File size: 9,369 Bytes
15c3265 9b7e0a7 15c3265 9b7e0a7 15c3265 9b7e0a7 15c3265 | 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 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 | """Run the end-to-end pipeline (document/persona/worldview layers)."""
from __future__ import annotations
import argparse
from pathlib import Path
import re
from typing import Dict, List
from fic_agent.config import RuntimeConfig
from fic_agent.ingest.pipeline import (
build_document_layer,
chunks_to_dicts,
dialogues_to_dicts,
extract_dialogue,
load_text_file,
save_jsonl,
)
from fic_agent.persona.profile import (
build_persona_profile,
render_persona_prompt,
save_persona_profile,
save_persona_prompt,
)
from fic_agent.worldview.worldview import (
build_worldview_notes,
build_worldview_notes_llm,
save_worldview_notes,
)
from fic_agent.retrieval.retriever import build_index_for_texts
def _norm_text(value: str) -> str:
return re.sub(r"\s+", " ", (value or "").strip()).lower()
def _select_worldview_for_character(
character: str,
worldview_notes: List[Dict],
character_chunk_ids: set[str],
top_n: int = 6,
) -> List[str]:
char_tokens = set(_norm_text(character).split())
ranked: List[tuple[int, str]] = []
for note in worldview_notes:
text = (note.get("text") or "").strip()
if not text:
continue
entity = _norm_text(str(note.get("entity") or ""))
text_l = _norm_text(text)
source_chunk = note.get("source_chunk")
score = 0
if source_chunk in character_chunk_ids:
score += 2
if entity and char_tokens and (set(entity.split()) & char_tokens):
score += 4
if char_tokens and any(re.search(rf"\b{re.escape(tok)}\b", text_l) for tok in char_tokens):
score += 3
if note.get("type") in {"rule", "event"} and source_chunk in character_chunk_ids:
score += 1
if score > 0:
ranked.append((score, text))
ranked.sort(key=lambda x: (x[0], len(x[1])), reverse=True)
selected: List[str] = []
seen = set()
for _, text in ranked:
k = _norm_text(text)
if k in seen:
continue
seen.add(k)
selected.append(text)
if len(selected) >= top_n:
break
if selected:
return selected
# Fallback: provide a small set of global rule/setting notes.
for note in worldview_notes:
t = (note.get("type") or "").lower()
if t not in {"rule", "setting", "event", "location", "organization"}:
continue
text = (note.get("text") or "").strip()
if not text:
continue
k = _norm_text(text)
if k in seen:
continue
seen.add(k)
selected.append(text)
if len(selected) >= min(3, top_n):
break
return selected
def main() -> None:
parser = argparse.ArgumentParser(description="fic-agent pipeline")
parser.add_argument("--input", required=True, help="Path to raw novel text")
parser.add_argument("--book-id", default="book", help="Book id")
parser.add_argument("--character", default=None, help="Character name for persona")
parser.add_argument("--characters", default=None, help="Comma-separated character list for speaker detection")
parser.add_argument(
"--all-characters",
action="store_true",
help="Generate persona/profile prompt for all characters in --characters (or all detected speakers if --characters is absent).",
)
parser.add_argument("--max-chars", type=int, default=2000)
parser.add_argument("--overlap", type=int, default=200)
parser.add_argument("--build-index", action="store_true", help="Build vector indexes")
parser.add_argument("--worldview-llm", action="store_true", help="Use LLM to extract worldview notes")
args = parser.parse_args()
cfg = RuntimeConfig()
raw_text = load_text_file(args.input)
chunks = build_document_layer(raw_text, book_id=args.book_id, max_chars=args.max_chars, overlap=args.overlap)
chunk_dicts = chunks_to_dicts(chunks)
processed_dir = Path(cfg.data_processed_dir)
processed_dir.mkdir(parents=True, exist_ok=True)
chunks_path = processed_dir / "chunks.jsonl"
save_jsonl(chunk_dicts, str(chunks_path))
character_candidates = None
if args.characters:
character_candidates = [c.strip() for c in args.characters.split(",") if c.strip()]
dialogues = extract_dialogue(chunks, cfg=cfg, character_candidates=character_candidates)
dialogues_path = processed_dir / "dialogues.jsonl"
save_jsonl(dialogues_to_dicts(dialogues), str(dialogues_path))
# Worldview notes
if args.worldview_llm:
worldview_notes = build_worldview_notes_llm(chunk_dicts, cfg)
else:
worldview_notes = build_worldview_notes(chunk_dicts)
worldview_path = processed_dir / "worldview_notes.jsonl"
save_worldview_notes(worldview_notes, str(worldview_path))
# Persona profile(s)
persona_outputs: List[tuple[str, Path, Path]] = []
persona_targets: List[str] = []
if args.all_characters:
if character_candidates:
persona_targets = character_candidates
else:
