mirror of
https://github.com/cocoindex-io/cocoindex-code.git
synced 2026-09-14 16:39:38 +08:00
Jiangzhou
9700655aa6
- Remove `dimensions` from the litellm whitelist in `_ACCEPTED_KWARGS`. Output dimension must be identical for indexing and query for vectors to be comparable, so it's a model-wide setting, not a per-side knob — exposing it under `indexing_params` / `query_params` invited misconfiguration. Updated comment template, README, design doc, and testing plan accordingly. - Plumb `indexing_params` into `create_embedder` and pass them as constructor kwargs to `PacedLiteLLMEmbedder`. The values land in `self._kwargs` and become defaults forwarded into every `litellm.aembedding` call — including paths that don't go through the `INDEXING_EMBED_PARAMS` context var (e.g. the dim probe in `_get_dim`). Per-call overrides (`query_params` spread at query time) still win because `_embed` overlays kwargs on top of `self._kwargs`. Sentence- transformers ignores `indexing_params` (its constructor doesn't accept arbitrary kwargs; `prompt_name` is per-call only).
97 lines
3.5 KiB
Python
97 lines
3.5 KiB
Python
"""Validation and resolution of embedder ``indexing_params`` / ``query_params``.
|
|
|
|
Runtime entry point is :func:`resolve_embedder_params`. The curated defaults
|
|
table lives in :mod:`embedder_defaults` and is used only by ``ccc init`` —
|
|
this module does not consult it.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import Any, NamedTuple
|
|
|
|
from .embedder_defaults import LEGACY_QUERY_PROMPT_MODELS
|
|
from .settings import EmbeddingSettings
|
|
|
|
__all__ = [
|
|
"EmbedderParams",
|
|
"accepted_kwargs_for",
|
|
"resolve_embedder_params",
|
|
"validate_params",
|
|
]
|
|
|
|
|
|
# Accepted kwargs per provider. Intentionally minimal — we only expose knobs
|
|
# that users have reason to tune AND that make sense per-side (indexing vs
|
|
# query). Excluded keys:
|
|
# - ``normalize_embeddings`` (sentence-transformers): query._l2_to_score
|
|
# assumes unit vectors.
|
|
# - ``encoding_format`` (litellm): litellm_embedder hardcodes "float".
|
|
_ACCEPTED_KWARGS: dict[str, frozenset[str]] = {
|
|
"sentence-transformers": frozenset({"prompt_name"}),
|
|
"litellm": frozenset({"input_type"}),
|
|
}
|
|
|
|
|
|
def accepted_kwargs_for(provider: str) -> frozenset[str]:
|
|
"""Return the set of accepted kwarg names for *provider*.
|
|
|
|
Raises ``ValueError`` on unknown providers.
|
|
"""
|
|
try:
|
|
return _ACCEPTED_KWARGS[provider]
|
|
except KeyError as e:
|
|
raise ValueError(f"Unknown provider: {provider!r}") from e
|
|
|
|
|
|
def validate_params(
|
|
provider: str,
|
|
indexing_params: dict[str, Any] | None,
|
|
query_params: dict[str, Any] | None,
|
|
) -> None:
|
|
"""Raise ``ValueError`` if either dict contains keys not accepted by *provider*."""
|
|
accepted = accepted_kwargs_for(provider)
|
|
for side, params in (("indexing_params", indexing_params), ("query_params", query_params)):
|
|
if not params:
|
|
continue
|
|
unknown = sorted(set(params) - accepted)
|
|
if unknown:
|
|
raise ValueError(
|
|
f"{side}: unknown key(s) {unknown!r} for provider {provider!r}. "
|
|
f"Accepted keys: {sorted(accepted)!r}."
|
|
)
|
|
|
|
|
|
class EmbedderParams(NamedTuple):
|
|
"""Params that will be spread into ``embedder.embed()`` calls at runtime."""
|
|
|
|
indexing: dict[str, Any] # never None; possibly empty
|
|
query: dict[str, Any] # never None; possibly empty
|
|
used_backward_compat: bool # True iff the legacy bridge fired
|
|
|
|
|
|
def resolve_embedder_params(settings: EmbeddingSettings) -> EmbedderParams:
|
|
"""Resolve the effective embedder params from user settings.
|
|
|
|
Whatever the user put in the file, verbatim, with one exception for
|
|
backward compatibility: if neither ``indexing_params`` nor ``query_params``
|
|
is set and the model was previously handled by the hardcoded
|
|
``_QUERY_PROMPT_MODELS`` path, fill in ``query = {'prompt_name': 'query'}``
|
|
and raise the ``used_backward_compat`` flag so the daemon emits a
|
|
handshake warning.
|
|
"""
|
|
indexing: dict[str, Any] = dict(settings.indexing_params) if settings.indexing_params else {}
|
|
query: dict[str, Any] = dict(settings.query_params) if settings.query_params else {}
|
|
used_backward_compat = False
|
|
|
|
if (
|
|
settings.indexing_params is None
|
|
and settings.query_params is None
|
|
and settings.provider == "sentence-transformers"
|
|
and settings.model in LEGACY_QUERY_PROMPT_MODELS
|
|
):
|
|
query = {"prompt_name": "query"}
|
|
used_backward_compat = True
|
|
|
|
validate_params(settings.provider, indexing, query)
|
|
return EmbedderParams(indexing=indexing, query=query, used_backward_compat=used_backward_compat)
|