Andrew-Chen-Wang

On very large pages, the snapshot can fill the entire context #1329

microsoft/playwright-mcp#1329

believe a good source of inspiration is Claude Code/Codex. There are two things that I'm thinking of which is:

Grab the HTML instead. Unfortunately, usually when the browser_snapshot is so huge, the html is also pretty big (in my case 12m HTML and above sometimes) Take a browser_screenshot and grab the HTML and browser_snapshot and use a "Grep" tool or an HTML parser (or however your agent is running) to whatever you need so that your MCP can do a browser action. Images should be much smaller. Here's some slop Claude code that worked for me very well. I'm using openai/agents, but this can be used for any custom or alternative agent package.

Andrew-Chen-Wang

import { getGlobalTraceProvider, setTracingDisabled, withTrace } from "@openai/agents"
import { Agent, run } from "@openai/agents"
import { openai } from "@scraper/utils/openai"
import { zodTextFormat } from "openai/helpers/zod"
import { afterAll, beforeAll, describe, expect, it } from "vitest"
import { z } from "zod"

const testSchema = z.object({
  randomNumber: z.number(),
})

Andrew-Chen-Wang

Currently, I'm working with Pulumi; here's a list of downsides.

• One thing that I noticed is a problem compared to a static configuration file like Terraform is that the destruction of resources is not in order. For instance, if I wanted to get rid of AWS Route53 records but some already existed, Pulumi would first create rather than destroy. In Route53, a single record type can only be defined once per name/host. This became a problem where I had to manually destroy the records first.
• Support is lacking due to a lack of a community
• Things don't work sometimes, and it's hard to debug (I'm basically facing the same issues as https://aws.plainenglish.io/why-we-abandoned-pulumi-after-3-months-and-went-back-to-terraform-03539d1859c1)

Terraform comparison:

Andrew-Chen-Wang

The repository for the assignment is public and Github does not allow the creation of private forks for public repositories.

The correct way of creating a private frok by duplicating the repo is documented here.

For this assignment the commands are:

1. Create a bare clone of the repository. (This is temporary and will be removed so just do it wherever.)

git clone --bare git@github.com:usi-systems/easytrace.git

Andrew-Chen-Wang

abandoned
able
absolute
adorable
adventurous
academic
acceptable
acclaimed
accomplished
accurate

Andrew-Chen-Wang

import uuid
import random
from datetime import datetime, UTC

def encode_time_fields(dt: datetime) -> int:
    year = dt.year
    month = dt.month
    hour = dt.hour
    minute = dt.minute
    millisecond = dt.microsecond // 1000

Andrew-Chen-Wang

import { SelectQueryBuilder } from "kysely";
import { DB } from "db";

/*
 // From kysely-codegen
 export type Generated<T> = T extends ColumnType<infer S, infer I, infer U>
  ? ColumnType<S, I | undefined, U>
  : ColumnType<T, T | undefined, T>;

 export type Int8 = ColumnType<string, bigint | number | string, bigint | number | string>;

Andrew-Chen-Wang

Fetch All Files from Firebase Hosting

This script fetches all of the files from the currently deployed version of a Firebase Hosting site. You must be signed in via the Firebase CLI and have "Site Viewer" permission on the site in question to be able to properly run the script.

Running via NPX

npx https://gist.github.com/mbleigh/9c8680cf319ace2f506f57380da66e7d <site_name>

Andrew-Chen-Wang

"""
Created by: @Andrew-Chen-Wang
Parallel calling OpenAI with rate limit handling.

Required packages:
- openai
- tenacity (optional if you remove the decorators. Useful for 500 errors)

Usage:
You must call with with_raw_response to get rate limit headers like:

Andrew-Chen-Wang

from bs4 import BeautifulSoup

# remove all attributes
def _remove_all_attrs(soup):
    for tag in soup.find_all(True): 
        tag.attrs = {}
    return soup

# remove all attributes except some tags
def _remove_all_attrs_except(soup):