feat: 新增岗位管理、提示词模板与资源标签,完善用户提示词与推荐问题接口
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
1 parent
2f414c6e08
commit
3855c30edd
52 files changed
+3143
-57
No files matched your search
+79
-12
@@ -18,6 +18,7 @@ post-compaction behavior).
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import threading
|
||||
from collections import OrderedDict
|
||||
from collections.abc import Awaitable, Callable, Collection
|
||||
from dataclasses import dataclass
|
||||
@@ -131,6 +132,27 @@ class _SummaryCacheEntry:
|
||||
reusable: bool
|
||||
|
||||
|
||||
# Process-wide summary cache shared across every middleware instance.
|
||||
#
|
||||
# ``make_lead_agent`` rebuilds the agent — and therefore a *fresh*
|
||||
# ``DeerFlowSummarizationMiddleware`` — on every run. If the cache lived on the
|
||||
# instance (``self._summary_cache``) it would be empty at the start of each
|
||||
# user turn, forcing a brand-new (blocking) summarization LLM call on *every*
|
||||
# turn of a long conversation. Hoisting it to module scope, keyed by
|
||||
# ``thread_id``, lets a summary computed on one turn be reused on the next,
|
||||
# so only the first threshold crossing (or a genuine re-compaction) pays for an
|
||||
# LLM call. A lock guards the non-atomic LRU eviction against concurrent
|
||||
# threads (subagents / parallel conversations share this process).
|
||||
_GLOBAL_SUMMARY_CACHE: "OrderedDict[str, _SummaryCacheEntry]" = OrderedDict()
|
||||
_GLOBAL_SUMMARY_CACHE_LOCK = threading.Lock()
|
||||
|
||||
|
||||
def _clear_global_summary_cache() -> None:
|
||||
"""Drop every cached summary. Intended for tests/diagnostics only."""
|
||||
with _GLOBAL_SUMMARY_CACHE_LOCK:
|
||||
_GLOBAL_SUMMARY_CACHE.clear()
|
||||
|
||||
|
||||
class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
|
||||
"""Summarization middleware with pre-compression hook dispatch and skill rescue."""
|
||||
|
||||
@@ -153,8 +175,9 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
|
||||
self._preserve_recent_skill_tokens = max(0, preserve_recent_skill_tokens)
|
||||
self._preserve_recent_skill_tokens_per_skill = max(0, preserve_recent_skill_tokens_per_skill)
|
||||
# Per-thread cache of the most recent summary: thread_id -> (summarized message ids, summary text).
|
||||
# Lets us reuse a summary across model calls instead of re-running the LLM every invocation.
|
||||
self._summary_cache: OrderedDict[str, _SummaryCacheEntry] = OrderedDict()
|
||||
# Points at the *process-wide* cache (not a fresh per-instance dict) so a
|
||||
# summary survives across runs/turns — see ``_GLOBAL_SUMMARY_CACHE``.
|
||||
self._summary_cache: OrderedDict[str, _SummaryCacheEntry] = _GLOBAL_SUMMARY_CACHE
|
||||
|
||||
@override
|
||||
def before_model(self, state: AgentState, runtime: Runtime) -> dict | None:
|
||||
@@ -208,6 +231,7 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
|
||||
return payload # type: ignore[return-value]
|
||||
# action == "summarize"
|
||||
messages_to_summarize, preserved, thread_id = payload # type: ignore[misc]
|
||||
self._emit_compacting_notice()
|
||||
self._fire_hooks(messages_to_summarize, preserved, getattr(request, "runtime", None))
|
||||
summary = self._create_summary(messages_to_summarize)
|
||||
return self._finish_summarize(messages_to_summarize, preserved, summary, thread_id)
|
||||
@@ -221,10 +245,34 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
|
||||
if action == "reuse":
|
||||
return payload # type: ignore[return-value]
|
||||
messages_to_summarize, preserved, thread_id = payload # type: ignore[misc]
|
||||
self._emit_compacting_notice()
|
||||
self._fire_hooks(messages_to_summarize, preserved, getattr(request, "runtime", None))
|
||||
summary = await self._acreate_summary(messages_to_summarize)
|
||||
return self._finish_summarize(messages_to_summarize, preserved, summary, thread_id)
|
||||
|
||||
@staticmethod
|
||||
def _emit_compacting_notice() -> None:
|
||||
"""Best-effort UI hint that a (blocking) summarization LLM call is starting.
|
||||
|
||||
Only the genuine summarize path calls this — the cheap cache-reuse path
|
||||
does no LLM work and stays silent. Emitted on LangGraph's ``custom``
|
||||
stream channel; the chat frontend renders it as a transient toast. A
|
||||
no-op when there is no active stream writer (e.g. unit tests, non-stream
|
||||
runs), so callers never branch.
