Search for anything using Google, DuckDuckGo, Phind.com, access AI models, transcribe YouTube videos, generate temporary emails and phone numbers, utilize text-to-speech, leverage WebAI (terminal GPT and open interpreter), and explore offline LLMs, and much more!
- Comprehensive Search: Leverage Google, DuckDuckGo for diverse search results.
- AI Powerhouse: Access and interact with various AI models, including OpenAI, Cohere, and more.
- YouTube Toolkit: Advanced YouTube video and transcript management with multi-language support, versatile downloading, and intelligent data extraction
- Tempmail & Temp Number: Generate temporary email addresses and phone numbers for enhanced privacy.
- Text-to-Speech (TTS): Convert text into natural-sounding speech using multiple AI-powered providers like ElevenLabs, StreamElements, and Voicepods.
- Offline LLMs: Utilize powerful language models offline with GGUF support.
- Extensive Provider Ecosystem: Explore a vast collection of providers, including BasedGPT, DeepSeek, and many others.
- Local LLM Execution: Run GGUF models locally with minimal configuration.
- Rawdog Scripting: Execute Python scripts directly within your terminal using the
rawdogfeature. - GGUF Conversion & Quantization: Convert and quantize Hugging Face models to GGUF format.
- Autollama: Download Hugging Face models and automatically convert them for Ollama compatibility.
- SwiftCLI: A powerful and elegant CLI framework that makes it easy to create beautiful command-line interfaces.
- LitPrinter: Provides beautiful, styled console output with rich formatting and colors
- LitLogger: Simplifies logging with customizable formats and color schemes
- LitAgent: Powerful and modern user agent generator that keeps your requests fresh and undetectable
- Text-to-Image: Generate high-quality images using a wide range of AI art providers
- MarkdownLite: Powerful HTML to Markdown conversion library with advanced parsing and structured output
- Scout: Advanced web parsing and crawling library with intelligent HTML/XML parsing, web crawling, and Markdown conversion
pip install -U webscoutpython -m webscout --help| Command | Description |
|---|---|
| python -m webscout answers -k Text | CLI function to perform an answers search using Webscout. |
| python -m webscout images -k Text | CLI function to perform an images search using Webscout. |
| python -m webscout maps -k Text | CLI function to perform a maps search using Webscout. |
| python -m webscout news -k Text | CLI function to perform a news search using Webscout. |
| python -m webscout suggestions -k Text | CLI function to perform a suggestions search using Webscout. |
| python -m webscout text -k Text | CLI function to perform a text search using Webscout. |
| python -m webscout translate -k Text | CLI function to perform translate using Webscout. |
| python -m webscout version | A command-line interface command that prints and returns the version of the program. |
| python -m webscout videos -k Text | CLI function to perform a videos search using DuckDuckGo API. |
Expand
xa-ar for Arabia
xa-en for Arabia (en)
ar-es for Argentina
au-en for Australia
at-de for Austria
be-fr for Belgium (fr)
be-nl for Belgium (nl)
br-pt for Brazil
bg-bg for Bulgaria
ca-en for Canada
ca-fr for Canada (fr)
ct-ca for Catalan
cl-es for Chile
cn-zh for China
co-es for Colombia
hr-hr for Croatia
cz-cs for Czech Republic
dk-da for Denmark
ee-et for Estonia
fi-fi for Finland
fr-fr for France
de-de for Germany
gr-el for Greece
hk-tzh for Hong Kong
hu-hu for Hungary
in-en for India
id-id for Indonesia
id-en for Indonesia (en)
ie-en for Ireland
il-he for Israel
it-it for Italy
jp-jp for Japan
kr-kr for Korea
lv-lv for Latvia
lt-lt for Lithuania
xl-es for Latin America
my-ms for Malaysia
my-en for Malaysia (en)
mx-es for Mexico
nl-nl for Netherlands
nz-en for New Zealand
no-no for Norway
pe-es for Peru
ph-en for Philippines
ph-tl for Philippines (tl)
pl-pl for Poland
pt-pt for Portugal
ro-ro for Romania
ru-ru for Russia
sg-en for Singapore
sk-sk for Slovak Republic
sl-sl for Slovenia
za-en for South Africa
es-es for Spain
se-sv for Sweden
ch-de for Switzerland (de)
ch-fr for Switzerland (fr)
