Dify Workflow Basics Interview Questions Hindi mein

Dify Workflow Basics कोर्स के 25 सवाल, हर सवाल के सही जवाब और explanation के साथ। जवाब देखने से पहले ख़ुद सोचें।

Getting to Know Dify

1. Which is part of Dify's feature set?

  1. A replacement for a SQL database
  2. Training foundation models from scratch
  3. Visual workflows, knowledge bases, model provider connections and an API for each app
  4. A mobile operating system
जवाब देखें

जवाब: Visual workflows, knowledge bases, model provider connections and an API for each app

Dify orchestrates models; it does not pretrain them.

पाठ: What Dify Is

2. Which app type suits "classify each incoming support ticket and return a category"?

  1. Workflow
  2. Chatflow with long memory
  3. Text-to-speech app
  4. No app is needed
जवाब देखें

जवाब: Workflow

One input, one structured output.

पाठ: App Types: Chatbot, Agent, Chatflow and Workflow

3. What is a main reason to self-host Dify?

  1. It removes the need for API keys
  2. It requires no maintenance
  3. It makes models free
  4. Keeping data and infrastructure under your own control
जवाब देखें

जवाब: Keeping data and infrastructure under your own control

Control comes with operational work.

पाठ: Setting Up: Cloud or Self-Hosted, and Model Providers

Workflow Building Blocks

4. Where are a workflow's input variables declared?

  1. In the end node
  2. In the start node
  3. In the model provider settings
  4. In the browser console
जवाब देखें

जवाब: In the start node

The start node defines what a run receives.

पाठ: Start Node and Variables

5. What happens if a prompt references a variable that has no value?

  1. The model always refuses
  2. It can render as empty text, quietly changing the prompt
  3. The variable is invented automatically
  4. Dify trains a new model
जवाब देखें

जवाब: It can render as empty text, quietly changing the prompt

Verify rendered prompts.

पाठ: LLM Nodes and Prompt Templates

6. Why keep the same output names on every branch?

  1. It reduces token usage to zero
  2. It speeds up the model
  3. Dify forbids different names
  4. So callers can rely on a consistent response shape
जवाब देखें

जवाब: So callers can rely on a consistent response shape

Predictable outputs simplify integration.

पाठ: End Node and Outputs

Knowledge Bases and Retrieval

7. What does chunk overlap help with?

  1. Reducing the chance that a fact is split across two chunks
  2. Making the index smaller
  3. Removing the need for embeddings
  4. Speeding up the LLM
जवाब देखें

जवाब: Reducing the chance that a fact is split across two chunks

Overlap keeps boundary context.

पाठ: Creating a Knowledge Base and Chunking

8. Why can keyword search return the wrong chunk for "how many days for a refund"?

  1. Refund chunks cannot be indexed
  2. Keyword search always returns nothing
  3. It matches shared words like "days" without understanding meaning
  4. Top K must be zero
जवाब देखें

जवाब: It matches shared words like "days" without understanding meaning

Hybrid search and reranking help.

पाठ: Retrieval Settings: Search Mode, Top K and Threshold

9. What should a grounded RAG app do when the context lacks the answer?

  1. Return the whole knowledge base
  2. Invent a plausible answer
  3. Say it could not find the answer
  4. Crash
जवाब देखें

जवाब: Say it could not find the answer

Honest refusals build trust.

पाठ: Grounded Answers From Retrieved Context

Routing, Classification and Iteration

10. Why order if/else conditions from specific to general?

  1. The first matching branch wins, so general conditions would catch specific cases
  2. Dify sorts them alphabetically anyway
  3. Order never matters
  4. Only one condition is allowed
जवाब देखें

जवाब: The first matching branch wins, so general conditions would catch specific cases

Specific rules first.

पाठ: If/Else Branches

11. How should you check a question classifier node?

  1. Check only one example
  2. Trust it without testing
  3. Count the number of classes
  4. Measure routing accuracy on labelled real messages
जवाब देखें

जवाब: Measure routing accuracy on labelled real messages

Routing errors cascade downstream.

पाठ: Question Classifier Node

12. What does an iteration node output?

  1. A single random item
  2. An array with one result per input item
  3. Nothing
  4. Only the first item's result
जवाब देखें

जवाब: An array with one result per input item

Map over items, then collect.

