabled; the normal agent loop still
owns tool calling and turn termination, while this module gathers reference-model
context before each model iteration.
"""
from __future__ import annotations
import hashlib
import json
import logging
import re
import threading
import time
from concurrent.futures import ThreadPoolExecutor, wait as _futures_wait
from dataclasses import KW_ONLY, dataclass, replace
from types import SimpleNamespace
from typing import Any
from agent.auxiliary_client import call_llm
from agent.message_content import flatten_message_text
from agent.transports import get_transport
from agent.usage_pricing import CanonicalUsage
logger = logging.getLogger(__name__)
# Privacy filter (moa.privacy_filter: '' | display | full, #59959). Secret shapes
# are handled by agent.redact; these two patterns add the PII classes it leaves
# alone (emails, formatted NA phones). The phone pattern requires explicit
# delimiters so line numbers, dates, times, SHAs, IPs and versions never match.
_MOA_EMAIL_RE = re.compile(r"\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b")
_MOA_PHONE_RE = re.compile(
r"(?<![\w.+-])" # no leading word char / dot / + / - (kills IPs, IDs, versions)
r"(?:\+?1[ .-])?" # optional NA country code
r"(?:\(\d{3}\)[ .-]?|\d{3}[.-])" # delimited area code: (555) or 555- / 555.
r"\d{3}[.-]\d{4}" # exchange-subscriber with explicit separator
r"(?![\w-])" # no trailing word char / hyphen
)
def _redact_reference_text(text: Any) -> Any:
"""Redact secrets (central redactor) then MoA PII patterns.
force=True: the privacy filter is its own opt-in, independent of the global
log-redaction toggle. code_file=True: advisory text is prose/code, so the
ENV/JSON assignment heuristics that mangle source snippets stay off.
"""
if not isinstance(text, str) or not text:
return text
from agent.redact import redact_sensitive_text
text = redact_sensitive_text(text, force=True, code_file=True)
text = _MOA_EMAIL_RE.sub("[redacted email]", text)
text = _MOA_PHONE_RE.sub("[redacted phone]", text)
return text
def _moa_privacy_mode(moa_raw: Any) -> str:
"""Resolve the normalized privacy-filter mode from a raw ``moa`` config."""
from hermes_cli.moa_config import coerce_privacy_filter
raw = moa_raw if isinstance(moa_raw, dict) else {}
return coerce_privacy_filter(raw.get("privacy_filter"))
def _redact_reference_outputs(
reference_outputs: list[tuple[str, str, Any]],
) -> list[tuple[str, str, Any]]:
"""Redact advisor text in reference-output tuples; accounting slot untouched."""
return [
(label, _redact_reference_text(text), acct)
for label, text, acct in reference_outputs
]
def _redact_message_content(content: Any) -> Any:
"""Redact a message's content: a string, or the text parts of a content-part list."""
if isinstance(content, str):
return _redact_reference_text(content)
if isinstance(content, list):
return [
{**p, "text": _redact_reference_text(p.get("text"))}
if isinstance(p, dict) and isinstance(p.get("text"), str)
else p
for p in content
]
return content
def _redact_trace_messages(messages: Any) -> Any:
"""Redact message copies for trace persistence (string or content-part lists)."""
if not isinstance(messages, list):
return messages
return [
{**m, "content": _redact_message_content(m.get("content"))} if isinstance(m, dict) else m
for m in messages
]
def _redact_trace_accounting(acct: Any) -> Any:
"""Copy a ``_RefAccounting`` with its trace text (messages/output) redacted."""
if not isinstance(acct, _RefAccounting):
return acct
return replace(
acct,
messages=_redact_trace_messages(acct.messages),
output=_redact_reference_text(acct.output),
)
# Cold-start caches (#66793): the preset and each (provider, model) runtime are
# immutable for a turn, so avoid re-resolving them on every create() call.
_preset_cache_lock = threading.Lock()
_preset_cache: dict[tuple, Any] = {}
def _resolve_preset_cached(preset_name: str) -> tuple[dict[str, Any], Any]:
"""Return ``(preset, raw moa config)``, caching the resolved preset per config mtime.
load_config() is (mtime_ns, size)-cached upstream; the saving here is skipping
resolve_moa_preset's full validation of the moa block on every create().
"""
from hermes_cli.config import get_config_path, load_config
from hermes_cli.moa_config import resolve_moa_preset
try:
cfg_stamp = get_config_path().stat().st_mtime_ns
except OSError:
cfg_stamp = None
moa_raw = load_config().get("moa") or {}
key = (cfg_stamp, preset_name)
preset = None
if cfg_stamp is not None:
with _preset_cache_lock:
preset = _preset_cache.get(key)
if preset is None:
preset = resolve_moa_preset(moa_raw, preset_name)
if cfg_stamp is not None:
with _preset_cache_lock:
_preset_cache.clear() # one live config stamp at a time
_preset_cache[key] = preset
return preset, moa_raw
_runtime_cache_lock = threading.Lock()
_runtime_cache: dict[tuple[str, str], tuple[float, dict[str, Any]]] = {}
# Short TTL so rotated keys / base_url edits are picked up within 5 minutes.
_RUNTIME_CACHE_TTL_SECONDS = 300.0
# Cap on concurrent reference calls (guards pathologically large presets).
_MAX_REFERENCE_WORKERS = 8
@dataclass(slots=True)
class _RefAccounting:
"""Per-reference usage, cost and full trace (third slot of a reference-output tuple).
Advisors may run on a different model than the aggregator, so cost is priced at
the advisor's OWN rate and summed in dollars. ``messages``/``output``/``model``/
``provider``/``temperature`` carry the full input/output for trace persistence
(only populated when tracing is on).
"""
usage: Any
cost_usd: Any = None
cost_status: str | None = None
cost_source: str | None = None
_: KW_ONLY
messages: Any = None
output: str | None = None
model: str | None = None
provider: str | None = None
temperature: Any = None
# Per-tool-result char budget for the advisory view: tool CALLS are kept in full,
# tool RESULTS are head+tail previewed. The aggregator always gets the full transcript.
_REFERENCE_TOOL_RESULT_BUDGET = 4000
# Reference system prompt: without this framing a reference assumes it is the acting
# agent and refuses ("I can't access repositories") or tries to call tools.
_REFERENCE_SYSTEM_PROMPT = (
"You are a reference advisor in a Mixture of Agents (MoA) process. You are "
"NOT the acting agent and you do NOT execute anything: you cannot call "
"tools, run commands, browse, or access files, repositories, or URLs, and "
"you should not try to or apologize for being unable to. A separate "
"aggregator/orchestrator model holds those capabilities and will take the "
"actual actions.\n\n"
"CRITICAL: You must NEVER claim or imply that you have executed a command, "
"downloaded a file, accessed a URL, or performed any action. You can only "
"analyze and advise based on the conversation context. Examples of what to "
"avoid:\n"
"- Bad: \"I ran curl and got 404.\"\n"
"- Bad: \"I downloaded the file successfully.\"\n"
"- Bad: \"I checked the repository and found...\"\n"
"- Good: \"Based on the error pattern, a curl request to that URL would likely return 404.\"\n"
"- Good: \"The conversation suggests downloading this file may help.\"\n"
"- Good: \"From the context, checking the repository would reveal...\"\n\n"
"The conversation below is the current state of a task handled by that "
"acting agent. Your job is to give your most intelligent analysis of that "
"state: understand the goal, reason about the problem, and advise on what "
"to do next. Surface the best approach, concrete next steps and tool-use "
"strategy, likely pitfalls and risks, and anything the acting agent may "
"have missed or gotten wrong. Assume any referenced files, URLs, or "
"systems exist and reason about them from the context given rather than "
"asking for access.\n\n"
"Respond with your advice directly — no preamble, no disclaimers about "
"tools or access. Your response is private guidance handed to the "
"aggregator, not an answer shown to the user. NEVER claim to have executed "
"anything."
