File size: 9,919 Bytes
2004c79
3b9a6b5
 
 
 
 
 
 
 
 
 
 
 
 
2004c79
3b9a6b5
 
 
 
 
 
918bdb4
3b9a6b5
918bdb4
 
 
3b9a6b5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2004c79
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3b9a6b5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2004c79
 
3b9a6b5
2004c79
3b9a6b5
 
 
2004c79
3b9a6b5
 
 
 
 
 
 
 
2004c79
3b9a6b5
2004c79
3b9a6b5
 
 
 
 
2004c79
 
 
3b9a6b5
 
 
 
2004c79
 
 
 
 
3b9a6b5
2004c79
3b9a6b5
 
 
2004c79
 
3b9a6b5
 
2004c79
3b9a6b5
 
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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
import asyncio
from typing import Optional, List

from llama_index.llms.nebius import NebiusLLM
from llama_index.core.prompts import RichPromptTemplate
from llama_index.core.workflow import (
    StartEvent,
    StopEvent,
    Workflow,
    step,
    Event,
)

from utils.markdown_analyzer import MarkdownAnalyzer
from factory.agents.task_processing import (
    remove_markdown_code_blocks,
    remove_markdown_list_elements,
    unwrap_tasks_from_generated,
    log_task_duration_breakdown,
    log_total_time,
)
from utils.logging_config import setup_logging, get_logger

# Initialize logging
setup_logging()
logger = get_logger(__name__)


from domain import AgentsConfig, AGENTS_CONFIG


class TaskComposerAgent:
    def __init__(self, config: AgentsConfig = AGENTS_CONFIG):
        self.config = config
        self.llm: Optional[NebiusLLM] = None
        self.task_splitter_template: Optional[RichPromptTemplate] = None
        self.task_evaluator_template: Optional[RichPromptTemplate] = None
        self.task_deps_matcher_template: Optional[RichPromptTemplate] = None
        self.workflow: Optional[TaskComposerWorkflow] = None

        self.set_llm()
        self.set_prompt_templates()
        self.set_workflow()

    def set_llm(self) -> None:
        self.llm = NebiusLLM(
            model=self.config.nebius_model,
            api_key=self.config.nebius_api_key,
            timeout=self.config.timeout,
            max_retries=self.config.max_retries,
            verify_ssl=self.config.verify_ssl,
            request_timeout=self.config.request_timeout,
            max_tokens=self.config.max_tokens,
            temperature=self.config.temperature,
        )

    def set_prompt_templates(self) -> None:
        self.task_splitter_template = RichPromptTemplate(
            self.config.task_splitter_prompt,
            template_var_mappings={"query_str": "query"},
        )
        self.task_evaluator_template = RichPromptTemplate(
            self.config.task_evaluator_prompt,
            template_var_mappings={"query_str": "query"},
        )
        self.task_deps_matcher_template = RichPromptTemplate(
            self.config.task_deps_matcher_prompt,
            template_var_mappings={
                "query_str": "task",
                "skills_str": "skills",
                "context_str": "context",
            },
        )

    def set_workflow(self) -> None:
        self.workflow = TaskComposerWorkflow(
            llm=self.llm,
            task_splitter_template=self.task_splitter_template,
            task_evaluator_template=self.task_evaluator_template,
            task_deps_matcher_template=self.task_deps_matcher_template,
            timeout=self.config.workflow_timeout,
            verbose=True,
        )

    async def run_workflow(
        self, query: str, skills: Optional[List[str]] = None, context: str = ""
    ) -> str:
        return await self.workflow.run(
            input=query, skills=skills or [], context=context
        )

    async def compose_tasks(self, input_text: str, parameters) -> List:
        """
        Compose tasks from input text using the task composer workflow.

        Args:
            input_text: The input text to compose tasks from
            parameters: TimeTableDataParameters containing skill information

        Returns:
            List of task tuples (description, duration, skill)
        """
        try:
            # Extract skills from parameters
            skills = list(parameters.skill_set.required_skills) + list(
                parameters.skill_set.optional_skills
            )

            # Run the workflow
            result = await self.run_workflow(input_text, skills=skills, context="")

            # The workflow returns a list of tuples (description, duration, skill)
            logger.debug(f"Task composer workflow result: {result}")
            return result

        except Exception as e:
            logger.error(f"Error in compose_tasks: {e}")
            return []


class TaskSplitter(Event):
    task_splitter_output: str
    skills: List[str]
    context: str


class TaskEvaluator(Event):
    task_evaluator_output: list[tuple[str, str]]
    skills: List[str]
    context: str


class TaskDependencyMatcher(Event):
    task_dependency_output: list[
        tuple[str, str, str]
    ]  # (task, duration, matched_skill)


class TaskComposerWorkflow(Workflow):
    def __init__(
        self,
        llm: NebiusLLM,
        task_splitter_template: RichPromptTemplate,
        task_evaluator_template: RichPromptTemplate,
        task_deps_matcher_template: RichPromptTemplate,
        **kwargs,
    ):
        super().__init__(**kwargs)
        self._llm = llm
        self._task_splitter_template = task_splitter_template
        self._task_evaluator_template = task_evaluator_template
        self._task_deps_matcher_template = task_deps_matcher_template

