from __future__ import annotations import json import re from hashlib import md5 from typing import TYPE_CHECKING, Any, Dict, List, NamedTuple, Tuple, Union from langchain_community.graphs.graph_document import GraphDocument from langchain_community.graphs.graph_store import GraphStore if TYPE_CHECKING: import psycopg2.extras class AGEQueryException(Exception): """Exception for the AGE queries.""" def __init__(self, exception: Union[str, Dict]) -> None: if isinstance(exception, dict): self.message = exception["message"] if "message" in exception else "unknown" self.details = exception["details"] if "details" in exception else "unknown" else: self.message = exception self.details = "unknown" def get_message(self) -> str: return self.message def get_details(self) -> Any: return self.details class AGEGraph(GraphStore): """ Apache AGE wrapper for graph operations. Args: graph_name (str): the name of the graph to connect to or create conf (Dict[str, Any]): the pgsql connection config passed directly to psycopg2.connect create (bool): if True and graph doesn't exist, attempt to create it *Security note*: Make sure that the database connection uses credentials that are narrowly-scoped to only include necessary permissions. Failure to do so may result in data corruption or loss, since the calling code may attempt commands that would result in deletion, mutation of data if appropriately prompted or reading sensitive data if such data is present in the database. The best way to guard against such negative outcomes is to (as appropriate) limit the permissions granted to the credentials used with this tool. See https://python.langchain.com/docs/security for more information. """ # python type mapping for providing readable types to LLM types = { "str": "STRING", "float": "DOUBLE", "int": "INTEGER", "list": "LIST", "dict": "MAP", "bool": "BOOLEAN", } # precompiled regex for checking chars in graph labels label_regex = re.compile("[^0-9a-zA-Z]+") def __init__( self, graph_name: str, conf: Dict[str, Any], create: bool = True ) -> None: """Create a new AGEGraph instance.""" self.graph_name = graph_name # check that psycopg2 is installed try: import psycopg2 except ImportError: raise ImportError( "Could not import psycopg2 python package. " "Please install it with `pip install psycopg2`." ) self.connection = psycopg2.connect(**conf) with self._get_cursor() as curs: # check if graph with name graph_name exists graph_id_query = ( """SELECT graphid FROM ag_catalog.ag_graph WHERE name = '{}'""".format( graph_name ) ) curs.execute(graph_id_query) data = curs.fetchone() # if graph doesn't exist and create is True, create it if data is None: if create: create_statement = """ SELECT ag_catalog.create_graph('{}'); """.format(graph_name) try: curs.execute(create_statement) self.connection.commit() except psycopg2.Error as e: raise AGEQueryException( { "message": "Could not create the graph", "detail": str(e), } ) else: raise Exception( ( 'Graph "{}" does not exist in the database ' + 'and "create" is set to False' ).format(graph_name) ) curs.execute(graph_id_query) data = curs.fetchone() # store graph id and refresh the schema self.graphid = data.graphid self.refresh_schema() def _get_cursor(self) -> psycopg2.extras.NamedTupleCursor: """ get cursor, load age extension and set search path """ try: import psycopg2.extras except ImportError as e: raise ImportError( "Unable to import psycopg2, please install with " "`pip install -U psycopg2`." ) from e cursor = self.connection.cursor(cursor_factory=psycopg2.extras.NamedTupleCursor) cursor.execute("""LOAD 'age';""") cursor.execute("""SET search_path = ag_catalog, "$user", public;""") return cursor def _get_labels(self) -> Tuple[List[str], List[str]]: """ Get all labels of a graph (for both edges and vertices) by querying the graph metadata table directly Returns Tuple[List[str]]: 2 lists, the first containing vertex labels and the second containing edge labels """ e_labels_records = self.query( """MATCH ()-[e]-() RETURN collect(distinct label(e)) as labels""" ) e_labels = e_labels_records[0]["labels"] if e_labels_records else [] n_labels_records = self.query( """MATCH (n) RETURN collect(distinct label(n)) as labels""" ) n_labels = n_labels_records[0]["labels"] if n_labels_records else [] return n_labels, e_labels def _get_triples(self, e_labels: List[str]) -> List[Dict[str, str]]: """ Get a set of distinct relationship types (as a list of dicts) in the graph to be used as context by an llm. Args: e_labels (List[str]): a list of edge labels to filter for Returns: List[Dict[str, str]]: relationships as a list of dicts in the format "{'start':, 'type':, 'end':}" """ # age query to get distinct relationship types try: import psycopg2 except ImportError as e: raise ImportError( "Unable to import psycopg2, please install with " "`pip install -U psycopg2`." ) from e triple_query = """ SELECT * FROM ag_catalog.cypher('{graph_name}', $$ MATCH (a)-[e:`{e_label}`]->(b) WITH a,e,b LIMIT 3000 RETURN DISTINCT labels(a) AS from, type(e) AS edge, labels(b) AS to LIMIT 10 $$) AS (f agtype, edge agtype, t agtype); """ triple_schema = [] # iterate desired edge types and add distinct relationship types to result with self._get_cursor() as curs: for label in e_labels: q = triple_query.format(graph_name=self.graph_name, e_label=label) try: curs.execute(q) data = curs.fetchall() for d in data: # use json.loads to convert returned # strings to python primitives triple_schema.append( { "start": json.loads(d.f)[0], "type": json.loads(d.edge), "end": json.loads(d.t)[0], } ) except psycopg2.Error as e: raise AGEQueryException( { "message": "Error fetching triples", "detail": str(e), } ) return triple_schema def _get_triples_str(self, e_labels: List[str]) -> List[str]: """ Get a set of distinct relationship types (as a list of strings) in the graph to be used as context by an llm. Args: e_labels (List[str]): a list of edge labels to filter for Returns: List[str]: relationships as a list of strings in the format "(:``)-[:``]->(:``)" """ triples = self._get_triples(e_labels) return self._format_triples(triples) @staticmethod def _format_triples(triples: List[Dict[str, str]]) -> List[str]: """ Convert a list of relationships from dictionaries to formatted strings to be better readable by an llm Args: triples (List[Dict[str,str]]): a list relationships in the form {'start':, 'type':, 'end':} Returns: List[str]: a list of relationships in the form "(:``)-[:``]->(:``)" """ triple_template = "(:`{start}`)-[:`{type}`]->(:`{end}`)" triple_schema = [triple_template.format(**triple) for triple in triples] return triple_schema def _get_node_properties(self, n_labels: List[str]) -> List[Dict[str, Any]]: """ Fetch a list of available node properties by node label to be used as context for an llm Args: n_labels (List[str]): a list of node labels to filter for Returns: List[Dict[str, Any]]: a list of node labels and their corresponding properties in the form "{ 'labels': , 'properties': [ { 'property': , 'type': },... ] }" """ try: import psycopg2 except ImportError as e: raise ImportError( "Unable to import psycopg2, please install with " "`pip install -U psycopg2`." ) from e # cypher query to fetch properties of a given label node_properties_query = """ SELECT * FROM ag_catalog.cypher('{graph_name}', $$ MATCH (a:`{n_label}`) RETURN properties(a) AS props LIMIT 100 $$) AS (props agtype); """ node_properties = [] with self._get_cursor() as curs: for label in n_labels: q = node_properties_query.format( graph_name=self.graph_name, n_label=label ) try: curs.execute(q) except psycopg2.Error as e: raise AGEQueryException( { "message": "Error fetching node properties", "detail": str(e), } ) data = curs.fetchall() # build a set of distinct properties s = set({}) for d in data: # use json.loads to convert to python # primitive and get readable type for k, v in json.loads(d.props).items(): s.add((k, self.types[type(v).__name__])) np = { "properties": [{"property": k, "type": v} for k, v in s], "labels": label, } node_properties.append(np) return node_properties def _get_edge_properties(self, e_labels: List[str]) -> List[Dict[str, Any]]: """ Fetch a list of available edge properties by edge label to be used as context for an llm Args: e_labels (List[str]): a list of edge labels to filter for Returns: List[Dict[str, Any]]: a list of edge labels and their corresponding properties in the form "{ 'labels': , 'properties': [ { 'property': , 'type': },... ] }" """ try: import psycopg2 except ImportError as e: raise ImportError( "Unable to import psycopg2, please install with " "`pip install -U psycopg2`." ) from e # cypher query to fetch properties of a given label edge_properties_query = """ SELECT * FROM ag_catalog.cypher('{graph_name}', $$ MATCH ()-[e:`{e_label}`]->() RETURN properties(e) AS props LIMIT 100 $$) AS (props agtype); """ edge_properties = [] with self._get_cursor() as curs: for label in e_labels: q = edge_properties_query.format( graph_name=self.graph_name, e_label=label ) try: curs.execute(q) except psycopg2.Error as e: raise AGEQueryException( { "message": "Error fetching edge properties", "detail": str(e), } ) data = curs.fetchall() # build a set of distinct properties s = set({}) for d in data: # use json.loads to convert to python # primitive and get readable type for k, v in json.loads(d.props).items(): s.add((k, self.types[type(v).