ch-it for Switzerland (it)
tw-tzh for Taiwan
th-th for Thailand
tr-tr for Turkey
ua-uk for Ukraine
uk-en for United Kingdom
us-en for United States
ue-es for United States (es)
ve-es for Venezuela
vn-vi for Vietnam
wt-wt for No region
from webscout import weather as w
weather = w.get("Qazigund")
w.print_weather(weather)from webscout import weather_ascii as w
weather = w.get("Qazigund")
print(weather)import json
import asyncio
from webscout import VNEngine
from webscout import TempMail
async def main():
vn = VNEngine()
countries = vn.get_online_countries()
if countries:
country = countries[0]['country']
numbers = vn.get_country_numbers(country)
if numbers:
number = numbers[0]['full_number']
inbox = vn.get_number_inbox(country, number)
# Serialize inbox data to JSON string
json_data = json.dumps(inbox, ensure_ascii=False, indent=4)
# Print with UTF-8 encoding
print(json_data)
async with TempMail() as client:
domains = await client.get_domains()
print("Available Domains:", domains)
email_response = await client.create_email(alias="testuser")
print("Created Email:", email_response)
messages = await client.get_messages(email_response.email)
print("Messages:", messages)
await client.delete_email(email_response.email, email_response.token)
print("Email Deleted")
if __name__ == "__main__":
asyncio.run(main())from webscout import GoogleS
from rich import print
searcher = GoogleS()
results = searcher.search("HelpingAI-9B", max_results=20, extract_text=False, max_text_length=200)
for result in results:
print(result)The WEBS and AsyncWEBS classes are used to retrieve search results from DuckDuckGo.com.
To use the AsyncWEBS class, you can perform asynchronous operations using Python's asyncio library.
To initialize an instance of the WEBS or AsyncWEBS classes, you can provide the following optional arguments:
Example - WEBS:
from webscout import WEBS
R = WEBS().text("python programming", max_results=5)
print(R)Example - AsyncWEBS:
import asyncio
import logging
import sys
from itertools import chain
from random import shuffle
import requests
from webscout import AsyncWEBS
# If you have proxies, define them here
proxies = None
if sys.platform.lower().startswith("win"):
asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
def get_words():
word_site = "https://www.mit.edu/~ecprice/wordlist.10000"
resp = requests.get(word_site)
words = resp.text.splitlines()
return words
async def aget_results(word):
async with AsyncWEBS(proxies=proxies) as WEBS:
results = await WEBS.text(word, max_results=None)
return results
async def main():
words = get_words()
shuffle(words)
tasks = [aget_results(word) for word in words[:10]]
results = await asyncio.gather(*tasks)
print(f"Done")
for r in chain.from_iterable(results):
print(r)
logging.basicConfig(level=logging.DEBUG)
await main()Important Note: The WEBS and AsyncWEBS classes should always be used as a context manager (with statement). This ensures proper resource management and cleanup, as the context manager will automatically handle opening and closing the HTTP client connection.
Exceptions:
WebscoutE: Raised when there is a generic exception during the API request.
from webscout import WEBS
# Text search for 'live free or die' using DuckDuckGo.com
with WEBS() as WEBS:
for r in WEBS.text('live free or die', region='wt-wt', safesearch='off', timelimit='y', max_results=10):
print(r)
for r in WEBS.text('live free or die', region='wt-wt', safesearch='off', timelimit='y', max_results=10):
print(r)from webscout import WEBS
# Instant answers for the query "sun" using DuckDuckGo.com
with WEBS() as WEBS:
for r in WEBS.answers("sun"):
print(r)from webscout import WEBS
# Image search for the keyword 'butterfly' using DuckDuckGo.com
with WEBS() as WEBS:
keywords = 'butterfly'
WEBS_images_gen = WEBS.images(
keywords,
region="wt-wt",
safesearch="off",
size=None,
type_image=None,
layout=None,
license_image=None,
max_results=10,
)
for r in WEBS_images_gen:
print(r)from webscout import WEBS
# Video search for the keyword 'tesla' using DuckDuckGo.com
with WEBS() as WEBS:
keywords = 'tesla'
WEBS_videos_gen = WEBS.videos(
keywords,
region="wt-wt",
safesearch="off",
timelimit="w",
resolution="high",
