Planning an AI Project

Welcome to the finale of Python Safari! Just as every great expedition requires planning, every programming project needs solid foundations. In this module, you will combine all the skills you've acquired into one comprehensive project!

Project Planning Methodology

1. Defining the Problem

1from dataclasses import dataclass
2from enum import Enum
3
4class ProjectType(Enum):
5    WEB_APP = "web_application"
6    API = "api_service"
7    ML_PIPELINE = "ml_pipeline"
8    RAG_SYSTEM = "rag_system"
9    AUTOMATION = "automation"
10
11@dataclass
12class ProjectDefinition:
13    """Project definition."""
14    name: str
15    problem_statement: str
16    target_users: list[str]
17    success_metrics: list[str]
18    project_type: ProjectType
19    tech_stack: list[str]
20
21# Example
22capstone_project = ProjectDefinition(
23    name="AI Document Assistant",
24    problem_statement="Users waste time searching for information in documents",
25    target_users=["analysts", "lawyers", "researchers"],
26    success_metrics=[
27        "70% reduction in search time",
28        "Answer accuracy > 90%",
29        "Response time < 2s"
30    ],
31    project_type=ProjectType.RAG_SYSTEM,
32    tech_stack=["Python", "FastAPI", "LlamaIndex", "Qdrant", "React"]
33)

2. User Stories

1@dataclass
2class UserStory:
3    """User story in Agile format."""
4    as_a: str
5    i_want: str
6    so_that: str
7    acceptance_criteria: list[str]
8    priority: int  # 1-5, 1 = highest
9
10stories = [
11    UserStory(
12        as_a="analyst",
13        i_want="ask questions about documents in natural language",
14        so_that="I can quickly find the information I need",
15        acceptance_criteria=[
16            "System understands questions in English",
17            "Response includes sources",
18            "Response time < 3s"
19        ],
20        priority=1
21    ),
22    UserStory(
23        as_a="administrator",
24        i_want="easily add new documents",
25        so_that="the knowledge base stays up to date",
26        acceptance_criteria=[
27            "Upload via drag & drop",
28            "Support for PDF, DOCX, TXT",
29            "Automatic indexing"
30        ],
31        priority=2
32    )
33]

3. Estimation and Sprint Planning

1from datetime import datetime, timedelta
2
3@dataclass
4class Task:
5    name: str
6    story_points: int  # Fibonacci: 1, 2, 3, 5, 8, 13
7    dependencies: list[str]
8    assigned_to: str = ""
9
10@dataclass
11class Sprint:
12    number: int
13    start_date: datetime
14    tasks: list[Task]
15    velocity: int = 20  # Story points per sprint
16
17    @property
18    def end_date(self) -> datetime:
19        return self.start_date + timedelta(days=14)
20
21    @property
22    def total_points(self) -> int:
23        return sum(t.story_points for t in self.tasks)
24
25# Project plan
26sprints = [
27    Sprint(1, datetime(2024, 1, 1), [
28        Task("Project setup", 2, []),
29        Task("CI/CD configuration", 3, ["Project setup"]),
30        Task("Basic API", 5, ["Project setup"]),
31        Task("Vector DB integration", 5, ["Basic API"]),
32    ]),
33    Sprint(2, datetime(2024, 1, 15), [
34        Task("RAG Pipeline", 8, ["Vector DB integration"]),
35        Task("Chat interface", 5, ["Basic API"]),
36        Task("Document upload", 5, ["RAG Pipeline"]),
37    ])
38]

Choosing the Technology Stack

1tech_decisions = {
2    "backend": {
3        "choice": "FastAPI",
4        "reasons": ["Async", "Type hints", "Auto docs", "Performance"]
5    },
6    "vector_db": {
7        "choice": "Qdrant",
8        "reasons": ["Self-hosted", "Filtering", "Rust performance"]
9    },
10    "llm": {
11        "choice": "OpenAI GPT-4",
12        "reasons": ["Quality", "Function calling", "Reliable"]
13    },
14    "rag_framework": {
15        "choice": "LlamaIndex",
16        "reasons": ["Flexibility", "Integrations", "Community"]
17    }
18}

Remember: a good plan is half the success. In the next lesson, we will design the system architecture!