__name__])) np = { "properties": [{"property": k, "type": v} for k, v in s], "type": label, } edge_properties.append(np) return edge_properties def refresh_schema(self) -> None: """ Refresh the graph schema information by updating the available labels, relationships, and properties """ # fetch graph schema information n_labels, e_labels = self._get_labels() triple_schema = self._get_triples(e_labels) node_properties = self._get_node_properties(n_labels) edge_properties = self._get_edge_properties(e_labels) # update the formatted string representation self.schema = f""" Node properties are the following: {node_properties} Relationship properties are the following: {edge_properties} The relationships are the following: {self._format_triples(triple_schema)} """ # update the dictionary representation self.structured_schema = { "node_props": {el["labels"]: el["properties"] for el in node_properties}, "rel_props": {el["type"]: el["properties"] for el in edge_properties}, "relationships": triple_schema, "metadata": {}, } @property def get_schema(self) -> str: """Returns the schema of the Graph""" return self.schema @property def get_structured_schema(self) -> Dict[str, Any]: """Returns the structured schema of the Graph""" return self.structured_schema @staticmethod def _get_col_name(field: str, idx: int) -> str: """ Convert a cypher return field to a pgsql select field If possible keep the cypher column name, but create a generic name if necessary Args: field (str): a return field from a cypher query to be formatted for pgsql idx (int): the position of the field in the return statement Returns: str: the field to be used in the pgsql select statement """ # remove white space field = field.strip() # if an alias is provided for the field, use it if " as " in field: return field.split(" as ")[-1].strip() # if the return value is an unnamed primitive, give it a generic name elif field.isnumeric() or field in ("true", "false", "null"): return f"column_{idx}" # otherwise return the value stripping out some common special chars else: return field.replace("(", "_").replace(")", "") @staticmethod def _wrap_query(query: str, graph_name: str) -> str: """ Convert a cypher query to an Apache Age compatible sql query by wrapping the cypher query in ag_catalog.cypher, casting results to agtype and building a select statement Args: query (str): a valid cypher query graph_name (str): the name of the graph to query Returns: str: an equivalent pgsql query """ # pgsql template template = """SELECT {projection} FROM ag_catalog.cypher('{graph_name}', $$ {query} $$) AS ({fields});""" # if there are any returned fields they must be added to the pgsql query if "return" in query.lower(): # parse return statement to identify returned fields fields = ( query.lower() .split("return")[-1] .split("distinct")[-1] .split("order by")[0] .split("skip")[0] .split("limit")[0] .split(",") ) # raise exception if RETURN * is found as we can't resolve the fields if "*" in [x.strip() for x in fields]: raise ValueError( "AGE graph does not support 'RETURN *'" + " statements in Cypher queries" ) # get pgsql formatted field names fields = [ AGEGraph._get_col_name(field, idx) for idx, field in enumerate(fields) ] # build resulting pgsql relation fields_str = ", ".join( [field.split(".")[-1] + " agtype" for field in fields] ) # if no return statement we still need to return a single field of type agtype else: fields_str = "a agtype" select_str = "*" return template.format( graph_name=graph_name, query=query, fields=fields_str, projection=select_str, ) @staticmethod def _record_to_dict(record: NamedTuple) -> Dict[str, Any]: """ Convert a record returned from an age query to a dictionary Args: record (): a record from an age query result Returns: Dict[str, Any]: a dictionary representation of the record where the dictionary key is the field name and the value is the value converted to a python type """ # result holder d = {} # prebuild a mapping of vertex_id to vertex mappings to be used # later to build edges vertices = {} for k in record._fields: v = getattr(record, k) # agtype comes back '{key: value}::type' which must be parsed if isinstance(v, str) and "::" in v: dtype = v.split("::")[-1] v = v.split("::")[0] if dtype == "vertex": vertex = json.loads(v) vertices[vertex["id"]] = vertex.get("properties") # iterate returned fields and parse appropriately for k in record._fields: v = getattr(record, k) if isinstance(v, str) and "::" in v: dtype = v.split("::")[-1] v = v.split("::")[0] else: dtype = "" if dtype == "vertex": d[k] = json.loads(v).get("properties") # convert edge from id-label->id by replacing id with node information # we only do this if the vertex was also returned in the query # this is an attempt to be consistent with neo4j implementation elif dtype == "edge": edge = json.loads(v) d[k] = ( vertices.get(edge["start_id"], {}), edge["label"], vertices.get(edge["end_id"], {}), ) else: d[k] = json.loads(v) if isinstance(v, str) else v return d def query(self, query: str, params: dict = {}) -> List[Dict[str, Any]]: """ Query the graph by taking a cypher query, converting it to an age compatible query, executing it and converting the result Args: query (str): a cypher query to be executed params (dict): parameters for the query (not used in this implementation) Returns: List[Dict[str, Any]]: a list of dictionaries containing the result set """ try: import psycopg2 except ImportError as e: raise ImportError( "Unable to import psycopg2, please install with " "`pip install -U psycopg2`." ) from e # convert cypher query to pgsql/age query wrapped_query = self._wrap_query(query, self.graph_name) # execute the query, rolling back on an error with self._get_cursor() as curs: try: curs.execute(wrapped_query) self.connection.commit() except psycopg2.Error as e: self.connection.rollback() raise AGEQueryException( { "message": "Error executing graph query: {}".format(query), "detail": str(e), } ) data = curs.fetchall() if data is None: result = [] # convert to dictionaries else: result = [self._record_to_dict(d) for d in data] return result @staticmethod def _format_properties( properties: Dict[str, Any], id: Union[str, None] = None ) -> str: """ Convert a dictionary of properties to a string representation that can be used in a cypher query insert/merge statement. Args: properties (Dict[str,str]): a dictionary containing node/edge properties id (Union[str, None]): the id of the node or None if none exists Returns: str: the properties dictionary as a properly formatted string """ props = [] # wrap property key in backticks to escape for k, v in properties.items(): prop = f"`{k}`: {json.dumps(v)}" props.append(prop) if id is not None and "id" not in properties: props.append( f"id: {json.dumps(id)}" if isinstance(id, str) else f"id: {id}" ) return "{" + ", ".join(props) + "}" @staticmethod def clean_graph_labels(label: str) -> str: """ remove any disallowed characters from a label and replace with '_' Args: label (str): the original label Returns: str: the sanitized version of the label """ return re.sub(AGEGraph.label_regex, "_", label) def add_graph_documents( self, graph_documents: List[GraphDocument], include_source: bool = False ) -> None: """ insert a list of graph documents into the graph Args: graph_documents (List[GraphDocument]): the list of documents to be inserted include_source (bool): if True add nodes for the sources with MENTIONS edges to the entities they mention Returns: None """ # query for inserting nodes node_insert_query = ( """ MERGE (n:`{label}` {properties}) """ if not include_source else """ MERGE (n:`{label}` {properties}) MERGE (d:Document {d_properties}) MERGE (d)-[:MENTIONS]->(n) """ ) # query for inserting edges edge_insert_query = """ MERGE (from:`{f_label}` {f_properties}) MERGE (to:`{t_label}` {t_properties}) MERGE (from)-[:`{r_label}` {r_properties}]->(to) """ # iterate docs and insert them for doc in graph_documents: # if we are adding sources, create an id for the source if include_source: if not doc.source.metadata.get("id"): doc.source.metadata["id"] = md5( doc.source.page_content.encode("utf-8") ).hexdigest() # insert entity nodes for node in doc.nodes: node.properties["id"] = node.id if include_source: query = node_insert_query.format( label=node.type, properties=self._format_properties(node.properties), d_properties=self._format_properties(doc.source.metadata), ) else: query = node_insert_query.format( label=AGEGraph.clean_graph_labels(node.type), properties=self._format_properties(node.properties), ) self.query(query) # insert relationships for edge in doc.relationships: edge.source.properties["id"] = edge.source.id edge.target.properties["id"] = edge.target.id inputs = { "f_label": AGEGraph.clean_graph_labels(edge.source.type), "f_properties": self._format_properties(edge.source.properties), "t_label": AGEGraph.clean_graph_labels(edge.target.type), "t_properties": self._format_properties(edge.target.properties), "r_label": AGEGraph.clean_graph_labels(edge.type).upper(), "r_properties": self._format_properties(edge.properties), } query = edge_insert_query.format(**inputs) self.query(query)