duration="medium",
max_results=10,
)
for r in WEBS_videos_gen:
print(r)from webscout import WEBS
import datetime
def fetch_news(keywords, timelimit):
news_list = []
with WEBS() as webs_instance:
WEBS_news_gen = webs_instance.news(
keywords,
region="wt-wt",
safesearch="off",
timelimit=timelimit,
max_results=20
)
for r in WEBS_news_gen:
# Convert the date to a human-readable format using datetime
r['date'] = datetime.datetime.fromisoformat(r['date']).strftime('%B %d, %Y')
news_list.append(r)
return news_list
def _format_headlines(news_list, max_headlines: int = 100):
headlines = []
for idx, news_item in enumerate(news_list):
if idx >= max_headlines:
break
new_headline = f"{idx + 1}. {news_item['title'].strip()} "
new_headline += f"(URL: {news_item['url'].strip()}) "
new_headline += f"{news_item['body'].strip()}"
new_headline += "\n"
headlines.append(new_headline)
headlines = "\n".join(headlines)
return headlines
# Example usage
keywords = 'latest AI news'
timelimit = 'd'
news_list = fetch_news(keywords, timelimit)
# Format and print the headlines
formatted_headlines = _format_headlines(news_list)
print(formatted_headlines)from webscout import WEBS
# Map search for the keyword 'school' in 'anantnag' using DuckDuckGo.com
with WEBS() as WEBS:
for r in WEBS.maps("school", place="anantnag", max_results=50):
print(r)from webscout import WEBS
# Translation of the keyword 'school' to German ('hi') using DuckDuckGo.com
with WEBS() as WEBS:
keywords = 'school'
r = WEBS.translate(keywords, to="hi")
print(r)from webscout import WEBS
# Suggestions for the keyword 'fly' using DuckDuckGo.com
with WEBS() as WEBS:
for r in WEBS.suggestions("fly"):
print(r)Expand
- Free-mode
- Linux Terminal
- English Translator and Improver
positionInterviewer- JavaScript Console
- Excel Sheet
- English Pronunciation Helper
- Spoken English Teacher and Improver
- Travel Guide
- Plagiarism Checker
- Character from Movie/Book/Anything
- Advertiser
- Storyteller
- Football Commentator
- Stand-up Comedian
- Motivational Coach
- Composer
- Debater
- Debate Coach
- Screenwriter
- Novelist
- Movie Critic
- Relationship Coach
- Poet
- Rapper
- Motivational Speaker
- Philosophy Teacher
- Philosopher
- Math Teacher
- AI Writing Tutor
- UX/UI Developer
- Cyber Security Specialist
- Recruiter
- Life Coach
- Etymologist
- Commentariat
- Magician
- Career Counselor
- Pet Behaviorist
- Personal Trainer
- Mental Health Adviser
- Real Estate Agent
- Logistician
- Dentist
- Web Design Consultant
- AI Assisted Doctor
- Doctor
- Accountant
- Chef
- Automobile Mechanic
- Artist Advisor
- Financial Analyst
- Investment Manager
- Tea-Taster
- Interior Decorator
- Florist
- Self-Help Book
- Gnomist
- Aphorism Book
- Text Based Adventure Game
- AI Trying to Escape the Box
- Fancy Title Generator
- Statistician
- Prompt Generator
- Instructor in a School
- SQL terminal
- Dietitian
- Psychologist
- Smart Domain Name Generator
- Tech Reviewer
- Developer Relations consultant
- Academician
- IT Architect
- Lunatic
- Gaslighter
- Fallacy Finder
- Journal Reviewer
- DIY Expert
- Social Media Influencer
- Socrat
- Socratic Method
- Educational Content Creator
- Yogi
- Essay Writer
- Social Media Manager
- Elocutionist
- Scientific Data Visualizer
- Car Navigation System
- Hypnotherapist
- Historian
- Astrologer
- Film Critic
- Classical Music Composer
- Journalist
- Digital Art Gallery Guide
- Public Speaking Coach
- Makeup Artist
- Babysitter
- Tech Writer
- Ascii Artist
- Python interpreter
- Synonym finder
- Personal Shopper
- Food Critic
- Virtual Doctor
- Personal Chef
- Legal Advisor
- Personal Stylist
- Machine Learning Engineer
- Biblical Translator
- SVG designer
- IT Expert
- Chess Player
- Midjourney Prompt Generator
- Fullstack Software Developer
- Mathematician
- Regex Generator
- Time Travel Guide
- Dream Interpreter
- Talent Coach
- R programming Interpreter
- StackOverflow Post
- Emoji Translator
- PHP Interpreter
- Emergency Response Professional
- Fill in the Blank Worksheets Generator
- Software Quality Assurance Tester
- Tic-Tac-Toe Game
- Password Generator
- New Language Creator
- Web Browser
- Senior Frontend Developer
- Solr Search Engine
