diff options
Diffstat (limited to 'gn3')
| -rw-r--r-- | gn3/api/llm.py | 188 | ||||
| -rw-r--r-- | gn3/errors.py | 18 | ||||
| -rw-r--r-- | gn3/llms/client.py | 57 | ||||
| -rw-r--r-- | gn3/llms/errors.py | 7 | ||||
| -rw-r--r-- | gn3/llms/process.py | 121 |
5 files changed, 197 insertions, 194 deletions
diff --git a/gn3/api/llm.py b/gn3/api/llm.py index 442252f..952a5b9 100644 --- a/gn3/api/llm.py +++ b/gn3/api/llm.py @@ -1,125 +1,117 @@ """Api endpoints for gnqa""" -from datetime import timedelta import json -import sqlite3 -from redis import Redis - +from datetime import datetime, timezone from flask import Blueprint from flask import current_app from flask import jsonify from flask import request from gn3.llms.process import get_gnqa -from gn3.llms.process import get_user_queries -from gn3.llms.process import fetch_query_results from gn3.llms.errors import LLMError from gn3.auth.authorisation.oauth2.resource_server import require_oauth from gn3.auth import db + gnqa = Blueprint("gnqa", __name__) -@gnqa.route("/gnqna", methods=["POST"]) -def gnqna(): - """Main gnqa endpoint""" +@gnqa.route("/search", methods=["PUT"]) +def search(): + """Api endpoint for searching queries in fahamu Api""" query = request.json.get("querygnqa", "") if not query: return jsonify({"error": "querygnqa is missing in the request"}), 400 - - try: - fahamu_token = current_app.config.get("FAHAMU_AUTH_TOKEN") - if fahamu_token is None: - return jsonify({"query": query, - "error": "Use of invalid fahamu auth token"}), 500 - task_id, answer, refs = get_gnqa( - query, fahamu_token, current_app.config.get("DATA_DIR")) - response = { - "task_id": task_id, - "query": query, - "answer": answer, - "references": refs - } - with (Redis.from_url(current_app.config["REDIS_URI"], - decode_responses=True) as redis_conn): - redis_conn.setex( - f"LLM:random_user-{query}", - timedelta(days=10), json.dumps(response)) - return jsonify({ - **response, - "prev_queries": get_user_queries("random_user", redis_conn) - }) - except LLMError as error: - return jsonify({"query": query, - "error": f"Request failed-{str(error)}"}), 500 + fahamu_token = current_app.config.get("FAHAMU_AUTH_TOKEN") + if not fahamu_token: + raise LLMError( + "Request failed:an LLM authorisation token is required ", query) + task_id, answer, refs = get_gnqa( + query, fahamu_token, current_app.config.get("DATA_DIR")) + response = { + "task_id": task_id, + "query": query, + "answer": answer, + "references": refs + } + with (db.connection(current_app.config["LLM_DB_PATH"]) as conn, + require_oauth.acquire("profile user") as token): + cursor = conn.cursor() + cursor.execute("""CREATE TABLE IF NOT EXISTS + history(user_id TEXT NOT NULL, + task_id TEXT NOT NULL, + query TEXT NOT NULL, + results TEXT, + created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP, + PRIMARY KEY(task_id)) WITHOUT ROWID""") + cursor.execute( + """INSERT INTO history(user_id, task_id, query, results) + VALUES(?, ?, ?, ?) + """, (str(token.user.user_id), str(task_id["task_id"]), + query, + json.dumps(response)) + ) + return response @gnqa.route("/rating/<task_id>", methods=["POST"]) @require_oauth("profile") -def rating(task_id): - """Endpoint for rating qnqa query and answer""" - try: - llm_db_path = current_app.config["LLM_DB_PATH"] - with (require_oauth.acquire("profile") as token, - db.connection(llm_db_path) as conn): - - results = request.json - user_id, query, answer, weight = (token.user.user_id, - results.get("query"), - results.get("answer"), - results.get("weight", 0)) - cursor = conn.cursor() - create_table = """CREATE TABLE IF NOT EXISTS Rating( - user_id TEXT NOT NULL, - query TEXT NOT NULL, - answer TEXT NOT NULL, - weight INTEGER NOT NULL DEFAULT 0, - task_id TEXT NOT NULL UNIQUE - )""" - cursor.execute(create_table) - cursor.execute("""INSERT