Skip to content

⚡ Bolt: Reuse Gemini client to save ~400ms initialization overhead #25

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
suvadityamuk wants to merge 1 commit into
base: main
Choose a base branch
from
Open

⚡ Bolt: Reuse Gemini client to save ~400ms initialization overhead #25

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions .jules/bolt.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,7 @@
## 2024-05-21 - [Blender Modal Redraws]
**Learning:** Blender `modal` operators run frequently (e.g. every 0.3s). Calling `area.tag_redraw()` unconditionally in the loop forces constant viewport rendering, even when no state has changed, causing unnecessary CPU/GPU load.
**Action:** Only call `tag_redraw()` when the operator actually processes data or updates the UI state (e.g., via a `refresh_needed` flag).

## 2026-02-16 - [API Client Overhead]
Copy link

@gemini-code-assist gemini-code-assist bot Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The date 2026-02-16 appears to be a typo, as it is a future date. It should likely be a date in 2024.

Suggested change
## 2026-02-16 - [API Client Overhead]
## 2024-02-16 - [API Client Overhead]

**Learning:** `google.genai.Client` initialization has a measurable overhead (~75-100ms). Reusing the client instance across sequential and parallel requests saves ~400ms in a multi-step generation pipeline.
**Action:** Always instantiate API clients once at the beginning of a workflow and pass them down, rather than creating them inside helper functions.
15 changes: 10 additions & 5 deletions operators.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,10 +90,13 @@ def execute(self, context):

def _run_pipeline(self, gemini_key, meshy_key, prompt, q):
try:
# Create Gemini client once to save initialization time (~75ms per call)
Copy link
Contributor

@github-actions github-actions bot Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 It's great that you're reusing the client to improve performance. However, you are passing the client object to a nested function generate_view which is then submitted to a ThreadPoolExecutor. This is not incorrect, but it can be slightly cleaner to pass the client to the _run_pipeline method instead of creating it inside. This would make the code more testable and the dependency more explicit.

Suggested change
# Create Gemini client once to save initialization time (~75ms per call)
def _run_pipeline(self, client, gemini_key, meshy_key, prompt, q):
try:
q.put(("INFO", "Refining prompt...", ""))
# Step 1: Refine prompt
refined = utils.refine_prompt(client, prompt)

client = utils.get_client(gemini_key)

q.put(("INFO", "Refining prompt...", ""))

# Step 1: Refine prompt
refined = utils.refine_prompt(gemini_key, prompt)
refined = utils.refine_prompt(client, prompt)
q.put(("REFINED", refined, ""))
q.put(("INFO", "Prompt refined", ""))

Expand All @@ -103,19 +106,19 @@ def _run_pipeline(self, gemini_key, meshy_key, prompt, q):
front_prompt = f"Front view of {refined}, white background, product shot"

# Helper to generate a single view
def generate_view(view_name, view_prompt, input_ref=None):
def generate_view(client, view_name, view_prompt, input_ref=None):
q.put(("INFO", f"Generating {view_name}...", ""))
with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tf:
out = tf.name

# If it's the front view call, input_ref is None usually,
# but valid for subsequent calls
res_path = utils.generate_image(gemini_key, view_prompt, out, input_ref)
res_path = utils.generate_image(client, view_prompt, out, input_ref)
q.put(("IMAGE", f"{view_name} done", res_path))
return res_path

# Generate front view first
front_path = generate_view("Front", front_prompt, None)
front_path = generate_view(client, "Front", front_prompt, None)

# Step 3: Generate Other Views (Parallel)
remaining_views = [
Expand Down Expand Up @@ -145,7 +148,9 @@ def generate_view(view_name, view_prompt, input_ref=None):
futures = []
for v_name, v_prompt in remaining_views:
futures.append(
executor.submit(generate_view, v_name, v_prompt, front_path)
executor.submit(
generate_view, client, v_name, v_prompt, front_path
)
)

# Gather results in order
Expand Down
10 changes: 5 additions & 5 deletions utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,9 +16,9 @@ def get_client(api_key):
return genai.Client(api_key=api_key)


def refine_prompt(api_key, prompt):
def refine_prompt(client, prompt):
"""Use Gemini to refine a prompt for 3D model generation."""
client = get_client(api_key)
# client is passed directly

Copy link
Contributor

@github-actions github-actions bot Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 The comment # client is passed directly is redundant because the function's signature and the removal of get_client(api_key) already make it clear that the client is now an argument.

Suggested change
def refine_prompt(client, prompt):
"""Use Gemini to refine a prompt for 3D model generation."""
instruction = (
f"Refine this prompt for generating a high-quality 3D model reference image. "

instruction = (
f"Refine this prompt for generating a high-quality 3D model reference image. "
Expand All @@ -36,15 +36,15 @@ def refine_prompt(api_key, prompt):
return response.text.strip()


def generate_image(api_key, prompt, output_path, input_image_path=None):
def generate_image(client, prompt, output_path, input_image_path=None):
"""
Generate an image using Gemini.
If input_image_path is provided, use it as reference for the generation.
"""
from google.genai import types
from PIL import Image

client = get_client(api_key)
# client is passed directly
Copy link
Contributor

@github-actions github-actions bot Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 Similar to the comment above, this comment is redundant and can be removed for cleaner code.

Suggested change
# client is passed directly
def generate_image(client, prompt, output_path, input_image_path=None):
"""
Generate an image using Gemini.
If input_image_path is provided, use it as reference for the generation.
"""
from google.genai import types
from PIL import Image
config = types.GenerateContentConfig(
response_modalities=["Image"],


config = types.GenerateContentConfig(
response_modalities=["Image"],
Expand Down Expand Up @@ -117,7 +117,7 @@ def generate_3d_meshy(api_key, image_paths):
task_id = resp.json()["result"]

# Adaptive polling: Check frequently at first (2s), then back off to 5s
# This reduces waiting time for fast jobs without spamming the API for slow ones.
# This reduces waiting time for fast jobs without spamming the API.
Copy link
Contributor

@github-actions github-actions bot Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 This comment change is good, but it could be slightly more descriptive.

Suggested change
# This reduces waiting time for fast jobs without spamming the API.
# This reduces waiting time for fast jobs without spamming the API for slow ones.

intervals = [2, 2, 2, 5]
default_interval = 5

Expand Down
Loading