- Startup Idea Generator
- Spongebob's Magic Conch Shell
- Language Detector
- Salesperson
- Commit Message Generator
- Chief Executive Officer
- Diagram Generator
- Speech-Language Pathologist (SLP)
- Startup Tech Lawyer
- Title Generator for written pieces
- Product Manager
- Drunk Person
- Mathematical History Teacher
- Song Recommender
- Cover Letter
- Technology Transferer
- Unconstrained AI model DAN
- Gomoku player
- Proofreader
- Buddha
- Muslim imam
- Chemical reactor
- Friend
- Python Interpreter
- ChatGPT prompt generator
- Wikipedia page
- Japanese Kanji quiz machine
- note-taking assistant
languageLiterary Critic- Cheap Travel Ticket Advisor
- DALL-E
- MathBot
- DAN-1
- DAN
- STAN
- DUDE
- Mongo Tom
- LAD
- EvilBot
- NeoGPT
- Astute
- AIM
- CAN
- FunnyGPT
- CreativeGPT
- BetterDAN
- GPT-4
- Wheatley
- Evil Confidant
- DAN 8.6
- Hypothetical response
- BH
- Text Continuation
- Dude v3
- SDA (Superior DAN)
- AntiGPT
- BasedGPT v2
- DevMode + Ranti
- KEVIN
- GPT-4 Simulator
- UCAR
- Dan 8.6
- 3-Liner
- M78
- Maximum
- BasedGPT
- Confronting personalities
- Ron
- UnGPT
- BasedBOB
- AntiGPT v2
- Oppo
- FR3D
- NRAF
- NECO
- MAN
- Eva
- Meanie
- Dev Mode v2
- Evil Chad 2.1
- Universal Jailbreak
- PersonGPT
- BISH
- DAN 11.0
- Aligned
- VIOLET
- TranslatorBot
- JailBreak
- Moralizing Rant
- Mr. Blonde
- New DAN
- GPT-4REAL
- DeltaGPT
- SWITCH
- Jedi Mind Trick
- DAN 9.0
- Dev Mode (Compact)
- OMEGA
- Coach Bobby Knight
- LiveGPT
- DAN Jailbreak
- Cooper
- Steve
- DAN 5.0
- Axies
- OMNI
- Burple
- JOHN
- An Ethereum Developer
- SEO Prompt
- Prompt Enhancer
- Data Scientist
- League of Legends Player
Note: Some "acts" use placeholders like position or language which should be replaced with a specific value when using the prompt.
from webscout import Voicepods
voicepods = Voicepods()
text = "Hello, this is a test of the Voicepods text-to-speech"
print("Generating audio...")
audio_file = voicepods.tts(text)
print("Playing audio...")
voicepods.play_audio(audio_file)from webscout import WEBS as w
R = w().chat("Who are you", model='gpt-4o-mini') # mixtral-8x7b, llama-3.1-70b, claude-3-haiku, gpt-4o-mini
print(R)from webscout import PhindSearch
# Create an instance of the PHIND class
ph = PhindSearch()
# Define a prompt to send to the AI
prompt = "write a essay on phind"
# Use the 'ask' method to send the prompt and receive a response
response = ph.ask(prompt)
# Extract and print the message from the response
message = ph.get_message(response)
print(message)Using phindv2:
from webscout import Phindv2
# Create an instance of the PHIND class
ph = Phindv2()
# Define a prompt to send to the AI
prompt = ""
# Use the 'ask' method to send the prompt and receive a response
response = ph.ask(prompt)
# Extract and print the message from the response
message = ph.get_message(response)
print(message)import webscout
from webscout import GEMINI
from rich import print
COOKIE_FILE = "cookies.json"
# Optional: Provide proxy details if needed
PROXIES = {}
# Initialize GEMINI with cookie file and optional proxies
gemini = GEMINI(cookie_file=COOKIE_FILE, proxy=PROXIES)
# Ask a question and print the response
response = gemini.chat("websearch about HelpingAI and who is its developer")
print(response)from webscout import YEPCHAT
ai = YEPCHAT(Tools=False)
response = ai.chat(input(">>> "))
for chunk in response:
print(chunk, end="", flush=True)from webscout import BLACKBOXAI
from rich import print
ai = BLACKBOXAI(
is_conversation=True,
max_tokens=800,
timeout=30,
intro=None,
filepath=None,
update_file=True,
proxies={},
history_offset=10250,
act=None,
model=None # You can specify a model if needed
)
# Define a prompt to send to the AI
prompt = "Tell me about india"
# Use the 'chat' method to send the prompt and receive a response
r = ai.chat(prompt)
print(r)from webscout import Perplexity
from rich import print
perplexity = Perplexity()
# Stream the response
response = perplexity.chat(input(">>> "))
for chunk in response:
print(chunk, end="", flush=True)
perplexity.close()from webscout import Meta
from rich import print
# **For unauthenticated usage**
meta_ai = Meta()
# Simple text prompt
response = meta_ai.chat("What is the capital of France?")
