Role of Python Flashcards

Question

What are the three types of **cell types** in a notebook?

Answer 1

* code * markdown * raw ## Footnote Each cell type serves a different purpose in the notebook.

Answer 2

The “In [3]” number you see in the UI ## Footnote It indicates how many times the cell has been executed.

Answer 3

A list of result objects (print output, rich displays, errors) ## Footnote This allows the notebook to remember what happened the last time it was run.

Answer 4

* Kernel info (language, display name) * Extensions config, authors, etc. ## Footnote Metadata is crucial for tools and pipelines that rely on these fields.

Answer 5

* Frontend * Jupyter server * Kernel ## Footnote Each component plays a vital role in rendering and executing the notebook.

Answer 6

FALSE ## Footnote All code cells communicate with the same kernel instance, leading to a shared global state.

Answer 7

* Frontend sends code to the kernel * Kernel executes code * Outputs are delivered back to frontend * Frontend renders outputs ## Footnote This process updates the cell’s execution count and outputs in memory.

Answer 8

* Executes code * Keeps variables in memory * Serves as the single source of truth for all variables ## Footnote The kernel provides a REPL-like loop for executing and inspecting code.

Answer 9

* Chunked experimentation * Re-runnable sections * Narrative structure ## Footnote Cells allow for organized and efficient coding and documentation.

Answer 10

* A lab notebook * A shareable artifact ## Footnote Outputs allow the notebook to retain and display results from previous runs.

Answer 11

* Parameterization / automation * Selective execution * Presentation changes ## Footnote Metadata enhances the functionality and usability of notebooks.

Answer 12

* Exploratory playground * Narrative analysis / reports * Experiment harness * Bridge to production code ## Footnote These functions illustrate the versatility of notebooks in data science and programming.

Answer 13

[Prompt templates and system prompts] ## Footnote This highlights how notebooks can be utilized for rapid experimentation.

Answer 14

Sequential execution per kernel while allowing user control over the sequence ## Footnote This model gives notebooks their hybrid feel of REPL and document.

Answer 15

**LLM wiring framework** ## Footnote It provides standard building blocks to connect models, prompts, tools, and memory into reusable chains and agents.

Answer 16

* LLMs and Chat Models * Prompts and Templates * Chains * Memory * Tools * Agents * Document loaders, text splitters, and vector stores * Callbacks / tracing / observability ## Footnote These components work together to facilitate the orchestration of LLM workflows.

Answer 17

* Abstractions around model providers * Unified interface for invoking models ## Footnote Examples include OpenAI, Anthropic, and local models like Ollama.

Answer 18

* Separate prompt design from execution code * Reuse templates across different chains or models * Make prompts programmable ## Footnote This allows for conditionals, few-shot examples, and multi-message chat prompts.

Answer 19

A **Chain** is a pipeline that takes inputs, calls components, and returns outputs ## Footnote It encapsulates multistep LLM workflows into reusable objects.

Answer 20

* Store conversation history or other state * Enable stateful chains/agents ## Footnote It helps agents remember previous user turns and recall facts from earlier in a session.

Answer 21

* Allow LLMs to act in the world * Call APIs, run code, query databases ## Footnote Tools are what agents choose between when deciding what to do next.

Answer 22

* Built on one or more LLMs * A set of tools ## Footnote Agents implement tool-augmented reasoning and can be composed into larger systems.

Answer 23

* Pull text from various sources * Break content into chunks ## Footnote They are part of the RAG stack for Retrieval-Augmented Generation.

Answer 24

* Log prompts and responses * Send traces to dashboards * Measure tokens, latency, cost ## Footnote They are crucial for debugging and improving agent behavior.

Answer 25

TRUE ## Footnote This helps avoid a ball of custom scripts and facilitates easier orchestration.

Answer 26

* See user input + current state * Ask the model what to do next * Execute the tool * Feed the result back into the model ## Footnote This process repeats until termination.

Answer 27

* Handles local orchestration for a single app or service * Encapsulates prompt logic, tools, RAG, and agent loops ## Footnote It can be integrated with other services for broader coordination.

Answer 28

* LLM / ChatModel * PromptTemplate * Chain * Memory * Tool * Agent * RAG stack (Loader → Splitter → Embeddings → VectorStore) * Callbacks / Tracing ## Footnote These components work together to create a structured system for interacting with language models.

Answer 29

* Provide a uniform way to talk to different models * Encapsulate provider-specific details * Allow higher-level components to treat model calls as nodes in a graph ## Footnote The LLM acts as a driver around an external model API.

Answer 30

* Static parts: system instructions, role, examples * Slots / variables: `{question}`, `{context}`, `{tools}` * Optional structure: multi-message representation ## Footnote It translates structured inputs into model-ready text/messages.

Answer 31

* Directed graph / pipeline of steps * Input schema (expected keys) * Output schema * May have sub-chains ## Footnote Chains encapsulate multi-step procedures into a single callable unit.

Answer 32

* Inject context into new prompts * Store conversation history, summaries, key entities * Act as a black-box store for context ## Footnote Memory allows chains/agents to access previous interactions.

Answer 33

* Name * Description * Input schema * Callable that implements the action ## Footnote Tools expose non-LLM capabilities to an LLM in a structured way.

Answer 34

* Implement reason–act–observe cycle * Read current state and ask LLM for next action * Execute tool calls and manage state ## Footnote Agents control the flow of operations based on user input and history.

Answer 35

* Document Loaders * Text Splitters * Embeddings * Vector Store ## Footnote The RAG stack is used for knowledge retrieval and semantic search.

Answer 36

* Read from various sources (PDFs, APIs) * Output a standard Document object ## Footnote They normalize heterogeneous sources into a common document abstraction.

Answer 37

* Break documents into retrieval-friendly chunks * Control recall vs precision tradeoffs ## Footnote They ensure chunks are meaningful and fit into context windows.

Answer 38

* Provide observability * Feed data to debugging tools and evaluation systems ## Footnote Callbacks allow monitoring of operations like model calls and tool calls.

Answer 39

* External input → Agent → Memory loading → (Optional RAG call) → Build prompt → Call LLM → Tool execution → Final answer ## Footnote This flow illustrates how components interact to process requests.

Role of Python Flashcards

(63 cards)