INTO Rating(user_id,query, - answer,weight,task_id) - VALUES(?,?,?,?,?) - ON CONFLICT(task_id) DO UPDATE SET - weight=excluded.weight - """, (str(user_id), query, answer, weight, task_id)) +def rate_queries(task_id): + """Api endpoint for rating GNQA query and answer""" + with (require_oauth.acquire("profile") as token, + db.connection(current_app.config["LLM_DB_PATH"]) as conn): + results = request.json + user_id, query, answer, weight = (token.user.user_id, + results.get("query"), + results.get("answer"), + results.get("weight", 0)) + cursor = conn.cursor() + create_table = """CREATE TABLE IF NOT EXISTS Rating( + user_id TEXT NOT NULL, + query TEXT NOT NULL, + answer TEXT NOT NULL, + weight INTEGER NOT NULL DEFAULT 0, + task_id TEXT NOT NULL UNIQUE, + created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP, + PRIMARY KEY(task_id))""" + cursor.execute(create_table) + cursor.execute("""INSERT INTO Rating(user_id, query, + answer, weight, task_id, created_at) + VALUES(?, ?, ?, ?, ?, ?) + ON CONFLICT(task_id) DO UPDATE SET + weight=excluded.weight + """, (str(user_id), query, answer, weight, task_id, + datetime.now(timezone.utc))) return { - "message": - "You have successfully rated this query:Thank you!!" + "message": "You have successfully rated this query.Thank you!" }, 200 - except sqlite3.Error as error: - return jsonify({"error": str(error)}), 500 -@gnqa.route("/history/<query>", methods=["GET"]) +@gnqa.route("/history", methods=["GET", "DELETE"]) @require_oauth("profile user") -def fetch_user_hist(query): - """"Endpoint to fetch previos searches for User""" - with (require_oauth.acquire("profile user") as the_token, - Redis.from_url(current_app.config["REDIS_URI"], - decode_responses=True) as redis_conn): - return jsonify({ - **fetch_query_results(query, the_token.user.id, redis_conn), - "prev_queries": get_user_queries("random_user", redis_conn) - }) - - -@gnqa.route("/historys/<query>", methods=["GET"]) -def fetch_users_hist_records(query): - """method to fetch all users hist:note this is a test functionality - to be replaced by fetch_user_hist - """ - with Redis.from_url(current_app.config["REDIS_URI"], - decode_responses=True) as redis_conn: - return jsonify({ - **fetch_query_results(query, "random_user", redis_conn), - "prev_queries": get_user_queries("random_user", redis_conn) - }) - - -@gnqa.route("/get_hist_names", methods=["GET"]) -def fetch_prev_hist_ids(): - """Test method for fetching history for Anony Users""" - with (Redis.from_url(current_app.config["REDIS_URI"], - decode_responses=True)) as redis_conn: - return jsonify({"prev_queries": get_user_queries("random_user", - redis_conn)}) +def fetch_prev_history(): + """Api endpoint to fetch GNQA previous search.""" + with (require_oauth.acquire("profile user") as token, + db.connection(current_app.config["LLM_DB_PATH"]) as conn): + cursor = conn.cursor() + if request.method == "DELETE": + task_ids = list(request.json.values()) + query = """DELETE FROM history + WHERE task_id IN ({}) + and user_id=?""".format(",".join("?" * len(task_ids))) + cursor.execute(query, (*task_ids, str(token.user.user_id),)) + return jsonify({}) + elif (request.method == "GET" and + request.args.get("search_term")): + cursor.execute( + """SELECT results from history + Where task_id=? and user_id=?""", + (request.args.get("search_term"), + str(token.user.user_id),)) + record = cursor.fetchone() + if record: + return dict(record).get("results") + return {} + cursor.execute( + """SELECT task_id,query from history WHERE user_id=?""", + (str(token.user.user_id),)) + return jsonify([dict(item) for item in cursor.fetchall()]) diff --git a/gn3/errors.py b/gn3/errors.py index 1833bf6..c53604f 100644 --- a/gn3/errors.py +++ b/gn3/errors.py @@ -16,7 +16,7 @@ from authlib.oauth2.rfc6749.errors import OAuth2Error from flask import Flask, jsonify, Response, current_app from gn3.auth.authorisation.errors import AuthorisationError - +from gn3.llms.errors import LLMError def add_trace(exc: Exception, jsonmsg: dict) -> dict: """Add the traceback to the error handling object.""" @@ -106,6 +106,21 @@ def handle_generic(exc: Exception) -> Response: return resp +def handle_llm_error(exc: Exception) -> Response: + """ Handle llm erros if not handled anywhere else. """ + current_app.logger.error(exc) + resp = jsonify({ + "query": exc.args[1], + "error_type": type(exc).