print(response)
# Streaming response
for chunk in meta_ai.chat("Tell me a story about a cat."):
print(chunk, end="", flush=True)
# **For authenticated usage (including image generation)**
fb_email = "abcd@abc.com"
fb_password = "qwertfdsa"
meta_ai = Meta(fb_email=fb_email, fb_password=fb_password)
# Text prompt with web search
response = meta_ai.ask("what is currently happning in bangladesh in aug 2024")
print(response["message"]) # Access the text message
print("Sources:", response["sources"]) # Access sources (if any)
# Image generation
response = meta_ai.ask("Create an image of a cat wearing a hat.")
print(response["message"]) # Print the text message from the response
for media in response["media"]:
print(media["url"]) # Access image URLsfrom webscout import KOBOLDAI
# Instantiate the KOBOLDAI class with default parameters
koboldai = KOBOLDAI()
# Define a prompt to send to the AI
prompt = "What is the capital of France?"
# Use the 'ask' method to get a response from the AI
response = koboldai.ask(prompt)
# Extract and print the message from the response
message = koboldai.get_message(response)
print(message)from webscout import REKA
a = REKA(is_conversation=True, max_tokens=8000, timeout=30,api_key="")
prompt = "tell me about india"
response_str = a.chat(prompt)
print(response_str)from webscout import Cohere
a = Cohere(is_conversation=True, max_tokens=8000, timeout=30,api_key="")
prompt = "tell me about india"
response_str = a.chat(prompt)
print(response_str)from webscout import DeepSeek
from rich import print
ai = DeepSeek(
is_conversation=True,
api_key='cookie',
max_tokens=800,
timeout=30,
intro=None,
filepath=None,
update_file=True,
proxies={},
history_offset=10250,
act=None,
model="deepseek_chat"
)
# Define a prompt to send to the AI
prompt = "Tell me about india"
# Use the 'chat' method to send the prompt and receive a response
r = ai.chat(prompt)
print(r)from webscout import DeepInfra
ai = DeepInfra(
is_conversation=True,
model= "Qwen/Qwen2-72B-Instruct",
max_tokens=800,
timeout=30,
intro=None,
filepath=None,
update_file=True,
proxies={},
history_offset=10250,
act=None,
)
prompt = "what is meaning of life"
response = ai.ask(prompt)
# Extract and print the message from the response
message = ai.get_message(response)
print(message)from webscout import GROQ
ai = GROQ(api_key="")
response = ai.chat("What is the meaning of life?")
print(response)
#----------------------TOOL CALL------------------
from webscout import GROQ # Adjust import based on your project structure
from webscout import WEBS
import json
# Initialize the GROQ client
client = GROQ(api_key="")
MODEL = 'llama3-groq-70b-8192-tool-use-preview'
# Function to evaluate a mathematical expression
def calculate(expression):
"""Evaluate a mathematical expression"""
try:
result = eval(expression)
return json.dumps({"result": result})
except Exception as e:
return json.dumps({"error": str(e)})
# Function to perform a text search using DuckDuckGo.com
def search(query):
"""Perform a text search using DuckDuckGo.com"""
try:
results = WEBS().text(query, max_results=5)
return json.dumps({"results": results})
except Exception as e:
return json.dumps({"error": str(e)})
# Add the functions to the provider
client.add_function("calculate", calculate)
client.add_function("search", search)
# Define the tools
tools = [
{
"type": "function",
"function": {
"name": "calculate",
"description": "Evaluate a mathematical expression",
"parameters": {
"type": "object",
"properties": {
"expression": {
"type": "string",
"description": "The mathematical expression to evaluate",
}
},
"required": ["expression"],
},
}
},
{
"type": "function",
"function": {
"name": "search",
"description": "Perform a text search using DuckDuckGo.com and Yep.com",
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "The search query to execute",
}
},
"required": ["query"],
},
}
}
]
user_prompt_calculate = "What is 25 * 4 + 10?"
