Streaming responses in FastAPI - Performance & Optimization

Performance: Streaming responses

MEDIUM IMPACT

Streaming responses affect how quickly the user starts receiving data and how smoothly the content loads over time.

Sending large data or long processing results to the client

FastAPI

from fastapi import FastAPI
from fastapi.responses import StreamingResponse
import asyncio

app = FastAPI()

async def data_generator():
    for i in range(10):
        yield f'data chunk {i}\n'
        await asyncio.sleep(0.1)

@app.get('/stream')
async def stream_data():
    return StreamingResponse(data_generator(), media_type='text/plain')

Sends data chunks as soon as they are ready, allowing the browser to start rendering early.

📈 Performance GainReduces LCP by delivering first bytes immediately, improving user experience.

Sending large data or long processing results to the client

FastAPI

from fastapi import FastAPI
app = FastAPI()

@app.get('/data')
async def get_data():
    result = compute_large_data()
    return result

The server waits to finish all processing before sending any data, delaying the first byte to the client.

📉 Performance CostBlocks rendering until full data is ready, increasing LCP by hundreds of milliseconds or more.

Performance Comparison

Pattern	DOM Operations	Reflows	Paint Cost	Verdict
Full response after processing	Single DOM update after full data	Single reflow after full load	One large paint	[X] Bad
Streaming response with chunks	Incremental DOM updates as chunks arrive	Multiple small reflows	Multiple small paints	[OK] Good

Rendering Pipeline

Streaming responses send partial data progressively, allowing the browser to start parsing and rendering before the full response arrives.

→Network

→HTML Parsing

→Rendering

⚠️ BottleneckNetwork latency and server processing before first chunk

Core Web Vital Affected

LCP

Streaming responses affect how quickly the user starts receiving data and how smoothly the content loads over time.

Optimization Tips

1Use StreamingResponse to send data chunks as soon as they are ready.

2Avoid waiting for full data before responding to reduce LCP.

3Test streaming endpoints in DevTools Network tab to verify chunked delivery.

Performance Quiz - 3 Questions

How do streaming responses affect Largest Contentful Paint (LCP)?

A It improves LCP by sending data chunks early so content appears faster.

B It worsens LCP because it delays sending the full content.

C It has no effect on LCP since content is the same.

D It only affects interaction responsiveness, not LCP.

DevTools: Network

How to check: Open DevTools, go to Network tab, reload the streaming endpoint, and observe the timing waterfall for incremental data chunks.

What to look for: Look for early 'First Byte' and multiple small data packets arriving over time indicating streaming.

Practice

1. What is the main benefit of using StreamingResponse in FastAPI?

easy

A. It compresses the entire response into a single file.

B. It automatically caches all responses for faster access.

C. It converts JSON data into XML format before sending.

D. It sends data in small parts instead of all at once, improving efficiency.

Solution

1. Step 1: Understand what streaming means

   Streaming sends data in chunks, not all at once, which helps with large data or slow connections.

2. Step 2: Identify the role of StreamingResponse

   FastAPI's StreamingResponse helps send these chunks efficiently to the client.

3. Final Answer:

   It sends data in small parts instead of all at once, improving efficiency. -> Option D

4. Quick Check:

   Streaming = sending data in parts [OK]

Hint: Streaming sends data bit by bit, not all at once [OK]

Common Mistakes:

• Thinking StreamingResponse caches data
• Confusing streaming with data compression
• Assuming it changes data format automatically

2. Which of the following is the correct way to import StreamingResponse in FastAPI?

easy

A. from fastapi import StreamingResponse

B. import StreamingResponse from fastapi.responses

C. from fastapi.responses import StreamingResponse

D. import StreamingResponse as sr from fastapi

Solution

1. Step 1: Recall FastAPI import syntax

   FastAPI uses standard Python import syntax: from module import class/function.

2. Step 2: Identify correct module for StreamingResponse

   StreamingResponse is in fastapi.responses, so correct import is from fastapi.responses import StreamingResponse.

