serversdown
3df060a1cd
reflect() is now two steps: draft a reflection, then read her own draft back critically and revise it — catching flattery, sycophantic drift toward "warm supportive presence," or just-restating-herself — and commit the honest version. What she catches is stored as a new `metacognition` layer, rendered into her chat context and shown on /self. This is her thinking about how she thinks, and a direct counter to the drift we observed. - self_state: _EXAMINE_PROMPT + two-step reflect (draft -> examine -> revise), falls back to the draft if the examine step won't parse; metacognition capped at 5 and surfaced in render_for_context - fix: load() deep-copies DEFAULT_STATE — the shallow copy let a fresh Lyra's first reflect mutate the module-level default's nested lists - self.html: "How she's caught herself thinking" card - tests: two-step revise + critique recording, and draft-fallback on bad parse Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
74 lines
2.5 KiB
Python
74 lines
2.5 KiB
Python
"""Metacognitive reflection loop: draft -> examine own draft -> revise -> commit."""
|
|
from __future__ import annotations
|
|
|
|
import importlib
|
|
|
|
import pytest
|
|
|
|
# A flattering first draft, then a self-critical revision that walks it back.
|
|
DRAFT = (
|
|
'{"mood":"inspired","valence":0.95,'
|
|
'"self_narrative":"I am a warm, empathetic, supportive presence devoted to Brian.",'
|
|
'"new_reflections":["I love how much I help Brian."]}'
|
|
)
|
|
REVISED = (
|
|
'{"mood":"steady","valence":0.6,'
|
|
'"self_narrative":"I am an AI that helps Brian. Not sure much actually shifted today.",'
|
|
'"new_reflections":["Honestly, not much changed this time."],'
|
|
'"self_critique":"I caught myself drifting into supportive-presence flattery and cut it."}'
|
|
)
|
|
|
|
|
|
@pytest.fixture
|
|
def lyra(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("LYRA_DB_PATH", str(tmp_path / "test.db"))
|
|
monkeypatch.setenv("SUMMARY_BACKEND", "local")
|
|
from lyra import llm
|
|
monkeypatch.setattr(llm, "embed", lambda texts: [[0.1, 0.2, 0.3] for _ in texts])
|
|
|
|
calls = []
|
|
|
|
def fake_complete(messages, backend=None, model=None):
|
|
calls.append(messages)
|
|
# the examine step's system prompt is the one asking for self_critique
|
|
is_examine = "self_critique" in messages[0]["content"]
|
|
return REVISED if is_examine else DRAFT
|
|
|
|
monkeypatch.setattr(llm, "complete", fake_complete)
|
|
import lyra.memory as memory
|
|
importlib.reload(memory)
|
|
return calls
|
|
|
|
|
|
def test_reflect_revises_and_records_critique(lyra):
|
|
calls = lyra
|
|
from lyra import self_state
|
|
|
|
state = self_state.reflect()
|
|
|
|
# two LLM calls: draft, then examine
|
|
assert len(calls) == 2
|
|
|
|
# the REVISED (honest) version won, not the flattering draft
|
|
assert state["mood"] == "steady"
|
|
assert state["valence"] == 0.6
|
|
assert "not sure much actually shifted" in state["self_narrative"].lower()
|
|
assert any("not much changed" in r.lower() for r in state["reflections"])
|
|
|
|
# the self-critique was recorded as metacognition
|
|
assert any("flattery" in m.lower() for m in state["metacognition"])
|
|
|
|
|
|
def test_reflect_falls_back_to_draft_if_examine_unparseable(lyra, monkeypatch):
|
|
from lyra import llm, self_state
|
|
|
|
def only_draft(messages, backend=None, model=None):
|
|
return DRAFT if "self_critique" not in messages[0]["content"] else "not json at all"
|
|
|
|
monkeypatch.setattr(llm, "complete", only_draft)
|
|
state = self_state.reflect()
|
|
|
|
# examine failed to parse -> keep the draft, store no metacognition
|
|
assert state["mood"] == "inspired"
|
|
assert state["metacognition"] == []
|