File size: 6,579 Bytes
6369972 |
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 |
"""
WBS Level 3: Create a Work Breakdown Structure (WBS) from a project plan.
https://en.wikipedia.org/wiki/Work_breakdown_structure
The "progressive elaboration" technique is used to decompose big tasks into smaller, more manageable subtasks.
"""
import os
import json
import time
from dataclasses import dataclass
from math import ceil
from typing import List, Optional
from uuid import uuid4
from dataclasses import dataclass
from pydantic import BaseModel, Field
from llama_index.core.llms.llm import LLM
from src.format_json_for_use_in_query import format_json_for_use_in_query
class WBSSubtask(BaseModel):
"""
A subtask.
"""
name: str = Field(
description="Short name of the subtask, such as: Prepare necessary documentation for permits, Conduct interviews with potential contractors, Secure final approval of funds."
)
description: str = Field(
description="Longer text that describes the subtask in more detail, such as: Objective, Scope, Steps, Deliverables."
)
resources_needed: list[str] = Field(
description="List of resources needed to complete the subtask. Example: ['Project manager', 'Architect', 'Engineer', 'Construction crew']."
)
class WBSTaskDetails(BaseModel):
"""
A big task in the project decomposed into a few smaller tasks.
"""
subtasks: list[WBSSubtask] = Field(
description="List of subtasks."
)
QUERY_PREAMBLE = f"""
Decompose a big task into smaller, more manageable subtasks.
Split the task into 3 to 5 subtasks.
Pick a subtask name that is short, max 7 words or max 60 characters.
Don't enumerate the subtasks with integers or letters.
Don't assign a uuid to the subtask.
"""
@dataclass
class CreateWBSLevel3:
"""
WBS Level 3: Creating a Work Breakdown Structure (WBS) from a project plan.
"""
query: str
response: dict
metadata: dict
tasks: list[dict]
task_uuids: list[str]
@classmethod
def format_query(cls, plan_json: dict, wbs_level1_json: dict, wbs_level2_json: list, wbs_level2_task_durations_json: dict, decompose_task_id: str) -> str:
if not isinstance(plan_json, dict):
raise ValueError("Invalid plan_json.")
if not isinstance(wbs_level1_json, dict):
raise ValueError("Invalid wbs_level1_json.")
if not isinstance(wbs_level2_json, list):
raise ValueError("Invalid wbs_level1_json.")
if not isinstance(wbs_level2_task_durations_json, list):
raise ValueError("Invalid wbs_level2_task_durations_json.")
if not isinstance(decompose_task_id, str):
raise ValueError("Invalid decompose_task_id.")
query = f"""
The project plan:
{format_json_for_use_in_query(plan_json)}
WBS Level 1:
{format_json_for_use_in_query(wbs_level1_json)}
WBS Level 2:
{format_json_for_use_in_query(wbs_level2_json)}
WBS Level 2 time estimates:
{format_json_for_use_in_query(wbs_level2_task_durations_json)}
Only decompose this task:
"{decompose_task_id}"
"""
return query
@classmethod
def execute(cls, llm: LLM, query: str, decompose_task_id: str) -> 'CreateWBSLevel3':
"""
Invoke LLM to decompose a big WBS level 2 task into smaller tasks.
"""
if not isinstance(llm, LLM):
raise ValueError("Invalid LLM instance.")
if not isinstance(query, str):
raise ValueError("Invalid query.")
if not isinstance(decompose_task_id, str):
raise ValueError("Invalid decompose_task_id.")
start_time = time.perf_counter()
sllm = llm.as_structured_llm(WBSTaskDetails)
response = sllm.complete(QUERY_PREAMBLE + query)
json_response = json.loads(response.text)
end_time = time.perf_counter()
duration = int(ceil(end_time - start_time))
metadata = dict(llm.metadata)
metadata["llm_classname"] = llm.class_name()
metadata["duration"] = duration
# Cleanup the json response from the LLM model, assign unique ids to each subtask.
parent_task_id = decompose_task_id
result_tasks = []
result_task_uuids = []
for subtask in json_response['subtasks']:
name = subtask['name']
description = subtask['description']
resources_needed = subtask['resources_needed']
uuid = str(uuid4())
subtask_item = {
"id": uuid,
"name": name,
"description": description,
"resources_needed": resources_needed,
"parent_id": parent_task_id
}
result_tasks.append(subtask_item)
result_task_uuids.append(uuid)
result = CreateWBSLevel3(
query=query,
response=json_response,
metadata=metadata,
tasks=result_tasks,
task_uuids=result_task_uuids
)
return result
def raw_response_dict(self, include_metadata=True, include_query=True) -> dict:
d = self.response.copy()
if include_metadata:
d['metadata'] = self.metadata
if include_query:
d['query'] = self.query
return d
if __name__ == "__main__":
from llama_index.llms.ollama import Ollama
# TODO: Eliminate hardcoded paths
basepath = '/Users/neoneye/Desktop/planexe_data'
def load_json(relative_path: str) -> dict:
path = os.path.join(basepath, relative_path)
print(f"loading file: {path}")
with open(path, 'r', encoding='utf-8') as f:
the_json = json.load(f)
return the_json
plan_json = load_json('002-project_plan.json')
wbs_level1_json = load_json('006-wbs_level1.json')
wbs_level2_json = load_json('008-wbs_level2.json')
wbs_level2_task_durations_json = load_json('012-task_durations.json')
decompose_task_id = "1c690f4a-ae8e-493d-9e47-6da58ef5b24c"
query = CreateWBSLevel3.format_query(plan_json, wbs_level1_json, wbs_level2_json, wbs_level2_task_durations_json, decompose_task_id)
model_name = "llama3.1:latest"
# model_name = "qwen2.5-coder:latest"
# model_name = "phi4:latest"
llm = Ollama(model=model_name, request_timeout=120.0, temperature=0.5, is_function_calling_model=False)
print(f"Query: {query}")
result = CreateWBSLevel3.execute(llm, query, decompose_task_id)
print("\n\nResponse:")
print(json.dumps(result.raw_response_dict(include_query=False), indent=2))
print("\n\nExtracted tasks:")
print(json.dumps(result.tasks, indent=2))
|