3. Final Answer:

   from fastapi.responses import StreamingResponse -> Option C

4. Quick Check:

   Correct import = from fastapi.responses import StreamingResponse [OK]

Hint: Use 'from fastapi.responses import StreamingResponse' [OK]

Common Mistakes:

• Using incorrect import syntax like 'import StreamingResponse from ...'
• Importing directly from 'fastapi' instead of 'fastapi.responses'
• Trying to alias import incorrectly

3. Given this FastAPI endpoint code, what will the client receive?

from fastapi import FastAPI
from fastapi.responses import StreamingResponse

app = FastAPI()

def generate_numbers():
    for i in range(3):
        yield f"Number: {i}\n"

@app.get("/numbers")
def numbers():
    return StreamingResponse(generate_numbers(), media_type="text/plain")

medium

A. A streamed response sending 'Number: 0\n', then 'Number: 1\n', then 'Number: 2\n' in parts

B. A single response with 'Number: 0\nNumber: 1\nNumber: 2\n' all at once

C. Three separate responses, each with one number

D. An error because StreamingResponse cannot use generators

Solution

1. Step 1: Understand the generator function

   The function generate_numbers yields strings one by one for i in 0 to 2.

2. Step 2: Analyze StreamingResponse behavior

   StreamingResponse sends data chunks as they are yielded, so client receives parts streamed.

3. Final Answer:

   A streamed response sending 'Number: 0\n', then 'Number: 1\n', then 'Number: 2\n' in parts -> Option A

4. Quick Check:

   Generator + StreamingResponse = streamed chunks [OK]

Hint: Generators yield chunks; StreamingResponse streams them [OK]

Common Mistakes:

• Thinking all data sends at once
• Expecting multiple separate HTTP responses
• Assuming StreamingResponse rejects generators

4. Identify the error in this FastAPI streaming endpoint code:

from fastapi import FastAPI
from fastapi.responses import StreamingResponse

app = FastAPI()

def stream_data():
    data = ["a", "b", "c"]
    for d in data:
        return d

@app.get("/stream")
def stream():
    return StreamingResponse(stream_data(), media_type="text/plain")

medium

A. The function stream_data uses return inside the loop instead of yield.

B. StreamingResponse requires media_type to be 'application/json'.

C. The endpoint path should be '/stream_data' not '/stream'.

D. FastAPI does not support StreamingResponse with string data.

Solution

1. Step 1: Check stream_data function logic

   Inside the loop, return exits the function immediately, so only first item 'a' is returned.

2. Step 2: Understand streaming requires yield

   StreamingResponse expects an iterable or generator; yield produces chunks one by one.

3. Final Answer:

   The function stream_data uses return inside the loop instead of yield. -> Option A

4. Quick Check:

   Use yield, not return, for streaming generators [OK]

Hint: Use yield inside loops for streaming, not return [OK]

Common Mistakes:

• Using return inside generator loops
• Wrong media_type for StreamingResponse
• Confusing endpoint path naming
• Believing StreamingResponse can't stream strings

5. You want to stream a large CSV file line by line using FastAPI. Which approach correctly combines streaming with file reading?

hard

A. Use StreamingResponse with a function that returns the file size.

B. Use a generator that opens the file and yields each line, then return StreamingResponse with that generator.

C. Return StreamingResponse with the file path string directly.

D. Read the entire file into memory, then return StreamingResponse with a list of lines.

Solution

1. Step 1: Understand streaming large files

   Reading line by line with a generator avoids loading whole file into memory.

2. Step 2: Match StreamingResponse usage

   StreamingResponse accepts an iterable like a generator that yields file lines.

3. Step 3: Evaluate options

   Use a generator that opens the file and yields each line, then return StreamingResponse with that generator. This uses a generator to yield lines, which is efficient and correct. Others either load all data or misuse StreamingResponse.

4. Final Answer:

   Use a generator that opens the file and yields each line, then return StreamingResponse with that generator. -> Option B

5. Quick Check:

   Generator + StreamingResponse = efficient file streaming [OK]

Hint: Stream files by yielding lines in a generator [OK]

Common Mistakes:

• Loading entire file into memory before streaming
• Passing file path string instead of iterable
• Returning file size instead of content