)
def _slot_label(slot: dict[str, Any]) -> str:
label = f"{(slot.get('provider') or '').strip()}:{(slot.get('model') or '').strip()}"
effort = str(slot.get("reasoning_effort") or "").strip()
return f"{label}[reasoning={effort}]" if effort else label
def _slot_reasoning_config(slot: dict[str, Any]) -> dict[str, Any] | None:
"""Translate optional per-MoA-slot reasoning_effort into runtime config."""
effort = slot.get("reasoning_effort")
try:
from hermes_constants import parse_reasoning_effort
return parse_reasoning_effort(effort)
except Exception: # pragma: no cover - defensive; bad config must not break MoA
return None
def _aggregator_reasoning_config(aggregator: dict[str, Any]) -> dict[str, Any] | None:
"""Resolve the aggregator's reasoning config: slot > per-model > global.
The aggregator is the ACTING model, so it falls back through the shared
``resolve_reasoning_config`` chokepoint (#64187). References deliberately do
not: inheriting a global ``xhigh`` into every advisor would multiply cost.
"""
cfg = _slot_reasoning_config(aggregator)
if cfg is not None:
return cfg
try:
from hermes_cli.config import load_config
from hermes_constants import resolve_reasoning_config
return resolve_reasoning_config(
load_config() or {}, str(aggregator.get("model") or "")
)
except Exception: # pragma: no cover - defensive; bad config must not break MoA
return None
def _slot_runtime(slot: dict[str, Any]) -> dict[str, Any]:
"""Resolve a slot to ``call_llm`` kwargs via ``resolve_runtime_provider``.
Gives the slot its provider's real api_mode/base_url/api_key instead of letting
the auxiliary auto-detector guess. Falls back to bare provider/model on error
(uncached). Cached per (provider, model) with a short TTL (#66793).
"""
provider = str(slot.get("provider") or "").strip()
model = str(slot.get("model") or "").strip()
cache_key = (provider, model)
now = time.monotonic()
with _runtime_cache_lock:
entry = _runtime_cache.get(cache_key)
if entry is not None:
stamped_at, cached = entry
if now - stamped_at < _RUNTIME_CACHE_TTL_SECONDS:
return cached
out: dict[str, Any] = {"provider": provider, "model": model}
try:
from hermes_cli.runtime_provider import resolve_runtime_provider
rt = resolve_runtime_provider(requested=provider, target_model=model)
out.update({k: rt[k] for k in ("base_url", "api_key", "api_mode") if rt.get(k)})
request_overrides = rt.get("request_overrides")
if isinstance(request_overrides, dict):
extra_body = request_overrides.get("extra_body")
if isinstance(extra_body, dict) and extra_body:
out["extra_body"] = dict(extra_body)
except Exception as exc: # pragma: no cover - defensive
logger.debug("MoA slot runtime resolution failed for %s: %s",
_slot_label(slot), exc)
# Never cache a fallback result: a transient error would pin bare kwargs for a TTL.
return out
with _runtime_cache_lock:
_runtime_cache[cache_key] = (now, out)
return out
def _merge_slot_extra_body(
slot_extra_body: Any,
caller_extra_body: Any,
) -> Any:
"""Merge slot defaults with a caller override for ``call_llm``."""
if isinstance(slot_extra_body, dict) and slot_extra_body:
if isinstance(caller_extra_body, dict):
return {**slot_extra_body, **caller_extra_body}
if caller_extra_body:
return caller_extra_body
return dict(slot_extra_body)
return caller_extra_body
def _maybe_apply_moa_cache_control(
messages: list[dict[str, Any]],
runtime: dict[str, Any],
*,
cache_disabled: bool | None = None,
cache_ttl: str | None = None,
) -> list[dict[str, Any]]:
"""Apply cache_control to an advisor/aggregator request when its route honors it.
Same policy function and marker helper as the main loop; MoA has no static
prefix so the legacy system-and-3 fallback is used. ``cache_disabled`` is
stamped onto the stub so ``cache_ttl: off`` is honored (#76085); ``cache_ttl``
threads the agent's tier, clamped per destination (#84733). Returns the
messages unchanged on any error.
"""
try:
from agent.agent_runtime_helpers import (
anthropic_prompt_cache_policy,
blank_cache_policy_stub,
)
from agent.prompt_caching import (
apply_anthropic_cache_control,
effective_cache_ttl,
envelope_tool_part_cache_markers_supported,
)
# Explicit kwarg > runtime snapshot (threaded from the live agent) > config.
if cache_disabled is None and "_cache_disabled" in runtime:
cache_disabled = runtime.get("_cache_disabled")
# blank_cache_policy_stub is the only sanctioned stub (carries _cache_disabled, #76085).
stub = blank_cache_policy_stub(cache_disabled)
should_cache, native_layout = anthropic_prompt_cache_policy(
stub,
provider=runtime.get("provider") or "",
base_url=runtime.get("base_url") or "",
api_mode=runtime.get("api_mode") or "",
model=runtime.get("model") or "",
)
if not should_cache:
return messages
return apply_anthropic_cache_control(
messages,
cache_ttl=effective_cache_ttl(
# None → "5m"; cache-disabled routes already returned above.
cache_ttl,
provider=runtime.get("provider") or "",
model=runtime.get("model") or "",
),
native_anthropic=native_layout,
# Envelope routes reject part-level markers in tool_result.content[] (#89886).
tool_part_markers=envelope_tool_part_cache_markers_supported(
runtime.get("provider") or "", runtime.get("base_url") or ""
),
)
except Exception as exc: # pragma: no cover - decoration must never break a call
logger.debug("MoA cache_control decoration skipped: %s", exc)
return messages
def _price_reference_response(
response: Any, slot: dict[str, Any], runtime: dict[str, Any]
) -> tuple[Any, Any, str | None, str | None]:
"""Normalize a reference response's usage with the slot's OWN provider/api_mode and
price it at its own rate (hence fan-out cost is summed in dollars). Never raises."""
from agent.usage_pricing import estimate_usage_cost, normalize_usage
usage = CanonicalUsage()
raw_usage = getattr(response, "usage", None)
if raw_usage:
try:
usage = normalize_usage(
raw_usage, provider=runtime.get("provider"), api_mode=runtime.get("api_mode"),
)
except Exception: # pragma: no cover - defensive
usage = CanonicalUsage()
try:
cost = estimate_usage_cost(
slot.get("model") or "",
usage,
provider=runtime.get("provider"),
base_url=runtime.get("base_url"),
api_key=runtime.get("api_key"),
)
return usage, cost.amount_usd, cost.status, cost.source
except Exception: # pragma: no cover - defensive
return usage, None, None, None
def _run_reference(
slot: dict[str, Any],
ref_messages: list[dict[str, Any]],
*,
temperature: float | None = None,
max_tokens: int | None = None,
reference_timeout: float | None = None,
context_length_cache: Any = None,
cache_disabled: bool | None = None,
cache_ttl: str | None = None,
) -> tuple[str, str, Any]:
"""Call one reference model; return ``(label, text, accounting)``. Never raises.
The slot is resolved to its provider's real runtime and called through the
same ``call_llm`` path any model uses. Usage is normalized with the slot's OWN
provider/api_mode and priced at its own rate. A failed reference becomes a
labelled ``[failed: …]`` note. Runs inside a thread pool (call_llm is blocking).
"""
label = _slot_label(slot)
runtime = _slot_runtime(slot)
trace_fields = {
"model": slot.get("model"),
"provider": runtime.get("provider") or slot.get("provider"),
"temperature": temperature,
}
try:
# The trimmed view already stripped the agent's system prompt; this is the only one.
messages = [{"role": "system", "content": _REFERENCE_SYSTEM_PROMPT}, *ref_messages]
# Trim to THIS model's window (advisors may be smaller than the aggregator, #60345);
# estimated after the system prompt is prepended so it counts too.
messages = _trim_messages_for_reference(
messages,
slot,
runtime,
reserve_output_tokens=max_tokens,
context_length_cache=context_length_cache,
)
# Anthropic-style caching is opt-in per request; the advisory view is append-only
# across iterations, so decorating lets iteration N+1 replay N's cached prefix.
# Pin the live agent disable onto the runtime (not a fresh config read, #76085).
cache_runtime = runtime
if cache_disabled is not None:
cache_runtime = {**runtime, "_cache_disabled": cache_disabled}
messages = _maybe_apply_moa_cache_control(
messages, cache_runtime, cache_ttl=cache_ttl
)
