File size: 1,856 Bytes
448d6a5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 | """Execute a thinking flow (Def 4.2): a DAG over thinking operators run in
topological order. Generate seeds the pool; Divergent branches into a candidate
pool; Convergent collapses the pool by selecting the best; the other operators
transform the current best idea. e(y | G; x_t, x_r) returns the final idea.
"""
from __future__ import annotations
from . import operators as ops
def execute_flow(seq, topic, related, model=None, seed=None, divergent_n=3):
"""seq: list of refinement operator names (Generate is always prepended).
Returns (final_idea, trace) where trace is the list of (op, idea) steps."""
trace = []
s = (seed or 0)
current = ops.op_generate(topic, related, model=model, seed=s)
pool = [current]
trace.append(("Generate", current))
for i, name in enumerate(seq):
s += 1
if name == "Divergent":
alts = ops.op_divergent_expand(topic, related, current, n=divergent_n, model=model, seed=s)
pool = pool + alts
current = alts[0]
elif name == "Convergent":
current = ops.op_convergent_select(topic, related, pool, model=model, seed=s)
pool = [current]
else:
fn = ops.OPERATORS.get(name)
if fn is None:
continue
current = fn(topic, related, current, model=model, seed=s)
pool[-1] = current
trace.append((name, current))
return current, trace
def idea_to_text(idea):
if not idea:
return ""
return (f"Title: {idea.get('title','')}\n"
f"Motivation/Problem: {idea.get('problem','')}\n"
f"Method: {idea.get('method','')}\n"
f"Evaluation: {idea.get('evaluation','')}")
def motivation_text(idea):
return f"Title: {idea.get('title','')}\nMotivation: {idea.get('problem','')}" if idea else ""
|