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

class API(Application):
    “””
    Base API template. The API is an extended txtai application, adding the ability to cluster API instances together.
    Downstream applications can extend this base template to add/modify functionality.
    “””
    def init(self, config, loaddata=True):
        super().init(config, loaddata)
        # Embeddings cluster
        self.cluster = None
        if self.config.get(“cluster”):
            self.cluster = Cluster(self.config[“cluster”])
    # pylint: disable=W0221
    def search(self, query, limit=None, weights=None, index=None, parameters=None, graph=False, request=None):
        # When search is invoked via the API, limit is set from the request
        # When search is invoked directly, limit is set using the method parameter
        limit = self.limit(request.query_params.get(“limit”) if request and hasattr(request, “query_params”) else limit)
        weights = self.weights(request.query_params.get(“weights”) if request and hasattr(request, “query_params”) else weights)
        index = request.query_params.get(“index”) if request and hasattr(request, “query_params”) else index
        parameters = request.query_params.get(“parameters”) if request and hasattr(request, “query_params”) else parameters
        graph = request.query_params.get(“graph”) if request and hasattr(request, “query_params”) else graph
        # Decode parameters
        parameters = json.loads(parameters) if parameters and isinstance(parameters, str) else parameters
        if self.cluster:
            return self.cluster.search(query, limit, weights, index, parameters, graph)
        return super().search(query, limit, weights, index, parameters, graph)
    def batchsearch(self, queries, limit=None, weights=None, index=None, parameters=None, graph=False):
        if self.cluster:
            return self.cluster.batchsearch(queries, self.limit(limit), weights, index, parameters, graph)
        return super().batchsearch(queries, limit, weights, index, parameters, graph)
    def add(self, documents):
        “””
        Adds a batch of documents for indexing.
        Downstream applications can override this method to also store full documents in an external system.
        Args:
            documents: list of {id: value, text: value}
        Returns:
            unmodified input documents
        “””
        if self.cluster:
            self.cluster.add(documents)
        else:
            super().add(documents)
        return documents
    def index(self):
        “””
        Builds an embeddings index for previously batched documents.
        “””
        if self.cluster:
            self.cluster.index()
        else:
            super().index()
    def upsert(self):
        “””
        Runs an embeddings upsert operation for previously batched documents.
        “””
        if self.cluster:
            self.cluster.upsert()
        else:
            super().upsert()
    def delete(self, ids):
        “””
        Deletes from an embeddings index. Returns list of ids deleted.
        Args:
            ids: list of ids to delete
        Returns:
            ids deleted
        “””
        if self.cluster:
            return self.cluster.delete(ids)
        return super().delete(ids)
    def reindex(self, config, function=None):
        “””
        Recreates this embeddings index using config. This method only works if document content storage is enabled.
        Args:
            config: new config
            function: optional function to prepare content for indexing
        “””
        if self.cluster:
            self.cluster.reindex(config, function)
        else:
            super().reindex(config, function)
    def count(self):
        “””
        Total number of elements in this embeddings index.
        Returns:
            number of elements in embeddings index
        “””
        if self.cluster:
            return self.cluster.count()
        return super().count()
    def limit(self, limit):
        “””
        Parses the number of results to return from the request. Allows range of 1-250, with a default of 10.
        Args:
            limit: limit parameter
        Returns:
            bounded limit
        “””
        # Return between 1 and 250 results, defaults to 10
        return max(1, min(250, int(limit) if limit else 10))
    def weights(self, weights):
        “””
        Parses the weights parameter from the request.
        Args:
            weights: weights parameter
        Returns:
            weights
        “””
        return float(weights) if weights else weights

list of {id: value, text: value}

unmodified input documents

51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

def add(self, documents):
    “””
    Adds a batch of documents for indexing.
    Downstream applications can override this method to also store full documents in an external system.
    Args:
        documents: list of {id: value, text: value}
    Returns:
        unmodified input documents
    “””
    if self.cluster:
        self.cluster.add(documents)
    else:
        super().add(documents)
    return documents