# Derive from extracted speakers when candidates are not provided.
seen = set()
for d in dialogues:
if d.speaker and d.speaker not in seen:
seen.add(d.speaker)
persona_targets.append(d.speaker)
elif args.character:
persona_targets = [args.character]
for character in persona_targets:
utterances: List[str] = [d.utterance for d in dialogues if d.speaker == character]
background_utterances: List[str] = [d.utterance for d in dialogues if d.speaker != character]
character_chunk_ids = {d.chunk_id for d in dialogues if d.speaker == character}
all_speakers = sorted({d.speaker for d in dialogues if d.speaker})
excluded_terms = sorted(set((character_candidates or []) + all_speakers))
character_worldview_notes = _select_worldview_for_character(
character=character,
worldview_notes=worldview_notes,
character_chunk_ids=character_chunk_ids,
top_n=6,
)
profile = build_persona_profile(
character,
utterances,
background_utterances=background_utterances,
excluded_terms=excluded_terms,
worldview_notes=character_worldview_notes,
)
safe_character = character.replace("/", "_")
persona_path = processed_dir / f"persona_{safe_character}.json"
save_persona_profile(profile, str(persona_path))
persona_prompt_path = processed_dir / f"persona_{safe_character}_prompt.txt"
save_persona_prompt(render_persona_prompt(profile), str(persona_prompt_path))
persona_outputs.append((character, persona_path, persona_prompt_path))
print(f"Saved chunks to {chunks_path}")
print(f"Saved dialogues to {dialogues_path}")
for character, persona_path, persona_prompt_path in persona_outputs:
print(f"Saved persona for {character} to {persona_path}")
print(f"Saved persona prompt for {character} to {persona_prompt_path}")
print(f"Saved worldview notes to {worldview_path}")
if args.build_index:
# Facts
build_index_for_texts(
[c["text"] for c in chunk_dicts],
[
{"id": c["chunk_id"], "text": c["text"], "chapter_id": c["chapter_id"]}
for c in chunk_dicts
],
cfg,
"facts",
)
# Persona
if dialogues:
build_index_for_texts(
[d.utterance for d in dialogues],
[
{
"id": f"dlg-{i}",
"text": d.utterance,
"speaker": d.speaker,
"chunk_id": d.chunk_id,
}
for i, d in enumerate(dialogues)
],
cfg,
"persona",
)
else:
print("No dialogue extracted; skipping persona index.")
# Worldview
if worldview_notes:
worldview_texts = []
worldview_meta = []
seen_worldview_text = set()
for i, w in enumerate(worldview_notes):
text = str(w.get("text", "")).strip()
if not text:
continue
norm = " ".join(text.split()).lower()
if norm in seen_worldview_text:
continue
seen_worldview_text.add(norm)
worldview_texts.append(text)
worldview_meta.append(
{
"id": f"wv-{i}",
"text": text,
"type": w.get("type"),
"entity": w.get("entity"),
"source_chunk": w.get("source_chunk"),
}
)
build_index_for_texts(
worldview_texts,
worldview_meta,
cfg,
"worldview",
)
else:
print("No worldview notes extracted; skipping worldview index.")
print("Built faiss indexes in", cfg.data_index_dir)
if __name__ == "__main__":
main()
|