Spaces:
Build error
Build error
File size: 7,965 Bytes
891fbea |
1 2 3 4 5 6 7 8 9 10 11 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 |
import json
import os
import ast
from pathlib import Path
from typing import List
import langchain.llms as LLM
from app.custom_langchain import ModAPIChain, APIResponse, FlexAPIChain, FlexAPIChainPayload
from app.utils import find_value, custom_api_prompt
from app.custom_api_prompts import API_REQUEST_PROMPT, \
API_REQUEST_PROMPT2, API_RESPONSE_PROMPT, API_RESPONSE_PROMPT2, PERMISSION_PROMPT
PATH = Path(os.path.abspath(os.path.dirname(__file__)))
# Open the YAML file
api_doc_path = PATH / 'dbr_api_docs'
def prepare_api_docs():
"""
Add environment variables to the individual
API docs.
"""
updated_api_docs = {}
for filename in os.listdir(api_doc_path):
with open(os.path.join(api_doc_path, filename)) as f:
updated_api_docs[filename] = f.read()
updated_api_docs[filename] = updated_api_docs[filename].replace(
"{DATABRICKS_HOST}", os.getenv('DATABRICKS_HOST'))
return updated_api_docs
def get_run(llm: LLM, runID: str, query: str, api_text: str, headers: dict) -> str:
"""
Get infos from an MLflow run.
Args:
llm (LLM): Large language model
runID (str): Run ID
query (str): Query
api_text (str): API documentation for a particular endpoint
headers (dict): Authorization for the Databricks API
Return:
(str): the response of the language model
"""
mlflow_run_get_chain = APIResponse.from_llm_and_api_docs(llm, api_text, headers=headers, verbose=True)
run_infos = json.loads(mlflow_run_get_chain.run(
f"""Please give me the infos about the run '{runID}'"""))
# Please return the f1 score
return llm(f"""{query} of the metrics : {find_value(run_infos, 'metrics')}""")
def trans_model(llm: LLM, query: str, api_text: str, headers: dict):
"""
Transition a MLflow model to another stage.
Args:
llm (LLM): Large language model
query (str): Query for the model
api_text (str): API documentation for a particular endpoint
headers (dict): Authorization for the Databricks API
"""
mlflow_model_transition_chain = FlexAPIChainPayload.from_llm_and_api_docs(
llm, api_text, headers=headers, api_url_prompt=API_REQUEST_PROMPT2,
api_response_prompt=API_RESPONSE_PROMPT2, verbose=True
)
mlflow_model_transition_chain.run(query)
def get_model(llm: LLM, model_query: str, api_text: str, headers: dict) -> str:
"""
Get all infos about a model and its particular stages
Args:
llm (LLM): Large language model
model_query (str): Query
api_text (str): API documentation for a particular endpoint
headers (dict): Authorization for the Databricks API
Return:
(str): response of the language model
"""
mlflow_model_get_chain = APIResponse.from_llm_and_api_docs(llm, api_text, headers=headers, verbose=True)
api_result = mlflow_model_get_chain.run(model_query)
return llm(
f"""What is the runID and their version of the latest version with stage 'None' \
and stage 'Production' respectively, given the following info: {api_result}"""
)
def batch_mod_permission(
logger, llm: LLM, updated_api_docs: dict, api_text: str, api_name: str,
headers: dict, permission_mod: str, jobs: List[str]
):
"""
Get or modify permissions for a batch of jobs.
Args:
logger (Logger): Logger object
llm (LLM): Large language model
updated_api_docs (dict): Updated API docs
api_text (str): API documentation for a particular endpoint
api_name (str): API name
headers (dict): Authorization for the Databricks API
permission_mod (str): Permission modification
jobs (list): List of jobs
"""
get_permission_txt = updated_api_docs['get_permissions.txt']
for job in jobs:
logger.info(f"Permission for job: {job}")
mod_permissions(
logger, llm, permission_mod, job, headers,
get_permission_txt, api_text, api_name
)
def get_permissions(logger, llm: LLM, api_text: str, headers: dict, jobID: str) -> dict:
"""
Get the permissions of a particular job.
Args:
logger (Logger): Logger object
llm (LLM): Large language model
api_text (str): API documentation for a particular endpoint
headers (dict): Authorization for the Databricks API
jobID (str): ID of the Databricks job
Return:
(dict): Access control list
"""
init_query = f"""Get the jobs permissions for jobID {jobID}"""
permission_get_chain = APIResponse.from_llm_and_api_docs(llm, api_text, headers=headers, verbose=True)
acc_control_list = {'access_control_list': json.loads(permission_get_chain.run(init_query))['access_control_list']}
logger.info(acc_control_list)
return acc_control_list
def mod_permissions(
logger,
llm: LLM,
permission_mod: str,
jobID: str, headers: dict,
get_permission_txt: str,
api_text: str, api_name: str
):
"""
Modify permissions for a job.
Args:
logger (Logger): Logger object
llm (LLM): Large language model
permission_mod (str): Permission modification
jobID (str): ID of the Databricks job
headers (dict): Authorization for the Databricks API
get_permission_txt (str): API text for getting permissions
api_text (str): API documentation for a particular endpoint
api_name (str): file name of the API documentation
"""
acc_control_list = get_permissions(logger, llm, get_permission_txt, headers, jobID)
# Update the permission, if the query demands it
if api_name == 'update_permissions.txt':
permission_prompt= PERMISSION_PROMPT.format(
acc_control_list=acc_control_list, permission_mod=permission_mod)
output_acc_list = llm(permission_prompt)
patch_payload = ast.literal_eval(output_acc_list.rstrip().lstrip()) #
# Reformat the general permission response into the form for the patch payload
for idx, _ in enumerate(patch_payload['access_control_list']):
patch_payload['access_control_list'][idx]['permission_level'] = patch_payload['access_control_list'][idx]['all_permissions'][0]['permission_level']
patch_payload['access_control_list'][idx].pop('all_permissions')
permission_update_chain = FlexAPIChain.from_llm_and_api_docs(
llm, api_text, headers=headers, verbose=True,
api_url_prompt=API_REQUEST_PROMPT, api_response_prompt=API_RESPONSE_PROMPT
) # use updated API Chain which can do post, put and patch also....
permission_update_chain.body = patch_payload
logger.info(f"Payload for permission update: {patch_payload}")
permission_update_chain(f"""Update the jobs permissions of jobID {jobID}""")
def get_job_infos(
llm: LLM, query: str, response_query: str, api_text: str, headers: dict
) -> str:
"""
Get particular infos about the list of existing jobs.
Args:
llm (LLM): Large language model
query (str): Query to get the list of jobs
response_query (str): Query to get the response
api_text (str): API documentation for a particular endpoint
headers (dict): Authorization for the Databricks API
Return:
(str): response from the large language model
"""
job_chain = ModAPIChain.from_llm_and_api_docs(
llm, api_text, headers=headers, verbose=True
)
custom_resp = f"This response contains a list of databricks jobs. {response_query}"
jobs_result = custom_api_prompt(llm, job_chain, query, custom_resp)
return jobs_result |