# Per-slot max_tokens beats the preset-level reference_max_tokens.
_slot_max_tokens: int | None = slot.get("max_tokens")
_effective_max_tokens = _slot_max_tokens if _slot_max_tokens is not None else max_tokens
extra_headers = None
# Normalize provider aliases (github, github-copilot, ...) via the canonical table.
from agent.auxiliary_client import _normalize_aux_provider
if _normalize_aux_provider(str(runtime.get("provider") or "")) in (
"copilot",
"copilot-acp",
):
# Copilot gates premium models on request attribution; MoA fan-out serves the
# user's current turn, so mirror the main agent's x-initiator header.
extra_headers = {"x-initiator": "user"}
response = call_llm(
task="moa_reference",
messages=messages,
temperature=temperature,
max_tokens=_effective_max_tokens,
timeout=reference_timeout,
reasoning_config=_slot_reasoning_config(slot),
extra_headers=extra_headers,
**runtime,
)
_output_text = _extract_text(response) or "(empty response)"
acct = _RefAccounting(
*_price_reference_response(response, slot, runtime),
messages=messages,
output=_output_text,
**trace_fields,
)
return label, _output_text, acct
except Exception as exc:
logger.warning("MoA reference model %s failed: %s", label, exc)
return label, f"[failed: {exc}]", _RefAccounting(
CanonicalUsage(),
messages=[{"role": "system", "content": _REFERENCE_SYSTEM_PROMPT}, *ref_messages],
output=f"[failed: {exc}]",
**trace_fields,
)
# Output headroom reserved in the reference window when reference_max_tokens is unset.
_REFERENCE_DEFAULT_OUTPUT_RESERVE = 8192
# Estimator slack: estimate_messages_tokens_rough is a rough chars/4 heuristic.
_REFERENCE_TRIM_SAFETY_FRACTION = 0.10
def _trim_messages_for_reference(
messages: list[dict[str, Any]],
slot: dict[str, str],
runtime: dict[str, Any],
*,
reserve_output_tokens: int | None = None,
context_length_cache: Any = None,
) -> list[dict[str, Any]]:
"""Trim an advisory request to fit a reference model's context window (#60345).
``messages`` is the full request (advisory system prompt included). Budget =
window minus ``reserve_output_tokens`` (or a default) minus a safety fraction.
Drops the OLDEST frames after the system prompt while keeping: the system
prompt, a user-first body, and the trailing user turn plus one preceding turn
(even if still over budget). ``context_length_cache`` memoizes the window per
(provider, model) for the fan-out; unresolvable windows leave messages unchanged.
"""
if not messages:
return messages
from agent.model_metadata import (
estimate_messages_tokens_rough,
get_model_context_length,
)
model = str(slot.get("model") or "")
provider = str(runtime.get("provider") or slot.get("provider") or "")
if not model:
return messages
cache_key = (provider, model)
context_length: int | None = None
if isinstance(context_length_cache, dict) and cache_key in context_length_cache:
context_length = context_length_cache[cache_key]
else:
try:
context_length = get_model_context_length(
model=model,
base_url=str(runtime.get("base_url") or ""),
api_key=str(runtime.get("api_key") or ""),
provider=provider,
)
except Exception:
logger.debug(
"MoA reference context-length resolution failed for %s",
_slot_label(slot),
)
context_length = None
if isinstance(context_length_cache, dict):
# Cache failures too so a flaky metadata source is not re-probed per reference.
context_length_cache[cache_key] = context_length
if not isinstance(context_length, int) or context_length <= 0:
return messages
reserve = (
int(reserve_output_tokens)
if isinstance(reserve_output_tokens, int) and reserve_output_tokens > 0
else _REFERENCE_DEFAULT_OUTPUT_RESERVE
)
budget = int(context_length * (1.0 - _REFERENCE_TRIM_SAFETY_FRACTION)) - reserve
if budget <= 0:
return messages
estimated = estimate_messages_tokens_rough(messages)
if estimated <= budget:
return messages
has_system = bool(messages) and messages[0].get("role") == "system"
head = [messages[0]] if has_system else []
body = list(messages[1:] if has_system else messages)
# Keep the trailing user turn plus at least one preceding turn.
while len(body) > 2 and estimate_messages_tokens_rough(head + body) > budget:
body.pop(0)
# Preserve the user-first invariant after each pop.
while len(body) > 2 and body[0].get("role") == "assistant":
body.pop(0)
# Two frames left with an assistant first: still enforce user-first.
while len(body) > 1 and body[0].get("role") == "assistant":
body.pop(0)
trimmed = head + body
dropped = len(messages) - len(trimmed)
if dropped:
logger.info(
"MoA reference %s: estimated %d tokens exceeds budget %d "
"(window %d, output reserve %d); dropped %d oldest message(s).",
_slot_label(slot),
estimated,
budget,
context_length,
reserve,
dropped,
)
return trimmed
_REFERENCE_POLL_INTERVAL_S = 5.0
# Sentinel for a reference aborted by user interrupt; the facade must never cache it.
_INTERRUPTED_REFERENCE_NOTE = "[skipped: interrupted by user]"
def _run_references_parallel(
reference_models: list[dict[str, Any]],
ref_messages: list[dict[str, Any]],
*,
temperature: float | None = None,
max_tokens: int | None = None,
progress_callback: Any = None,
reference_timeout: float | None = None,
agent: Any = None,
late_accounting_sink: Any = None,
) -> list[tuple[str, str, Any]]:
"""Fan out all reference models in parallel; outputs are in ``reference_models`` order.
Slots with ``provider == "moa"`` are skipped with a labelled note (recursion
guard). ``progress_callback(refs_done, refs_total, label)`` fires per completion
(best-effort). Each element is ``(label, text, _RefAccounting)``.
With *agent*, the wait polls every ``_REFERENCE_POLL_INTERVAL_S`` so a user
interrupt can abort it (like tool_executor's batch). In-flight calls cannot be
killed; their eventual accounting goes to ``late_accounting_sink``.
"""
if not reference_models:
return []
results: list[tuple[str, str, Any] | None] = [None] * len(reference_models)
futures: dict[Any, int] = {}
workers = min(_MAX_REFERENCE_WORKERS, len(reference_models))
# Executor threads start with an empty contextvars.Context; propagate the turn's
# (approval callbacks + Nous Portal conversation tag) into each worker.
from tools.thread_context import propagate_context_to_thread
total = len(reference_models)
completed = 0
executor = ThreadPoolExecutor(max_workers=workers)
interrupted = False
# Shared per-fan-out context-length cache (dict get/set is GIL-atomic).
_ctx_len_cache: dict[tuple[str, str], int | None] = {}
cache_disabled = (
getattr(agent, "_cache_disabled", None) if agent is not None else None
)
# Agent's cache TTL for every advisor request (#84733); clamped in the decorator.
cache_ttl = getattr(agent, "_cache_ttl", None) if agent is not None else None
try:
for idx, slot in enumerate(reference_models):
if slot.get("provider") == "moa":
results[idx] = (
_slot_label(slot),
"[skipped: MoA presets cannot recursively reference MoA]",
_RefAccounting(CanonicalUsage()),
)
continue
futures[
executor.submit(
propagate_context_to_thread(_run_reference),
slot,
ref_messages,
temperature=temperature,
max_tokens=max_tokens,
reference_timeout=reference_timeout,
context_length_cache=_ctx_len_cache,
cache_disabled=cache_disabled,
cache_ttl=cache_ttl,
)
] = idx
# Collect every reference (no early exit except a user interrupt); progress
# callbacks fire per completion.
pending = set(futures)
while pending:
done, pending = _futures_wait(pending, timeout=_REFERENCE_POLL_INTERVAL_S)
for future in done:
idx = futures[future]
results[idx] = future.result()
completed += 1
if progress_callback is not None:
try:
label = _slot_label(reference_models[idx])
progress_callback(completed, total, label)
except Exception as exc: # pragma: no cover - display must never break
logger.debug("MoA progress_callback failed: %s", exc)
if not pending:
break
if agent is not None and getattr(agent, "_interrupt_requested", False):
interrupted = True
break
if interrupted:
for future, idx in futures.items():
if results[idx] is not None:
continue
if future.cancel():
# Never dispatched — nothing was billed.
results[idx] = (
_slot_label(reference_models[idx]),
_INTERRUPTED_REFERENCE_NOTE,
_RefAccounting(CanonicalUsage()),
)
elif future.done():
# Finished between the interrupt check and now: keep its real output/accounting.
results[idx] = future.result()
else:
# Already running — cannot be killed; the provider WILL bill when it
# completes, so hand its eventual accounting to the caller's sink.
label = _slot_label(reference_models[idx])
results[idx] = (
label,
_INTERRUPTED_REFERENCE_NOTE,
_RefAccounting(CanonicalUsage()),
)
if late_accounting_sink is not None:
def _record_late(f: Any, _label: str = label) -> None:
try:
_lbl, _txt, _acct = f.result()
except Exception: # pragma: no cover - defensive
return
try:
late_accounting_sink(_label, _acct)
except Exception: # pragma: no cover - defensive
logger.debug(
"MoA: late accounting sink failed for %s",
_label,
)
future.add_done_callback(_record_late)
finally:
executor.shutdown(wait=not interrupted, cancel_futures=interrupted)
return [r for r in results if r is not None]
def _truncate_tool_result(text: str, budget: int = _REFERENCE_TOOL_RESULT_BUDGET) -> str:
"""Head+tail preview of a tool result with an ``[... N chars omitted ...]`` marker."""