object content

optional list of corresponding uids

optional field to set

documents

402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430

def addobject(self, data, uid, field):
    “””
    Helper method that builds a batch of object documents.
    Args:
        data: object content
        uid: optional list of corresponding uids
        field: optional field to set
    Returns:
        documents
    “””
    # Raise error if index is not writable
    if not self.config.get(“writable”):
        raise ReadOnlyError(“Attempting to add documents to a read-only index (writable != True)”)
    documents = []
    for x, content in enumerate(data):
        if field:
            row = {“id”: uid[x], field: content} if uid else {field: content}
        elif uid:
            row = (uid[x], content)
        else:
            row = content
        documents.append(row)
    return self.add(documents)

queries text

optional list of text, otherwise runs search queries

optional limit if texts is None

list of dict per input text per query where a higher token scores represents higher importance relative to the query

613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630

def batchexplain(self, queries, texts=None, limit=10):
    “””
    Explains the importance of each input token in text for a list of queries.
    Args:
        query: queries text
        texts: optional list of text, otherwise runs search queries
        limit: optional limit if texts is None
    Returns:
        list of dict per input text per query where a higher token scores represents higher importance relative to the query
    “””
    if self.embeddings:
        with self.lock:
            return self.embeddings.batchexplain(queries, texts, limit)
    return None

queries text

list of text

list of {id: value, score: value} per query

572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592

def batchsimilarity(self, queries, texts):
    “””
    Computes the similarity between list of queries and list of text. Returns a list
    of {id: value, score: value} sorted by highest score per query, where id is the
    index in texts.
    Args:
        queries: queries text
        texts: list of text
    Returns:
        list of {id: value, score: value} per query
    “””
    # Use similarity instance if available otherwise fall back to embeddings model
    if “similarity” in self.pipelines:
        return [[{“id”: uid, “score”: float(score)} for uid, score in r] for r in self.pipelines“similarity”]
    if self.embeddings:
        return [[{“id”: uid, “score”: float(score)} for uid, score in r] for r in self.embeddings.batchsimilarity(queries, texts)]
    return None

list of text

category for instruction-based embeddings

index name, if applicable

embeddings arrays

650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666

def batchtransform(self, texts, category=None, index=None):
    “””
    Transforms list of text into embeddings arrays.
    Args:
        texts: list of text
        category: category for instruction-based embeddings
        index: index name, if applicable
    Returns:
        embeddings arrays
    “””
    if self.embeddings:
        return [[float(x) for x in result] for result in self.embeddings.batchtransform(texts, category, index)]
    return None

number of elements in embeddings index

121
122
123
124
125
126
127
128
129
130
131
132

def count(self):
    “””
    Total number of elements in this embeddings index.
    Returns:
        number of elements in embeddings index
    “””
    if self.cluster:
        return self.cluster.count()
    return super().count()

list of ids to delete

ids deleted

 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105

def delete(self, ids):
    “””
    Deletes from an embeddings index. Returns list of ids deleted.
    Args:
        ids: list of ids to delete
    Returns:
        ids deleted
    “””
    if self.cluster:
        return self.cluster.delete(ids)
    return super().delete(ids)

query text

optional list of text, otherwise runs search query

optional limit if texts is None

list of dict per input text where a higher token scores represents higher importance relative to the query

594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611

def explain(self, query, texts=None, limit=10):
    “””
    Explains the importance of each input token in text for a query.
    Args:
        query: query text
        texts: optional list of text, otherwise runs search query
        limit: optional limit if texts is None
    Returns:
        list of dict per input text where a higher token scores represents higher importance relative to the query
    “””
    if self.embeddings:
        with self.lock:
            return self.embeddings.explain(query, texts, limit)
    return None

list of {name: value, query: value, question: value, snippet: value}

optional list of text

list of {name: value, answer: value}

668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687

def extract(self, queue, texts=None):
    “””
    Extracts answers to input questions.
    Args:
        queue: list of {name: value, query: value, question: value, snippet: value}
        texts: optional list of text
    Returns:
        list of {name: value, answer: value}
    “””
    if self.embeddings and “extractor” in self.pipelines:
        # Get extractor instance
        extractor = self.pipelines[“extractor”]
        # Run extractor and return results as dicts
        return extractor(queue, texts)
    return None