|
||||
"""
|
||||
try:
|
||||
from langgraph.config import get_stream_writer
|
||||
|
||||
writer = get_stream_writer()
|
||||
except Exception:
|
||||
return
|
||||
if writer is None:
|
||||
return
|
||||
try:
|
||||
writer({"type": "context_compacting", "message": "正在压缩上下文以容纳更长对话…"})
|
||||
except Exception:
|
||||
logger.debug("summarization: failed to emit context_compacting notice", exc_info=True)
|
||||
|
||||
def _plan_compression(
|
||||
self,
|
||||
request: ModelRequest,
|
||||
@@ -243,23 +291,30 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
|
||||
if not messages:
|
||||
return None
|
||||
self._ensure_message_ids(messages)
|
||||
thread_id = self._resolve_request_thread_id(request)
|
||||
|
||||
cached = self._summary_cache.get(thread_id) if thread_id else None
|
||||
# Threshold is the first gate: while the *full* current conversation is
|
||||
# still under the configured trigger, there is nothing to compress —
|
||||
# pass it through untouched. No cache lookup, no summary substitution.
|
||||
# Compression only ever engages once the real context crosses the
|
||||
# trigger, which is exactly when it is needed.
|
||||
if not self._should_summarize(messages, self.token_counter(messages)):
|
||||
return None
|
||||
|
||||
thread_id = self._resolve_request_thread_id(request)
|
||||
if thread_id is None:
|
||||
logger.debug("summarization: thread_id unresolved; cross-turn summary cache disabled for this call")
|
||||
cached = self._get_cache_entry(thread_id) if thread_id else None
|
||||
present_ids = {m.id for m in messages}
|
||||
if cached is not None and cached.reusable and cached.summarized_ids <= present_ids:
|
||||
preserved = [m for m in messages if m.id not in cached.summarized_ids]
|
||||
effective = [self._summary_message(cached.summary_text), *preserved]
|
||||
if not self._should_summarize(effective, self.token_counter(effective)):
|
||||
# Refresh LRU recency and reuse without an LLM call.
|
||||
self._summary_cache.move_to_end(thread_id)
|
||||
self._touch_cache_entry(thread_id)
|
||||
return ("reuse", effective)
|
||||
# New messages have pushed the compressed context back over the
|
||||
# threshold — fall through and re-summarize a larger prefix.
|
||||
|
||||
if not self._should_summarize(messages, self.token_counter(messages)):
|
||||
return None
|
||||
|
||||
cutoff_index = self._determine_cutoff_index(messages)
|
||||
if cutoff_index <= 0:
|
||||
return None
|
||||
@@ -294,12 +349,24 @@ class DeerFlowSummarizationMiddleware(SummarizationMiddleware):
|
||||
|
||||
return compressed
|
||||
|
||||
def _get_cache_entry(self, thread_id: str) -> _SummaryCacheEntry | None:
|
||||
"""Return the cached entry for ``thread_id`` (thread-safe read)."""
|
||||
with _GLOBAL_SUMMARY_CACHE_LOCK:
|
||||
return self._summary_cache.get(thread_id)
|
||||
|
||||
def _touch_cache_entry(self, thread_id: str) -> None:
|
||||
"""Refresh LRU recency for ``thread_id`` if still present (thread-safe)."""
|
||||
with _GLOBAL_SUMMARY_CACHE_LOCK:
|
||||
if thread_id in self._summary_cache:
|
||||
self._summary_cache.move_to_end(thread_id)
|
||||
|
||||
def _store_cache_entry(self, thread_id: str, entry: _SummaryCacheEntry) -> None:
|
||||
"""Insert/refresh a cache entry, evicting the oldest thread if over capacity."""
|
||||
self._summary_cache[thread_id] = entry
|
||||
self._summary_cache.move_to_end(thread_id)
|
||||
while len(self._summary_cache) > _SUMMARY_CACHE_MAX_THREADS:
|
||||
self._summary_cache.popitem(last=False)
|
||||
with _GLOBAL_SUMMARY_CACHE_LOCK:
|
||||
self._summary_cache[thread_id] = entry
|
||||
self._summary_cache.move_to_end(thread_id)
|
||||
while len(self._summary_cache) > _SUMMARY_CACHE_MAX_THREADS:
|
||||
self._summary_cache.popitem(last=False)
|
||||
|
||||
@staticmethod
|
||||
def _resolve_request_thread_id(request: ModelRequest) -> str | None:
|
||||
|
||||
Reference in new issue
Block a user