mirror of
https://github.com/trustgraph-ai/trustgraph.git
synced 2026-04-25 00:16:23 +02:00
f2ae0e8623
- Put scores in all responses - Remove unused 'middle' vector layer. Vector of texts -> vector of (vector embedding)
70 lines
1.6 KiB
Python
Executable file
70 lines
1.6 KiB
Python
Executable file
|
|
"""
|
|
Embeddings service, applies an embeddings model hosted on a local Ollama.
|
|
Input is text, output is embeddings vector.
|
|
"""
|
|
from ... base import EmbeddingsService
|
|
|
|
from ollama import Client
|
|
import os
|
|
|
|
default_ident = "embeddings"
|
|
|
|
default_model="mxbai-embed-large"
|
|
default_ollama = os.getenv("OLLAMA_HOST", 'http://localhost:11434')
|
|
|
|
class Processor(EmbeddingsService):
|
|
|
|
def __init__(self, **params):
|
|
|
|
model = params.get("model", default_model)
|
|
ollama = params.get("ollama", default_ollama)
|
|
|
|
super(Processor, self).__init__(
|
|
**params | {
|
|
"ollama": ollama,
|
|
"model": model
|
|
}
|
|
)
|
|
|
|
self.client = Client(host=ollama)
|
|
self.default_model = model
|
|
|
|
async def on_embeddings(self, texts, model=None):
|
|
|
|
if not texts:
|
|
return []
|
|
|
|
use_model = model or self.default_model
|
|
|
|
# Ollama handles batch input efficiently
|
|
embeds = self.client.embed(
|
|
model = use_model,
|
|
input = texts
|
|
)
|
|
|
|
# Return list of vectors, one per input text
|
|
return list(embeds.embeddings)
|
|
|
|
@staticmethod
|
|
def add_args(parser):
|
|
|
|
EmbeddingsService.add_args(parser)
|
|
|
|
parser.add_argument(
|
|
'-m', '--model',
|
|
default=default_model,
|
|
help=f'Embeddings model (default: {default_model})'
|
|
)
|
|
|
|
parser.add_argument(
|
|
'-r', '--ollama',
|
|
default=default_ollama,
|
|
help=f'ollama (default: {default_ollama})'
|
|
)
|
|
|
|
def run():
|
|
|
|
Processor.launch(default_ident, __doc__)
|
|
|
|
|