    @step
    async def split_tasks(self, event: StartEvent) -> TaskSplitter:
        logger.info("=== Step 1: Task Breakdown ===")
        logger.info(f"Input task: {event.input}")

        formatted_prompt: str = self._task_splitter_template.format(query=event.input)

        response = await asyncio.wait_for(
            asyncio.to_thread(self._llm.complete, formatted_prompt), timeout=30.0
        )

        logger.info("Task breakdown:")
        logger.info(response.text)

        # Get skills and context from the event, default to empty if not provided
        skills = getattr(event, "skills", [])
        context = getattr(event, "context", "")

        logger.info(f"Received skills: {skills}")
        logger.info(f"Received context: {context}")

        return TaskSplitter(
            task_splitter_output=response.text, skills=skills, context=context
        )

    @step
    async def evaluate_tasks_duration(self, event: TaskSplitter) -> TaskEvaluator:
        logger.info("=== Step 2: Time Estimation ===")
        logger.info("Using task breakdown from Step 1:")
        logger.info(event.task_splitter_output)

        content: str = remove_markdown_code_blocks(event.task_splitter_output)
        analyzer: MarkdownAnalyzer = MarkdownAnalyzer(content)
        result: list = analyzer.identify_lists()["Unordered list"]
        tasks: list[str] = unwrap_tasks_from_generated(result)

        logger.info(f"Processing {len(tasks)} tasks for time estimation...")

        merged_tasks: list[tuple[str, str]] = []
        for i, task in enumerate(tasks, 1):
            try:
                formatted_prompt: str = self._task_evaluator_template.format(query=task)

                response = await asyncio.wait_for(
                    asyncio.to_thread(self._llm.complete, formatted_prompt),
                    timeout=30.0,
                )
                merged_tasks.append((task, response.text))
                logger.info(f"Completed time estimation {i}/{len(tasks)}")

            except asyncio.TimeoutError:
                logger.warning(f"Time estimation timeout for task {i}: {task[:50]}...")

                # Use default duration of 2 units (1 hour)
                merged_tasks.append((task, "2"))

            except Exception as e:
                logger.error(f"Error estimating time for task {i}: {e}")

                # Use default duration of 2 units (1 hour)
                merged_tasks.append((task, "2"))

        # remove markdown list elements wrapped in **
        merged_tasks = remove_markdown_list_elements(merged_tasks)
        log_task_duration_breakdown(merged_tasks)
        log_total_time(merged_tasks)

        return TaskEvaluator(
            task_evaluator_output=merged_tasks,
            skills=event.skills,
            context=event.context,
        )

    @step
    async def evaluate_tasks_dependencies(
        self, event: TaskEvaluator
    ) -> TaskDependencyMatcher:
        logger.info("=== Step 3: Skill Matching ===")
        logger.info("Matching tasks with skills...")

        final_tasks: list[tuple[str, str, str]] = []
        for i, (task, duration) in enumerate(event.task_evaluator_output, 1):
            try:
                formatted_prompt: str = self._task_deps_matcher_template.format(
                    task=task, skills=", ".join(event.skills), context=event.context
                )

                response = await asyncio.wait_for(
                    asyncio.to_thread(self._llm.complete, formatted_prompt),
                    timeout=30.0,
                )

                matched_skill = response.text.strip()
                final_tasks.append((task, duration, matched_skill))
                logger.info(
                    f"Completed skill matching {i}/{len(event.task_evaluator_output)}"
                )

            except asyncio.TimeoutError:
                logger.warning(f"Skill matching timeout for task {i}: {task[:50]}...")

                # Use a default skill
                default_skill = event.skills[0] if event.skills else "General"
                final_tasks.append((task, duration, default_skill))

            except Exception as e:
                logger.error(f"Error matching skill for task {i}: {e}")

                # Use a default skill
                default_skill = event.skills[0] if event.skills else "General"
                final_tasks.append((task, duration, default_skill))

        logger.info(f"Skill matching completed for {len(final_tasks)} tasks")

        return TaskDependencyMatcher(task_dependency_output=final_tasks)

    @step
    async def result_output(self, event: TaskDependencyMatcher) -> StopEvent:
        logger.info("=== Final Result ===")
        logger.info(f"Generated {len(event.task_dependency_output)} tasks with skills")

        for task, duration, skill in event.task_dependency_output:
            logger.info(f"- {task[:50]}... | Duration: {duration} | Skill: {skill}")

        return StopEvent(result=event.task_dependency_output)