from __future__ import annotations import hashlib from collections.abc import Sequence from typing import Any class FakeLLM: chat_model = "fake-chat-model" embedding_model = "fake-embedding-model" def complete_json(self, system: str, user: str) -> dict[str, Any]: if "community" in system.casefold(): return { "title": "Atlas product community", "summary": "Acme created Atlas, a graph retrieval product.", } return { "entities": [ { "canonical_name": "Acme", "entity_type": "Organization", "description": "Acme is a technology organization.", "aliases": ["Acme Corp"], }, { "canonical_name": "Atlas", "entity_type": "Product", "description": "Atlas is a graph retrieval product.", "aliases": [], }, ], "relationships": [ { "source": "Acme", "target": "Atlas", "relationship_type": "CREATED", "description": "Acme created Atlas.", } ], } def complete_text(self, system: str, user: str) -> str: return "Acme created Atlas. [S1]" def embed(self, texts: Sequence[str]) -> list[list[float]]: vectors: list[list[float]] = [] for text in texts: normalized = text.casefold() digest = hashlib.sha256(normalized.encode("utf-8")).digest() vectors.append( [ float("atlas" in normalized), float("acme" in normalized), digest[0] / 255, ] ) return vectors