eqa-search/eqa_search.py

# Copyright 2021 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from towhee import ops, pipe, AutoPipes, AutoConfig


@AutoConfig.register
class EnhancedQASearchConfig:
    """
    Config of pipeline
    """
    def __init__(self):
        # config for sentence_embedding
        self.model = 'all-MiniLM-L6-v2'
        self.openai_api_key = None
        self.embedding_device = -1
        # config for search_milvus
        self.host = '127.0.0.1'
        self.port = '19530'
        self.collection_name = 'chatbot'
        self.top_k = 5
        self.user = None
        self.password = None
        # config for similarity evaluation
        self.threshold = 0.6
        # config for llm
        self.llm_model = 'openai'


_hf_models = ops.sentence_embedding.transformers().get_op().supported_model_names()
_sbert_models = ops.sentence_embedding.sbert().get_op().supported_model_names()
_openai_models = ['text-embedding-ada-002', 'text-similarity-davinci-001',
                  'text-similarity-curie-001', 'text-similarity-babbage-001',
                  'text-similarity-ada-001']


def _get_embedding_op(config):
    if config.embedding_device == -1:
        device = 'cpu'
    else:
        device = config.embedding_device

    if config.model in _hf_models:
        return True, ops.sentence_embedding.transformers(
            model_name=config.model, device=device
        )

    if config.model in _sbert_models:
        return True, ops.sentence_embedding.sbert(
            model_name=config.model, device=device
        )

    if config.model in _openai_models:
        return False, ops.sentence_embedding.openai(
            model_name=config.model, api_key=config.openai_api_key
        )

    raise RuntimeError('Unknown model: [%s], only support: %s' % (config.model, _hf_models + _openai_models))


def _get_similarity_evaluation_op(config):
    return lambda x: [i for i in x if i[1] >= config.threshold]


def _get_llm_op(config):
    if config.llm_model.lower() == 'openai':
        return ops.LLM.OpenAI(api_key=config.openai_api_key)
    if config.llm_model.lower() == 'dolly':
        return ops.LLM.Dolly()

    raise RuntimeError('Unknown llm model: [%s], only support \'openai\' and \'dolly\'' % (config.model))


@AutoPipes.register
def enhanced_qa_search_pipe(config):
    allow_triton, sentence_embedding_op = _get_embedding_op(config)
    sentence_embedding_config = {}
    if allow_triton:
        if config.embedding_device >= 0:
            sentence_embedding_config = AutoConfig.TritonGPUConfig(device_ids=[config.embedding_device], max_batch_size=128)
        else:
            sentence_embedding_config = AutoConfig.TritonCPUConfig()

    search_milvus_op = ops.ann_search.milvus_client(
        host=config.host,
        port=config.port,
        collection_name=config.collection_name,
        limit=config.top_k,
        output_fields=['text'],
        metric_type='IP',
        user=config.user,
        password=config.password,
    )

    llm_op = _get_llm_op(config)

    p = (
        pipe.input('question', 'history')
            .map('question', 'embedding', sentence_embedding_op, config=sentence_embedding_config)
            .map('embedding', 'embedding', ops.towhee.np_normalize())
            .map('embedding', 'result', search_milvus_op)
    )

    # if config.similarity_evaluation:
    if config.threshold:
        sim_eval_op = _get_similarity_evaluation_op(config)
        p = p.map('result', 'result', sim_eval_op)
    
    p = (
        p.map('result', 'docs', lambda x:[i[2] for i in x])
            .map(('question', 'docs', 'history'), 'prompt', ops.prompt.question_answer(llm_name=config.llm_model))
            .map('prompt', 'answer', llm_op)
    )

    return p.output('answer')