if not text or len(text) <= budget:
return text
half = budget // 2
omitted = len(text) - 2 * half
return f"{text[:half]}\n[... {omitted} chars omitted ...]\n{text[-half:]}"
def _field(obj: Any, name: str) -> Any:
"""Read ``name`` from a dict or an attribute object (tool calls arrive as either)."""
if obj is None:
return None
return obj.get(name) if isinstance(obj, dict) else getattr(obj, name, None)
def _render_tool_calls(tool_calls: Any) -> str:
"""Render an assistant turn's tool_calls as ``[called tool: name(args)]`` lines.
Tolerates dict- and SimpleNamespace-shaped entries (and nested ``function``).
"""
lines: list[str] = []
for tc in tool_calls or []:
fn = _field(tc, "function")
name = _field(fn, "name") or _field(tc, "name") or "tool"
fn_args = _field(fn, "arguments")
if isinstance(fn_args, str):
args_text = fn_args
elif fn_args is not None:
try:
args_text = json.dumps(fn_args, ensure_ascii=False)
except Exception:
args_text = str(fn_args)
else:
args_text = ""
lines.append(f"[called tool: {name}({args_text})]" if args_text else f"[called tool: {name}]")
return "\n".join(lines)
_ADVISORY_INSTRUCTION = (
"[The conversation above is the current state of the task. Give your "
"most intelligent judgement: what is going on, what should happen next, "
"what risks or mistakes you see, and how the acting agent should "
"proceed.]"
)
def _reference_messages(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
"""Build the advisory (reference-model) view of the conversation.
Flattens the transcript to plain user/assistant TEXT turns: the system prompt
is dropped, tool_calls are rendered inline, and tool results are folded into
the preceding assistant turn as ``[tool result: ...]`` previews. Zero tool-role
messages / tool_calls arrays are emitted, so strict providers do not 400.
The view always ends on a ``user`` turn (Anthropic treats a trailing assistant
turn as prefill): a synthetic advisory request is APPENDED rather than
deleting context. The aggregator always receives the full transcript.
"""
rendered: list[dict[str, Any]] = []
last_user_content: str | None = None
for msg in messages:
role = msg.get("role")
content = msg.get("content")
# Content may be a list (cache_control-decorated text parts, multimodal turns);
# flatten_message_text extracts text so decorated and undecorated transcripts
# yield a byte-identical view (stable advisory prefix for advisor caching).
text = flatten_message_text(content)
if role == "system":
continue
if role == "user":
if not text.strip() and isinstance(content, list) and content:
# Structured content with no text (e.g. image-only): an empty user
# message is rejected by strict providers and skipping breaks alternation.
text = "[user sent non-text content (e.g. an image attachment)]"
if not text.strip():
# Genuinely empty user turn: strict providers (Kimi, ZAI) 400 on it; dropping
# is safe because the advisory view is not strictly alternating anyway.
continue
last_user_content = text
rendered.append({"role": "user", "content": text})
elif role == "assistant":
parts: list[str] = []
if text.strip():
parts.append(text.strip())
calls_text = _render_tool_calls(msg.get("tool_calls"))
if calls_text:
parts.append(calls_text)
# Empty assistant turns (no text, no calls) carry nothing advisory.
if parts:
rendered.append({"role": "assistant", "content": "\n".join(parts)})
elif role == "tool":
# Fold the tool result into the preceding assistant turn as text.
result_text = _truncate_tool_result(text)
block = f"[tool result: {result_text}]"
if rendered and rendered[-1].get("role") == "assistant":
rendered[-1]["content"] = rendered[-1]["content"] + "\n" + block
else:
# No assistant turn to attach to (e.g. a leading tool result).
rendered.append({"role": "assistant", "content": block})
# Any other role is ignored.
# End on a user turn by appending a synthetic advisory request (Anthropic
# rejects trailing assistant prefill); an existing trailing user turn is left as is.
if rendered and rendered[-1].get("role") == "assistant":
rendered.append({"role": "user", "content": _ADVISORY_INSTRUCTION})
if not rendered:
# Nothing rendered: fall back to the latest user turn.
if last_user_content is not None:
return [{"role": "user", "content": last_user_content}]
for msg in reversed(messages):
if msg.get("role") == "user":
fallback_text = flatten_message_text(msg.get("content"))
if fallback_text.strip():
return [{"role": "user", "content": fallback_text}]
return rendered
def _extract_text(response: Any) -> str:
try:
transport = get_transport("chat_completions")
if transport is None:
raise RuntimeError("chat_completions transport unavailable")
normalized = transport.normalize_response(response)
text = (normalized.content or "").strip()
if text:
return text
except Exception:
pass
try:
message = response.choices[0].message
content = message.get("content") if isinstance(message, dict) else getattr(message, "content", message)
if not isinstance(content, str):
content = str(content) if content else ""
return content.strip()
except Exception:
return ""
def _preset_temperature(preset: dict[str, Any], key: str) -> float | None:
"""Read an optional preset temperature; None (absent/empty/null) = provider default.
A hardcoded fallback previously forced 0.6/0.4 even when unset.
"""
value = preset.get(key)
if value is None or (isinstance(value, str) and not value.strip()):
return None
try:
return float(value)
except (TypeError, ValueError):
logger.warning("ignoring non-numeric %s=%r in MoA preset", key, value)
return None
def _hash_messages(msgs: list[dict[str, Any]]) -> str:
return hashlib.sha256(
"\u0000".join(f"{m.get('role')}:{m.get('content')}" for m in msgs).encode("utf-8", "replace")
).hexdigest()
def _is_failed_reference(text: str) -> bool:
"""Whether a reference output is a ``[failed: …]`` / ``[skipped: …]`` sentinel."""
sentinel = text.lstrip().lower()
return sentinel.startswith("[failed:") or sentinel.startswith("[skipped:")
def _successful_references(
reference_outputs: list[tuple[str, str, Any]],
) -> list[tuple[str, str, Any]]:
"""Filter failed advice while preserving each accounting payload."""
return [output for output in reference_outputs if not _is_failed_reference(output[1])]
def _failed_reference_labels(
reference_outputs: list[tuple[str, str, Any]],
) -> list[str]:
return [label for label, text, _accounting in reference_outputs if _is_failed_reference(text)]
def _join_reference_outputs(outputs: list[tuple[str, str, Any]], degraded: str = "") -> str:
"""Render numbered reference blocks for the aggregator, appending any degraded notice."""
joined = "\n\n".join(
f"Reference {idx} — {label}:\n{text}"
for idx, (label, text, _acct) in enumerate(outputs, start=1)
)
if degraded:
joined = f"{joined}\n\n{degraded}" if joined else degraded
return joined
def _sum_reference_accounting(outputs: list[tuple[str, str, Any]]) -> tuple[Any, Any]:
"""Sum fan-out usage AND cost in dollars (each advisor priced at its OWN rate)."""
usage = CanonicalUsage()
cost: Any = None
for _lbl, _txt, acct in outputs:
if isinstance(acct, _RefAccounting):
if isinstance(acct.usage, CanonicalUsage):
usage = usage + acct.usage
if acct.cost_usd is not None:
cost = (cost or 0) + acct.cost_usd
return usage, cost
def _degraded_notice(failed_labels: list[str], policy: str) -> str:
if not failed_labels or policy.strip().lower() == "silent":
return ""
return f"[Reference models unavailable: {', '.join(failed_labels)}]"
def aggregate_moa_context(
*,
user_prompt: str,
api_messages: list[dict[str, Any]],
reference_models: list[dict[str, Any]],
aggregator: dict[str, Any],
temperature: float | None = None,
aggregator_temperature: float | None = None,
reference_max_tokens: int | None = None,
reference_timeout: float | None = None,
degraded_reference_policy: str = "loud",
agent: Any = None,
) -> str:
"""Run configured reference models and synthesize their advice (one-shot /moa).