पाठ: Iteration Over Lists

13. What problem does the variable aggregator solve?

  1. It trains the classifier
  2. It lets downstream nodes use one variable regardless of which branch ran
  3. It deletes branches
  4. It uploads documents
जवाब देखें

जवाब: It lets downstream nodes use one variable regardless of which branch ran

Merge results from alternative paths.

पाठ: Variable Aggregator and Merging Branches

Code, Extraction and External Calls

14. What does a Dify code node's main function return?

  1. Nothing; it only prints
  2. A web page
  3. A trained model
  4. A dictionary whose keys become output variables
जवाब देखें

जवाब: A dictionary whose keys become output variables

Returned keys feed downstream nodes.

पाठ: Code Nodes

15. Why validate parameter extractor output?

  1. The model can omit fields or return wrong types
  2. Extraction is always perfect
  3. Validation trains the model
  4. It reduces the number of nodes
जवाब देखें

जवाब: The model can omit fields or return wrong types

Trust, but check.

पाठ: Parameter Extraction

16. Where should API credentials for an HTTP node live?

  1. In the end node output
  2. Inside the system prompt
  3. In secure environment or secret settings, not in prompts
  4. In the user's message
जवाब देखें

जवाब: In secure environment or secret settings, not in prompts

Keep secrets out of model-visible text.

पाठ: HTTP Requests, Tools and Plugins

Chatflows, Memory and Agents

17. What do conversation variables provide in a chatflow?

  1. Document chunking
  2. Faster models
  3. State that persists across turns
  4. API keys
जवाब देखें

जवाब: State that persists across turns

Remember important facts explicitly.

पाठ: Chatflows Versus Workflows

18. What is a cost of a very large memory window?

  1. More tokens per call and more chance of distraction by stale details
  2. The model forgets everything
  3. It disables retrieval
  4. It removes the end node
जवाब देखें

जवाब: More tokens per call and more chance of distraction by stale details

More context is not always better.

पाठ: Memory and Context Length

19. When is a fixed workflow better than an agent?

  1. Never
  2. When the task is completely open-ended
  3. When you want less predictability
  4. When the steps are known and always the same
जवाब देखें

जवाब: When the steps are known and always the same

Predictability is valuable.

पाठ: Agent Nodes and Tool Use

Publishing and Calling Apps

20. Where should the Dify app API key be used?

  1. In a public JavaScript bundle
  2. In your backend server, never in browser code
  3. In the prompt text
  4. In the end node output
जवाब देखें

जवाब: In your backend server, never in browser code

Keys in browsers can be stolen.

पाठ: Calling a Workflow Through the API

21. When is a streamed answer complete?

  1. After the first text chunk
  2. When the workflow-finished event arrives
  3. Immediately after the request is sent
  4. When the browser tab closes
जवाब देखें

जवाब: When the workflow-finished event arrives

Wait for the completion event.

पाठ: Streaming Responses

22. What is the DSL export useful for?

  1. Replacing the knowledge base
  2. Training models
  3. Storing user passwords
  4. Backups, moving apps between environments and reviewing changes
जवाब देखें

जवाब: Backups, moving apps between environments and reviewing changes

Treat app definitions like code.

पाठ: Publishing, Web Apps and Versions

Debugging, Cost and Production Readiness

23. An answer is wrong; what should you check early?

  1. Whether routing and retrieval behaved as expected
  2. Only the font of the web app
  3. The number of API keys
  4. Nothing; republish
जवाब देखें

जवाब: Whether routing and retrieval behaved as expected

Most failures start upstream of the LLM.

पाठ: Logs, Tracing and Debugging

24. Which change usually saves the most in the example?

  1. Renaming variables
  2. Removing the free retrieval node
  3. Reducing the large answer model's input tokens
  4. Changing the app icon
जवाब देखें

जवाब: Reducing the large answer model's input tokens

Target the dominant cost.

पाठ: Cost per Run

25. Which item belongs on a Dify launch checklist?

  1. Arithmetic done by the LLM
  2. API key embedded in the web page
  3. Routing accuracy measured on labelled messages
  4. No test inputs
जवाब देखें

जवाब: Routing accuracy measured on labelled messages

Measure, secure, version.

पाठ: A Production Readiness Checklist

← कोर्स पर वापस जाएँ