71
72
73
74
75
76
77
78
79

def index(self):
    “””
    Builds an embeddings index for previously batched documents.
    “””
    if self.cluster:
        self.cluster.index()
    else:
        super().index()

text|list

list of labels

list of {id: value, score: value} per text element

689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710

def label(self, text, labels):
    “””
    Applies a zero shot classifier to text using a list of labels. Returns a list of
    {id: value, score: value} sorted by highest score, where id is the index in labels.
    Args:
        text: text|list
        labels: list of labels
    Returns:
        list of {id: value, score: value} per text element
    “””
    if “labels” in self.pipelines:
        # Text is a string
        if isinstance(text, str):
            return [{“id”: uid, “score”: float(score)} for uid, score in self.pipelines“labels”]
        # Text is a list
        return [[{“id”: uid, “score”: float(score)} for uid, score in result] for result in self.pipelines“labels”]
    return None

pipeline name

pipeline positional arguments

pipeline keyword arguments

pipeline results

712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731

def pipeline(self, name, args, **kwargs):
    “””
    Generic pipeline execution method.
    Args:
        name: pipeline name
        args: pipeline positional arguments
        kwargs: pipeline keyword arguments
    Returns:
        pipeline results
    “””
    # Backwards compatible with previous pipeline function arguments
    args = args[0] if args and len(args) == 1 and isinstance(args[0], tuple) else args
    if name in self.pipelines:
        return self.pipelines[name](args, **kwargs)
    return None

new config

optional function to prepare content for indexing

107
108
109
110
111
112
113
114
115
116
117
118
119

def reindex(self, config, function=None):
    “””
    Recreates this embeddings index using config. This method only works if document content storage is enabled.
    Args:
        config: new config
        function: optional function to prepare content for indexing
    “””
    if self.cluster:
        self.cluster.reindex(config, function)
    else:
        super().reindex(config, function)

query text

list of text

list of {id: value, score: value}

550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570

def similarity(self, query, texts):
    “””
    Computes the similarity between query and list of text. Returns a list of
    {id: value, score: value} sorted by highest score, where id is the index
    in texts.
    Args:
        query: query text
        texts: list of text
    Returns:
        list of {id: value, score: value}
    “””
    # Use similarity instance if available otherwise fall back to embeddings model
    if “similarity” in self.pipelines:
        return [{“id”: uid, “score”: float(score)} for uid, score in self.pipelines“similarity”]
    if self.embeddings:
        return [{“id”: uid, “score”: float(score)} for uid, score in self.embeddings.similarity(query, texts)]
    return None

input text

category for instruction-based embeddings

index name, if applicable

embeddings array

632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648

def transform(self, text, category=None, index=None):
    “””
    Transforms text into embeddings arrays.
    Args:
        text: input text
        category: category for instruction-based embeddings
        index: index name, if applicable
    Returns:
        embeddings array
    “””
    if self.embeddings:
        return [float(x) for x in self.embeddings.transform(text, category, index)]
    return None

81
82
83
84
85
86
87
88
89

def upsert(self):
    “””
    Runs an embeddings upsert operation for previously batched documents.
    “””
    if self.cluster:
        self.cluster.upsert()
    else:
        super().upsert()

755
756
757
758
759
760
761
762
763

def wait(self):
    “””
    Closes threadpool and waits for completion.
    “””
    if self.pool:
        self.pool.close()
        self.pool.join()
        self.pool = None

workflow name

elements to process

processed elements

733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753

def workflow(self, name, elements):
    “””
    Executes a workflow.
    Args:
        name: workflow name
        elements: elements to process
    Returns:
        processed elements
    “””
    if hasattr(elements, “len“) and hasattr(elements, “getitem“):
        # Convert to tuples and return as a list since input is sized
        elements = [tuple(element) if isinstance(element, list) else element for element in elements]
    else:
        # Convert to tuples and return as a generator since input is not sized
        elements = (tuple(element) if isinstance(element, list) else element for element in elements)
    # Execute workflow
    return self.workflowsname