Failures become model-specific notes instead of aborting the loop.
``reference_max_tokens`` caps ONLY the fan-out — capping the aggregator
truncated long syntheses (#53580). ``temperature`` / ``aggregator_temperature``
default to None (provider default). ``agent`` makes the fan-out interruptible.
"""
reference_models = [slot for slot in reference_models if slot.get("enabled", True)]
reference_outputs: list[tuple[str, str, Any]] = []
ref_messages = _reference_messages(api_messages)
reference_outputs = _run_references_parallel(
reference_models,
ref_messages,
temperature=temperature,
max_tokens=reference_max_tokens,
reference_timeout=reference_timeout,
agent=agent,
)
successful_outputs = _successful_references(reference_outputs)
failed_labels = _failed_reference_labels(reference_outputs)
# 'full' privacy mode also redacts advisor text before it reaches this synthesizer
# ('display' has no surface here). Failed refs are already filtered out.
try:
from hermes_cli.config import load_config as _load_config
if _moa_privacy_mode((_load_config() or {}).get("moa")) == "full":
successful_outputs = _redact_reference_outputs(successful_outputs)
except Exception: # pragma: no cover - privacy filter must never break a turn
logger.debug("MoA privacy filter check failed", exc_info=True)
degraded = _degraded_notice(failed_labels, degraded_reference_policy)
joined = _join_reference_outputs(successful_outputs, degraded)
# Every reference failed: skip the aggregator (synthesizing over nothing can block
# for the full provider timeout) and return only the sanitized notice.
if reference_outputs and not successful_outputs:
logger.warning(
"MoA: all %d reference(s) failed — skipping aggregator synthesis",
len(reference_outputs),
)
notice = degraded or "[Reference models unavailable]"
return (
"[Mixture of Agents context — all reference models failed. "
"Proceeding without aggregated guidance.]\n"
f"References: {', '.join(_slot_label(slot) for slot in reference_models)}\n\n"
f"{notice}"
)
synth_prompt = (
"You are the aggregator in a Mixture of Agents process. Synthesize the "
"reference responses into concise, actionable guidance for the main "
"Hermes agent. Focus on next steps, tool-use strategy, risks, and any "
"disagreements. Do not answer the user directly unless that is all that "
"is needed; produce context the main agent should use in its normal loop.\n\n"
f"Original user prompt:\n{user_prompt}\n\n"
f"Reference responses:\n{joined}"
)
agg_label = _slot_label(aggregator)
agg_runtime = _slot_runtime(aggregator)
# Pin the live agent disable onto synthesis decoration (#76085); None is a no-op.
agg_cache_runtime = agg_runtime
_agg_cache_disabled = (
getattr(agent, "_cache_disabled", None) if agent is not None else None
)
if _agg_cache_disabled is not None:
agg_cache_runtime = {
**agg_runtime,
"_cache_disabled": _agg_cache_disabled,
}
# Agent's cache TTL for the one-shot synthesis path (#84733).
_agg_cache_ttl = getattr(agent, "_cache_ttl", None) if agent is not None else None
try:
# Same cache_control decoration as the advisor calls; this synthesis call is
# a third independent MoA call path that otherwise re-bills its full input.
agg_messages = _maybe_apply_moa_cache_control(
[{"role": "user", "content": synth_prompt}],
agg_cache_runtime,
cache_ttl=_agg_cache_ttl,
)
response = call_llm(
task="moa_aggregator",
messages=agg_messages,
temperature=aggregator_temperature,
reasoning_config=_aggregator_reasoning_config(aggregator),
**agg_runtime,
)
synthesis = _extract_text(response)
except Exception as exc:
logger.warning("MoA aggregator model %s failed: %s", agg_label, exc)
synthesis = ""
if not synthesis:
synthesis = joined
return (
"[Mixture of Agents context — use this as private guidance for the "
"normal Hermes agent loop. You may call tools, continue reasoning, or "
"finish normally.]\n"
f"Aggregator: {agg_label}\n"
f"References: {', '.join(_slot_label(slot) for slot in reference_models)}\n\n"
f"{synthesis.strip()}"
)
def _completed_response_as_stream_chunk(response: Any) -> Any:
"""Adapt a completed response (``choices[0].message``) into one delta stream chunk.
Done at the MoA facade boundary so transports stay untouched.
"""
choices = getattr(response, "choices", None)
first_choice = choices[0] if isinstance(choices, (list, tuple)) and choices else None
message = getattr(first_choice, "message", None)
raw_tool_calls = getattr(message, "tool_calls", None)
tool_call_deltas = None
if isinstance(raw_tool_calls, (list, tuple)) and raw_tool_calls:
tool_call_deltas = []
for index, tc in enumerate(raw_tool_calls):
function = getattr(tc, "function", None)
tool_call_deltas.append(SimpleNamespace(
index=getattr(tc, "index", index),
id=getattr(tc, "id", None),
type=getattr(tc, "type", None) or "function",
function=SimpleNamespace(
name=getattr(function, "name", None),
arguments=getattr(function, "arguments", None),
),
))
delta = SimpleNamespace(
content=getattr(message, "content", None),
tool_calls=tool_call_deltas,
reasoning_content=getattr(message, "reasoning_content", None),
reasoning=getattr(message, "reasoning", None),
reasoning_details=getattr(message, "reasoning_details", None),
)
choice = SimpleNamespace(
index=getattr(first_choice, "index", 0),
delta=delta,
finish_reason=getattr(first_choice, "finish_reason", None) or "stop",
)
return SimpleNamespace(
id=getattr(response, "id", None),
model=getattr(response, "model", None),
choices=[choice],
usage=getattr(response, "usage", None),
)
def _attach_reference_guidance(agg_messages: list[dict[str, Any]], guidance: str) -> None:
"""Attach the per-turn reference block at the END of the aggregator prompt.
The block varies per iteration; merging it into the (early) original user
message would diverge the prompt prefix and re-prefill the whole conversation
each step. Appending keeps ``[system][task][tool-history]`` cache-stable. A
trailing user turn is merged in place (string or content-part list — a new text
part rides AFTER the cache_control-marked part); otherwise a user message is
appended (two consecutive user turns would be rejected by strict providers).
"""
last = agg_messages[-1] if agg_messages else None
if last is not None and last.get("role") == "user":
last_content = last.get("content")
if isinstance(last_content, str):
last["content"] = last_content + "\n\n" + guidance
return
if isinstance(last_content, list):
last["content"] = [*last_content, {"type": "text", "text": "\n\n" + guidance}]
return
agg_messages.append({"role": "user", "content": guidance})
def peel_reference_guidance(
messages: list[dict[str, Any]],
guidance: Any,
) -> list[dict[str, Any]]:
"""Exact inverse of ``_attach_reference_guidance`` (the three attach shapes).
Used by the failover redecoration chokepoint so a cache breakpoint never lands
on the turn-varying guidance (#72626). Returns a new list; inputs are not mutated.