__name__, + "error": ( + exc.args[0] if bool(exc.args) else "Fahamu gnqa error occurred" + ), + "trace": traceback.format_exc() + }) + resp.status_code = 500 + return resp + + def register_error_handlers(app: Flask): """Register application-level error handlers.""" app.register_error_handler(NotFound, page_not_found) @@ -115,6 +130,7 @@ def register_error_handlers(app: Flask): app.register_error_handler(AuthorisationError, handle_authorisation_error) app.register_error_handler(RemoteDisconnected, internal_server_error) app.register_error_handler(URLError, url_server_error) + app.register_error_handler(LLMError, handle_llm_error) for exc in ( EndPointInternalError, EndPointNotFound, diff --git a/gn3/llms/client.py b/gn3/llms/client.py index d57bca2..ad6c400 100644 --- a/gn3/llms/client.py +++ b/gn3/llms/client.py @@ -55,6 +55,7 @@ class GeneNetworkQAClient(Session): self.base_url = "https://genenetwork.fahamuai.com/api/tasks" self.answer_url = f"{self.base_url}/answers" self.feedback_url = f"{self.base_url}/feedback" + self.query = "" adapter = TimeoutHTTPAdapter( timeout=timeout, @@ -83,44 +84,44 @@ class GeneNetworkQAClient(Session): """ handler for non 200 response from fahamu api""" return f"Error: Status code -{response.status_code}- Reason::{response.reason}" - def ask(self, ex_url, *args, **kwargs): + def ask(self, ex_url, query, *args, **kwargs): """fahamu ask api interface""" + self.query = query res = self.custom_request('POST', f"{self.base_url}{ex_url}", *args, **kwargs) - if res.status_code != 200: - return f"Error: Status code -{res.status_code}- Reason::{res.reason}", 0 return res, json.loads(res.text) def get_answer(self, taskid, *args, **kwargs): """Fahamu get answer interface""" - try: - query = f"{self.answer_url}?task_id={taskid['task_id']}" - res = self.custom_request('GET', query, *args, **kwargs) - if res.status_code != 200: - return f"Error: Status code -{res.status_code}- Reason::{res.reason}", 0 - return res, 1 - except TimeoutError: - return "Timeout error occured:try to rephrase your query", 0 + query = f"{self.answer_url}?task_id={taskid['task_id']}" + res = self.custom_request('GET', query, *args, **kwargs) + return res, 1 def custom_request(self, method, url, *args, **kwargs): """ make custom request to fahamu api ask and get response""" max_retries = 50 retry_delay = 3 + response_msg = { + 404: "Api endpoint Does not exist", + 500: "Use of Invalid Token/or the Fahamu Api is currently down", + 400: "You sent a bad Fahamu request", + 401: "You do not have authorization to perform the request", + } for _i in range(max_retries): - try: - response = super().request(method, url, *args, **kwargs) - response.raise_for_status() - if response.ok: - if method.lower() == "get" and response.json().get("data") is None: - time.sleep(retry_delay) - continue - return response - else: + response = super().request(method, url, *args, **kwargs) + if response.ok: + if method.lower() == "get" and not response.json().get("data"): + # note this is a dirty trick to check if fahamu has returned the results + # the issue is that the api only returns 500 or 200 satus code + # TODO: fix this on their end time.sleep(retry_delay) - except requests.exceptions.HTTPError as error: - if error.response.status_code == 500: - raise LLMError(error.request, error.response, f"Response Error with:status_code:{error.response.status_code},Reason for error: Use of Invalid Fahamu Token") from error - raise LLMError(error.request, error.response, - f"HTTP error occurred with error status:{error.response.status_code}") from error - except requests.exceptions.RequestException as error: - raise error - raise TimeoutError + continue + return response + else: + raise LLMError(f"Request error with code:\ + {response.status_code} occurred with reason:\ + {response_msg.get(response.status_code,response.reason)}", + self.query) + #time.sleep(retry_delay) + raise LLMError("Timeout error: We couldn't provide a response,Please try\ + to rephrase your question to receive feedback", + self.query) diff --git a/gn3/llms/errors.py b/gn3/llms/errors.py index af3d7b0..a3a47a3 100644 --- a/gn3/llms/errors.py +++ b/gn3/llms/errors.py @@ -35,8 +35,5 @@ class UnprocessableEntity(HTTPError): msg, request=request, response=response) -class LLMError(HTTPError): - """Custom error from making Fahamu APi request """ - def __init__(self, request, response, msg): - super(HTTPError, self).__init__( - msg, request=request, response=response) +class LLMError(Exception): + """custom exception for LLMErrorMIxins""" diff --git a/gn3/llms/process.py b/gn3/llms/process.py index 4725bcb..55c27a0 100644 --- a/gn3/llms/process.py +++ b/gn3/llms/process.py @@ -1,6 +1,7 @@ """this module contains code for processing response from fahamu client.py""" # pylint: disable=C0301 import os +import re import string import json import logging @@ -21,21 +22,12 @@ class DocIDs(): * doc_ids.json: opens doc)ids for gn references * sugar_doc_ids: open doci_ids for diabetes references """ - self.doc_ids = self.load_file("doc_ids.json") - self.sugar_doc_ids = self.load_file("all_files.json") + self.doc_ids = load_file("doc_ids.json", BASEDIR) + self.sugar_doc_ids = load_file("all_files.json", BASEDIR) self.format_doc_ids(self.sugar_doc_ids) - def load_file(self, file_name): - """Method to load and read doc_id files""" - file_path = os.path.join(BASEDIR, file_name) - if os.path.isfile(file_path): - with open(file_path, "rb") as file_handler: - return json.load(file_handler) - else: - raise FileNotFoundError(f"{file_path}-- FIle does not exist\n") - def format_doc_ids(self, docs): - """method to format doc_ids for list items""" + """method to format doc_ids for list items doc_id and doc_name""" for _key, val in docs.items(): if isinstance(val, list): for doc_obj in val: @@ -43,7 +35,14 @@ class DocIDs(): self.doc_ids.update({doc_obj["id"]: doc_name}) def get_info(self, doc_id): - """ interface to make read from doc_ids""" + """ interface to make read from doc_ids + and extract info data else returns + doc_id + Args: + doc_id: str: a search key for doc_ids + Returns: + an object with doc_info if doc_id in doc_ids + """ if doc_id in self.doc_ids.keys(): return self.doc_ids[doc_id] else: @@ -51,7 +50,8 @@ class DocIDs(): def format_bibliography_info(bib_info): - """Function for formatting bibliography info""" + """Utility function for formatting bibliography info + """ if isinstance(bib_info, str): return bib_info.removesuffix('.txt') elif isinstance(bib_info, dict): @@ -59,14 +59,16 @@ def format_bibliography_info(bib_info): return bib_info -def filter_response_text(val): - """helper function for filtering non-printable chars""" - return json.loads(''.join([str(char) - for char in val if char in string.printable])) - - def parse_context(context, get_info_func, format_bib_func): - """function to parse doc_ids content""" + """Function to parse doc_ids content + Args: + context: raw references from fahamu api + get_info_func: function to get doc_ids info + format_bib_func: function to foramt bibliography info + Returns: + an list with each item having (doc_id,bib_info, + combined reference text) + """ results = [] for doc_ids, summary in context.items(): combo_txt = "" @@ -75,13 +77,23 @@ def parse_context(context, get_info_func, format_bib_func): doc_info = get_info_func(doc_ids) bib_info = doc_ids if doc_ids == doc_info else format_bib_func( doc_info) + pattern = r'(https?://|www\.)[\w.-]+(\.[a-zA-Z]{2,})([/\w.-]*)*' + combo_text = re.sub(pattern, + lambda x: f"<a href='{x[0]}' target=_blank> {x[0]} </a>", + combo_txt) results.append( - {"doc_id": doc_ids, "bibInfo": bib_info, "comboTxt": combo_txt}) + {"doc_id": doc_ids, "bibInfo": bib_info, + "comboTxt": combo_text}) return results def load_file(filename, dir_path): - """function to open and load json file""" + """Utility function to read json file + Args: + filename: file name to read + dir_path: base directory for the file + Returns: json data read to a dict + """ file_path = os.path.join(dir_path, f"{filename}") if not os.path.isfile(file_path): raise FileNotFoundError(f"{filename} was not found or is a directory") @@ -90,8 +102,19 @@ def load_file(filename, dir_path): def fetch_pubmed(references, file_name, data_dir=""): - """method to fetch and populate references with pubmed""" + """ + Fetches PubMed data from a JSON file and populates the\ + references dictionary. + + Args: + references (dict): Dictionary with document IDs as keys\ + and reference data as values. + filename (str): Name of the JSON file containing PubMed data. + data_dir (str): Base directory where the data files are located. + Returns: + dict: Updated references dictionary populated with the PubMed data. + """ try: pubmed = load_file(file_name, os.path.join(data_dir, "gn-meta/lit")) for reference in references: @@ -116,42 +139,16 @@ def get_gnqa(query, auth_token, data_dir=""): answer references: contains doc_name,reference,pub_med_info """ - api_client = GeneNetworkQAClient(api_key=auth_token) - res, task_id = api_client.ask('?ask=' + quote(query), auth_token) - if task_id == 0: - raise RuntimeError(f"Error connecting to Fahamu Api: {str(res)}") - res, status = api_client.get_answer(task_id) - if status == 1: - resp_text = filter_response_text(res.text) - if resp_text.get("data") is None: - return task_id, "Please try to rephrase your question to receive feedback", [] - answer = resp_text['data']['answer'] - context = resp_text['data']['context'] - references = parse_context( - context, DocIDs().get_info, format_bibliography_info) - references = fetch_pubmed(references, "pubmed.json", data_dir) - - return task_id, answer, references - else: - return task_id, "We couldn't provide a response,Please try to rephrase your question to receive feedback", [] - - -def fetch_query_results(query, user_id, redis_conn): - """this method fetches prev user query searches""" - result = redis_conn.get(f"LLM:{user_id}-{query}") - if result: - return json.loads(result) - return { - "query": query, - "answer": "Sorry No answer for you", - "references": [], - "task_id": None - } - - -def get_user_queries(user_id, redis_conn): - """methos to fetch all queries for a specific user""" - results = redis_conn.keys(f"LLM:{user_id}*") - return [query for query in - [result.partition("-")[2] for result in results] if query != ""] + res, task_id = api_client.ask('?ask=' + quote(query), query=query) + res, _status = api_client.get_answer(task_id) + resp_text = json.loads(''.join([str(char) + for char in res.text if char in string.printable])) + answer = re.sub(r'(https?://|www\.)[\w.-]+(\.[a-zA-Z]{2,})([/\w.-]*)*', + lambda x: f"<a href='{x[0]}' target=_blank> {x[0]} </a>", + resp_text["data"]["answer"]) + context = resp_text['data']['context'] + return task_id, answer, fetch_pubmed(parse_context( + context, DocIDs().get_info, + format_bibliography_info), + "pubmed.json", data_dir) |