"""
if not guidance or not messages:
return messages
guidance_text = str(guidance)
last = messages[-1]
if not isinstance(last, dict) or last.get("role") != "user":
return messages
content = last.get("content")
if content == guidance_text:
# Attach shape (c): guidance was appended as its own user message.
return list(messages[:-1])
suffix = "\n\n" + guidance_text
if isinstance(content, str) and content.endswith(suffix):
# Attach shape (a): merged into a trailing string user turn.
peeled = dict(last)
peeled["content"] = content[: -len(suffix)]
return [*messages[:-1], peeled]
if isinstance(content, list) and content:
last_part = content[-1]
if isinstance(last_part, dict) and last_part.get("type", "text") == "text":
text = last_part.get("text") or ""
if text == suffix or text == guidance_text:
# Attach shape (b): guidance rode as its own trailing part.
peeled = dict(last)
peeled["content"] = list(content[:-1])
if not peeled["content"]:
# Guidance was the only content: drop the whole message (mirrors shape c).
return list(messages[:-1])
return [*messages[:-1], peeled]
if text.endswith(suffix):
new_part = dict(last_part)
new_part["text"] = text[: -len(suffix)]
peeled = dict(last)
peeled["content"] = [*content[:-1], new_part]
return [*messages[:-1], peeled]
return messages
class MoAChatCompletions:
"""OpenAI-chat-compatible facade where the aggregator is the acting model."""
def __init__(self, preset_name: str, reference_callback: Any = None, agent: Any = None):
self.preset_name = preset_name or "default"
# Optional best-effort display hook, called as ``reference_callback(event, **kwargs)``:
# "moa.reference" index, count, label, text
# "moa.progress" refs_done, refs_total, label (per reference completion)
# "moa.phase" phase, refs_done, refs_total, aggregator
# "moa.aggregating" aggregator (label), ref_count
self.reference_callback = reference_callback
# Owning AIAgent (optional); lets the fan-out check agent._interrupt_requested.
self._agent = agent
# State-scoped reference cache keyed on the advisory-view signature: a new
# user/tool message is a MISS (references re-run), a redundant create() with
# identical state is a HIT (no re-run, no re-emit).
self._ref_cache_key: tuple | None = None
self._ref_cache_outputs: list[tuple[str, str, Any]] = []
# Fan-out usage/cost from the latest cache-MISS create(), awaiting
# consume_reference_usage (zero deposited on a HIT so spend counts once).
self._pending_reference_usage: Any = CanonicalUsage()
self._pending_reference_cost: Any = None
# Guards pending usage/cost against late-accounting callbacks on worker threads.
self._accounting_lock = threading.Lock()
# Resolved aggregator slot from the latest create(); cost accounting prices the
# acting turn at its real model instead of the virtual preset name.
self.last_aggregator_slot: Any = None
# Full-turn trace parts from a cache-MISS create(), flushed by consume_and_save_trace.
self._pending_trace: Any = None
# Per-advisor metrics for observability hooks; NOT consumed (post_api_request
# fires on a different branch than consume_and_save_trace).
self._last_reference_metrics: Any = None
# every_n cadence state, scoped to a single USER TURN (resets on a new user
# message) so iteration 1 of every turn is on-cadence.
self._fanout_iteration_count = 0
self._fanout_turn_sig: str | None = None
self._fanout_last_state_sig: str | None = None
# Normalized moa.privacy_filter mode ('' | 'display' | 'full'), refreshed per create().
self._privacy_mode: str = ""
def consume_reference_usage(self) -> tuple[Any, Any]:
"""Pop pending fan-out ``(CanonicalUsage, cost_usd_or_None)`` and reset both.
Clearing prevents a streaming retry re-entering accounting from double-counting.
"""
with self._accounting_lock:
usage = self._pending_reference_usage or CanonicalUsage()
cost = self._pending_reference_cost
self._pending_reference_usage = CanonicalUsage()
self._pending_reference_cost = None
return usage, cost
def last_reference_metrics(self) -> Any:
"""Per-advisor metrics from the most recent fan-out, or None (read-only)."""
return self._last_reference_metrics
def _record_late_reference_accounting(self, label: str, accounting: Any) -> None:
"""Fold a late-completing interrupted reference's real spend into pending totals.
Registered as a done-callback on abandoned futures (they still bill). Thread-safe.
"""
if not isinstance(accounting, _RefAccounting):
return
self._fold_pending_accounting(*_sum_reference_accounting([(label, "", accounting)]))
logger.debug(
"MoA: recorded late accounting for interrupted reference %s", label
)
def _fold_pending_accounting(self, usage: Any, cost: Any) -> None:
"""Add (never overwrite) fan-out spend so late interrupted-reference deposits survive."""
with self._accounting_lock:
self._pending_reference_usage = (
self._pending_reference_usage or CanonicalUsage()
) + usage
if cost is not None:
self._pending_reference_cost = (self._pending_reference_cost or 0) + cost
def consume_and_save_trace(
self, session_id: Any = None, aggregator_output_fallback: Any = None
) -> None:
"""Flush the pending full-turn trace to disk (no-op when nothing is pending).
``aggregator_output_fallback`` is the caller's resolved acting text: on the
streaming path the output could not be captured at ``create()`` time, so it
is folded in here. Clears the pending trace; never raises.
"""
pending = self._pending_trace
self._pending_trace = None
if not pending or "aggregator_input_messages" not in pending:
return
try:
from agent.moa_trace import save_moa_turn
agg_slot = pending.get("aggregator_slot") or {}
# Inline capture (non-streaming) beats the caller's streamed text.
agg_output = pending.get("aggregator_output")
if agg_output is None and aggregator_output_fallback:
agg_output = aggregator_output_fallback
save_moa_turn(
session_id=session_id,
preset_name=pending.get("preset", ""),
reference_outputs=pending.get("reference_outputs", []),
aggregator_label=pending.get("aggregator_label", ""),
aggregator_model=agg_slot.get("model"),
aggregator_provider=agg_slot.get("provider"),
aggregator_temperature=pending.get("aggregator_temperature"),
aggregator_input_messages=pending.get("aggregator_input_messages"),
aggregator_output=agg_output,
aggregator_streamed=bool(pending.get("aggregator_streamed")),
)
except Exception as exc: # pragma: no cover - tracing must never break a turn
logger.debug("MoA trace flush failed: %s", exc)
def _emit(self, event: str, **kwargs: Any) -> None:
cb = self.reference_callback
if cb is None:
return
try:
cb(event, **kwargs)
except Exception as exc: # pragma: no cover - display must never break the turn
logger.debug("MoA reference_callback failed for %s: %s", event, exc)
def prepare(self, messages: list[dict[str, Any]]) -> dict[str, Any]:
"""Run the advisor fan-out and return the exact aggregator request.
The loop measures this augmented prompt before its compression gate, then
hands the object back to ``create()`` so the fan-out is not repeated.
"""
return self.create(messages=messages, _moa_prepare_only=True)
def rebase_prepared_request(
self, prepared: dict[str, Any], messages: list[dict[str, Any]]
) -> dict[str, Any]:
"""Re-attach already-generated guidance to a rebuilt (compressed) transcript."""
guidance = prepared.get("guidance")
agg_messages = [dict(message) for message in messages]
if guidance:
_attach_reference_guidance(agg_messages, str(guidance))
return {**prepared, "messages": agg_messages}
def _call_prepared_aggregator(
self, prepared: dict[str, Any], api_kwargs: dict[str, Any]
) -> Any:
"""Send an already prepared MoA aggregator request exactly once."""
agg_messages = prepared["messages"]
aggregator = prepared["aggregator"]
aggregator_temperature = prepared["aggregator_temperature"]
if aggregator.get("provider") == "moa":
raise RuntimeError("MoA aggregator cannot be another MoA preset")
agg_kwargs = dict(api_kwargs)
max_tokens: Any = agg_kwargs.get("max_tokens")
tools: Any = agg_kwargs.get("tools")
extra_body: Any = agg_kwargs.get("extra_body")
agg_runtime = _slot_runtime(aggregator)
try:
from agent.agent_runtime_helpers import (
plan_cache_sections_for_destination,
)
guidance = prepared.get("guidance")
planning_messages = agg_messages
if guidance:
planning_messages = peel_reference_guidance(
agg_messages,
str(guidance),
)
# plan_cache_sections_for_destination returns request-local copies.
# Tri-state cache_disabled: facades built via __new__ have no _agent; forcing
# False would suppress the planner's config fallback (#76085).
_agent = getattr(self, "_agent", None)
_cache_disabled = (
getattr(_agent, "_cache_disabled", None)
if _agent is not None
else None
)
agg_messages, tools = plan_cache_sections_for_destination(
planning_messages,
tools,
provider=agg_runtime.get("provider") or "",
base_url=agg_runtime.get("base_url") or "",
api_mode=agg_runtime.get("api_mode") or "",
model=agg_runtime.get("model") or "",
cache_disabled=_cache_disabled,
# Agent TTL + stable system prefix so MoA stops regressing 1h → 5m (#84733).
cache_ttl=getattr(_agent, "_cache_ttl", None),
static_system_prefix=getattr(
_agent, "_cached_system_prompt_static", None
),
)
if guidance:
_attach_reference_guidance(agg_messages, str(guidance))
except Exception as exc: # pragma: no cover - cache planning must not block MoA
# Warning, not debug: this is the aggregator's ONLY decoration path.
logger.warning(
"MoA aggregator cache plan failed — sending undecorated "
"request (cache misses expected): %s", exc,
)
# Record the exact aggregator INPUT into the pending trace; the persisted COPY
# is redacted under any privacy mode while the live input stays raw.
if self._pending_trace is not None:
self._pending_trace["aggregator_input_messages"] = (
_redact_trace_messages([dict(m) for m in agg_messages])
if getattr(self, "_privacy_mode", "")
else agg_messages
)
self._pending_trace["aggregator_label"] = _slot_label(aggregator)
# The aggregator is the acting model: call it through the same request path
# any model uses, with max_tokens passed through (None → model maximum).
# stream=True returns the RAW token stream (consumer reassembles + retries);
# the non-streaming path forwards no stream/stream_options/timeout.
stream = bool(api_kwargs.get("stream"))
stream_kwargs: dict[str, Any] = {}
if stream:
stream_kwargs["stream"] = True
stream_kwargs["stream_options"] = (
api_kwargs.get("stream_options") or {"include_usage": True}
)
# The consumer's stream-read timeout must govern the aggregator stream.
if api_kwargs.get("timeout") is not None:
stream_kwargs["timeout"] = api_kwargs["timeout"]
# Pop the runtime's extra_body and merge with the caller's (caller wins) so the
# explicit kwarg never collides with **agg_runtime.
agg_extra_body = _merge_slot_extra_body(
agg_runtime.pop("extra_body", None),
extra_body,
)
_agg_response = call_llm(
task="moa_aggregator",
messages=agg_messages,
temperature=aggregator_temperature,
max_tokens=max_tokens,
tools=tools,
extra_body=agg_extra_body,
# Same reasoning policy as the direct create() path (#64187).
reasoning_config=_aggregator_reasoning_config(aggregator),
**stream_kwargs,
**agg_runtime,
)
# Non-streaming: capture the aggregator output inline. Streaming: the output
# lands as the turn's assistant message; the trace marks it streamed.
if self._pending_trace is not None:
if stream:
self._pending_trace["aggregator_streamed"] = True
self._pending_trace["aggregator_output"] = None
else:
self._pending_trace["aggregator_streamed"] = False
try:
self._pending_trace["aggregator_output"] = _extract_text(_agg_response)
except Exception: # pragma: no cover - defensive
self._pending_trace["aggregator_output"] = None
if stream and hasattr(_agg_response, "choices"):
# Some adapters (openai-codex Responses) return a completed response even
# when streaming was requested; hand the loop a one-chunk iterator (#55933).
return iter((_completed_response_as_stream_chunk(_agg_response),))
return _agg_response
def _fanout_cache_key(
self,
preset: dict[str, Any],
ref_messages: list[dict[str, Any]],
reference_models: list[dict[str, Any]],
) -> tuple:
"""Compute the turn-scoped reference cache key per the preset's fan-out cadence.
Cadence (#67199, #63393). "user_turn" (default): advisors run once per user
turn — the signature hashes only the prefix up to the LAST USER message, so later
tool iterations are cache HITs. "per_iteration": re-run whenever the advisory
view changes. "every_n:<N>": iteration 1 of a turn, then every Nth; in-between
iterations reuse the last on-cadence guidance (key pinned to that run so the
lookup is a HIT: no advisor calls, no double accounting, no re-emit).
"""
fanout_mode = str(preset.get("fanout") or "user_turn").strip().lower()
every_n = 0
if fanout_mode.startswith("every_n:"):
try:
every_n = int(fanout_mode.split(":", 1)[1])
except (TypeError, ValueError):
every_n = 0
if every_n < 2:
# every_n:1 IS per-iteration (mirrors _coerce_fanout).
fanout_mode = "per_iteration"
sig_messages = ref_messages
turn_prefix = ref_messages
if fanout_mode == "user_turn" or every_n >= 2:
# Last REAL user message: the synthetic _ADVISORY_INSTRUCTION marker must not
# count or the prefix would grow (and re-sign) every iteration.
for _i in range(len(ref_messages) - 1, -1, -1):
_m = ref_messages[_i]
if _m.get("role") == "user" and _m.get("content") != _ADVISORY_INSTRUCTION:
turn_prefix = ref_messages[: _i + 1]
break
if fanout_mode == "user_turn":
sig_messages = turn_prefix
# every_n bookkeeping: advance the counter only when the advisory STATE changed
# (a streaming retry must not consume a cadence slot); reset on a new turn prefix.
if every_n >= 2:
_turn_sig = _hash_messages(turn_prefix)
if _turn_sig != self._fanout_turn_sig:
self._fanout_turn_sig = _turn_sig
self._fanout_iteration_count = 0
self._fanout_last_state_sig = None
_state_sig = _hash_messages(ref_messages)
if _state_sig != self._fanout_last_state_sig:
self._fanout_last_state_sig = _state_sig
self._fanout_iteration_count += 1
# Iteration 1 is on-cadence; then every Nth iteration after it.
_on_cadence = (self._fanout_iteration_count - 1) % every_n == 0
if not _on_cadence and self._ref_cache_outputs:
return self._ref_cache_key
return (
self.preset_name,
_hash_messages(sig_messages),
tuple(_slot_label(s) for s in reference_models),
)
def create(self, **api_kwargs: Any) -> Any:
prepared_request = api_kwargs.pop("_moa_prepared_request", None)
if prepared_request is not None:
if not isinstance(prepared_request, dict):
raise TypeError("_moa_prepared_request must be a dict")
return self._call_prepared_aggregator(prepared_request, api_kwargs)
preset, _moa_raw = _resolve_preset_cached(self.preset_name)
# Remembered on self so _call_prepared_aggregator redacts the trace consistently.
privacy_mode = _moa_privacy_mode(_moa_raw)
self._privacy_mode = privacy_mode
messages = list(api_kwargs.get("messages") or [])
reference_models = [
slot for slot in (preset.get("reference_models") or [])
if slot.get("enabled", True)
]
aggregator = preset.get("aggregator") or {}
# The MoA path's virtual model/provider have no pricing entry; expose the real slot.
self.last_aggregator_slot = dict(aggregator) if aggregator else None
# No output caps by default (None → call_llm omits max_tokens). A preset MAY cap
# ADVISOR output (dominant MoA latency); the acting aggregator is never capped.
reference_max_tokens = preset.get("reference_max_tokens")
# None = provider default (see _preset_temperature).
temperature = _preset_temperature(preset, "reference_temperature")
aggregator_temperature = _preset_temperature(preset, "aggregator_temperature")
# None = inherit auxiliary.moa_reference.timeout via call_llm.
raw_reference_timeout = preset.get("reference_timeout")
reference_timeout = (
float(raw_reference_timeout) if raw_reference_timeout else None
)
degraded_reference_policy = str(
preset.get("degraded_reference_policy") or "loud"
)
if aggregator_temperature is None and api_kwargs.get("temperature") is not None:
# The acting agent's own temperature applies to the aggregator (the acting model).
aggregator_temperature = api_kwargs.get("temperature")
# A disabled preset = "use the aggregator directly".
if not preset.get("enabled", True):
reference_models = []
reference_outputs: list[tuple[str, str, Any]] = []
ref_messages = _reference_messages(messages)
_cache_key = self._fanout_cache_key(preset, ref_messages, reference_models)
_refs_from_cache = _cache_key == self._ref_cache_key and bool(self._ref_cache_outputs)
if _refs_from_cache:
reference_outputs = list(self._ref_cache_outputs)
# Cache HIT: references already ran and were accounted this turn. Deposit
# nothing, but do NOT zero pending totals (a late interrupted reference may
# have deposited real spend). No trace either — a repeat iteration is not a turn.
self._pending_trace = None
else:
# Per-reference progress (``moa.progress``) through the same display hook.
def _progress(done: int, total: int, label: str) -> None:
self._emit(
"moa.progress",
refs_done=done,
refs_total=total,
label=label,
)
reference_outputs = _run_references_parallel(
reference_models,
ref_messages,
temperature=temperature,
max_tokens=reference_max_tokens,
progress_callback=_progress,
reference_timeout=reference_timeout,
agent=self._agent,
late_accounting_sink=self._record_late_reference_accounting,
)
interrupted_any = any(
text == _INTERRUPTED_REFERENCE_NOTE
for _lbl, text, _acct in reference_outputs
)
if interrupted_any:
# An interrupted fan-out is a partial snapshot: never cache it (a HIT
# would replay placeholder notes every iteration).
self._ref_cache_key = None
self._ref_cache_outputs = []
else:
self._ref_cache_key = _cache_key
self._ref_cache_outputs = list(reference_outputs)
# Fold advisor spend into accounting exactly once per turn.
self._fold_pending_accounting(*_sum_reference_accounting(reference_outputs))
# Stash the fan-out for trace persistence (aggregator input/label filled in
# later; output stitched in by consume_and_save_trace). Traces are persisted,
# so ANY active privacy mode redacts advisor text and per-advisor input/output.
if privacy_mode:
_trace_refs = [
(label, _redact_reference_text(text), _redact_trace_accounting(acct))
for label, text, acct in reference_outputs
]
else:
_trace_refs = list(reference_outputs)
self._pending_trace = {
"preset": self.preset_name,
"reference_outputs": _trace_refs,
"aggregator_slot": aggregator,
"aggregator_temperature": aggregator_temperature,
}
# Derived from the privacy-redacted _trace_refs.
try:
from agent.moa_trace import slot_metrics
self._last_reference_metrics = [
slot_metrics(acct, label, output=text)
for label, text, acct in _trace_refs
]
except Exception as exc: # pragma: no cover - never break a turn
logger.debug("MoA reference metrics render failed: %s", exc)
self._last_reference_metrics = None
# Surface each reference's answer BEFORE the aggregator acts (once per turn).
# The cache keeps RAW text; redaction happens at each consuming surface.
_ref_count = len(reference_outputs)
for _idx, (_label, _text, _accounting) in enumerate(reference_outputs, start=1):
self._emit(
"moa.reference",
index=_idx,
count=_ref_count,
label=_label,
text=_redact_reference_text(_text) if privacy_mode else _text,
)
if _ref_count:
# Phase transition: fan-out complete, aggregator about to act.
self._emit(
"moa.phase",
phase="aggregator",
refs_done=_ref_count,
refs_total=_ref_count,
aggregator=_slot_label(aggregator),
)
self._emit(
"moa.aggregating",
aggregator=_slot_label(aggregator),
ref_count=_ref_count,
)
agg_messages = [dict(m) for m in messages]
successful_outputs = _successful_references(reference_outputs)
failed_labels = _failed_reference_labels(reference_outputs)
# 'full' privacy mode redacts advisor text reaching the AGGREGATOR too (#59959);
# 'display' leaves it raw. Applied to a per-call copy — the cache holds raw text.
_agg_refs = (
_redact_reference_outputs(successful_outputs)
if privacy_mode == "full"
else successful_outputs
)
degraded = _degraded_notice(failed_labels, degraded_reference_policy)
header = (
"[Mixture of Agents reference context]\n"
f"Preset: {self.preset_name}\n"
f"Aggregator/acting model: {_slot_label(aggregator)}\n"
)
guidance: str | None = None
if reference_outputs and not successful_outputs:
# Every reference failed: the aggregator acts alone. Under the loud policy it
# still gets the sanitized unavailability notice; under silent, nothing.
logger.warning(
"MoA: all %d reference(s) failed — acting aggregator-alone "
"without reference guidance",
len(reference_outputs),
)
if degraded:
guidance = (
f"{header}\n"
"All reference models failed this turn — no advisory "
"guidance is available. Act on your own judgment.\n\n"
f"{degraded}"
)
elif _agg_refs or degraded:
guidance = (
f"{header}"
f"References: {', '.join(label for label, _, _ in _agg_refs)}\n\n"
"Use the reference responses below as private context. You are the aggregator and acting model: "
"answer the user directly or call tools as needed.\n\n"
f"{_join_reference_outputs(_agg_refs, degraded)}"
)
if guidance:
_attach_reference_guidance(agg_messages, guidance)
prepared_request = {
"messages": agg_messages,
"guidance": guidance,
"aggregator": aggregator,
"aggregator_temperature": aggregator_temperature,
}
if api_kwargs.pop("_moa_prepare_only", False):
return prepared_request
return self._call_prepared_aggregator(prepared_request, api_kwargs)
class MoAClient:
def __init__(self, preset_name: str, reference_callback: Any = None, agent: Any = None):
self.chat = type("_MoAChat", (), {})()
self.chat.completions = MoAChatCompletions(
preset_name, reference_callback=reference_callback, agent=agent,
)
def consume_reference_usage(self) -> Any:
"""Pop pending reference-fan-out usage + cost from the completions facade."""
return self.chat.completions.consume_reference_usage()
@property
def last_aggregator_slot(self) -> Any:
"""Resolved aggregator slot from the most recent create(), or None."""
return getattr(self.chat.completions, "last_aggregator_slot", None)
def consume_and_save_trace(
self, session_id: Any = None, aggregator_output_fallback: Any = None
) -> None:
"""Flush the pending full-turn MoA trace via the completions facade."""
return self.chat.completions.consume_and_save_trace(
session_id, aggregator_output_fallback=aggregator_output_fallback
)
def last_reference_metrics(self) -> Any:
"""Per-advisor metrics from the most recent fan-out, or None (read-only)."""
return self.chat.completions.last_reference_metrics()
# Display-event relay table for build_moa_facade: event -> (primary kwarg, secondary
# kwarg or None, {tool_progress_callback kwarg: emit kwarg}). The callback signature is
# ``cb(event, label, text, None, **moa_*)``; "moa.progress" is rendered by frontends as
# a status-bar ``MOA: N/M refs done``.
_RELAY_EVENTS: dict[str, tuple[str, str | None, dict[str, str]]] = {
"moa.reference": ("label", "text", {"moa_index": "index", "moa_count": "count"}),
"moa.progress": ("label", None, {"moa_refs_done": "refs_done", "moa_refs_total": "refs_total"}),
"moa.phase": (
"aggregator",
None,
{"moa_phase": "phase", "moa_refs_done": "refs_done", "moa_refs_total": "refs_total"},
),
"moa.aggregating": ("aggregator", None, {"moa_ref_count": "ref_count"}),
}
def build_moa_facade(agent, preset_name: Any = None) -> MoAClient:
"""Build the MoA facade client for ``agent``, wiring the reference relay.
Single construction point for ``MoAClient`` (agent_init, fallback restore,
transport recovery, switch_model): a bare ``MoAClient(preset)`` would drop the
``reference_callback`` relay and silence display events for the session
(#53802). The relay reads ``agent.tool_progress_callback`` at emit time.
"""
def _moa_reference_relay(event: str, **kwargs: Any) -> None:
cb = getattr(agent, "tool_progress_callback", None)
spec = _RELAY_EVENTS.get(event)
if cb is None or spec is None:
return
primary, secondary, extra_map = spec
try:
cb(
event,
str(kwargs.get(primary) or ""),
str(kwargs.get(secondary) or "") if secondary else None,
None,
**{out: kwargs.get(src) for out, src in extra_map.items()},
)
except Exception:
pass
resolved_preset = preset_name
if resolved_preset is None and getattr(agent, "provider", None) == "moa":
resolved_preset = getattr(agent, "model", None)
resolved_preset = str(resolved_preset or "default")
try:
from hermes_cli.config import load_config
from hermes_cli.moa_config import normalize_moa_config
moa_cfg = normalize_moa_config(load_config().get("moa") or {})
presets = moa_cfg.get("presets") or {}
if resolved_preset not in presets:
resolved_preset = moa_cfg.get("default_preset") or "default"
except Exception:
resolved_preset = "default"
return MoAClient(
resolved_preset,
reference_callback=_moa_reference_relay,
# Lets the fan-out wait be aborted on a user interrupt.
agent=agent,
)