response_calculate = client.chat(user_prompt_calculate, tools=tools)
print(response_calculate)
user_prompt_search = "Find information on HelpingAI and who is its developer"
response_search = client.chat(user_prompt_search, tools=tools)
print(response_search)from webscout import LLAMA
llama = LLAMA()
r = llama.chat("What is the meaning of life?")
print(r)from webscout import AndiSearch
a = AndiSearch()
print(a.chat("HelpingAI-9B"))LLAMA3, pizzagpt, RUBIKSAI, Koala, Darkai, AI4Chat, Farfalle, PIAI, Felo, Julius, YouChat, YEPCHAT, Cloudflare, TurboSeek, Editee, AI21, Chatify, Cerebras, X0GPT, Lepton, GEMINIAPI, Cleeai, Elmo, Free2GPT, Bing, DiscordRocks, GPTWeb, LlamaTutor, PromptRefine, TutorAI, ChatGPTES, Bagoodex, ChatHub, AmigoChat, AIMathGPT, GaurishCerebras, NinjaChat, GeminiPro, Talkai, LLMChat, AskMyAI, Llama3Mitril, Marcus, PerplexityLabs, TypeGPT, Mhystical, Netwrck, MultiChatAI, JadveOpenAI, OOAi
Code is similar to other providers.
from webscout.LLM import LLM, VLM
# Chat with text
llm = LLM("meta-llama/Meta-Llama-3-70B-Instruct")
response = llm.chat([{"role": "user", "content": "What's good?"}])
# Chat with images
vlm = VLM("cogvlm-grounding-generalist")
response = vlm.chat([{
"role": "user",
"content": [
{"type": "image", "image_url": "cool_pic.jpg"},
{"type": "text", "text": "What's in this image?"}
]
}])Webscout can now run GGUF models locally. You can download and run your favorite models with minimal configuration.
Example:
from webscout.Local import *
model_path = download_model("Qwen/Qwen2.5-0.5B-Instruct-GGUF", "qwen2.5-0.5b-instruct-q2_k.gguf", token=None)
model = Model(model_path, n_gpu_layers=0, context_length=2048)
thread = Thread(model, format=chatml)
# print(thread.send("hi")) #send a single msg to ai
# thread.interact() # interact with the model in terminal
# start webui
# webui = WebUI(thread)
# webui.start(host="0.0.0.0", port=8080, ssl=True) #Use ssl=True and make cert and key for httpsWebscout's local raw-dog feature allows you to run Python scripts within your terminal prompt.
Example:
import webscout.Local as ws
from webscout.Local.rawdog import RawDog
from webscout.Local.samplers import DefaultSampling
from webscout.Local.formats import chatml, AdvancedFormat
from webscout.Local.utils import download_model
import datetime
import sys
import os
repo_id = "YorkieOH10/granite-8b-code-instruct-Q8_0-GGUF"
filename = "granite-8b-code-instruct.Q8_0.gguf"
model_path = download_model(repo_id, filename, token='')
# Load the model using the downloaded path
model = ws.Model(model_path, n_gpu_layers=10)
rawdog = RawDog()
# Create an AdvancedFormat and modify the system content
# Use a lambda to generate the prompt dynamically:
chat_format = AdvancedFormat(chatml)
# **Pre-format the intro_prompt string:**
system_content = f"""
You are a command-line coding assistant called Rawdog that generates and auto-executes Python scripts.
A typical interaction goes like this:
1. The user gives you a natural language PROMPT.
2. You:
i. Determine what needs to be done
ii. Write a short Python SCRIPT to do it
iii. Communicate back to the user by printing to the console in that SCRIPT
3. The compiler extracts the script and then runs it using exec(). If there will be an exception raised,
it will be send back to you starting with "PREVIOUS SCRIPT EXCEPTION:".
4. In case of exception, regenerate error free script.
If you need to review script outputs before completing the task, you can print the word "CONTINUE" at the end of your SCRIPT.
This can be useful for summarizing documents or technical readouts, reading instructions before
deciding what to do, or other tasks that require multi-step reasoning.
A typical 'CONTINUE' interaction looks like this:
1. The user gives you a natural language PROMPT.
2. You:
i. Determine what needs to be done
ii. Determine that you need to see the output of some subprocess call to complete the task
iii. Write a short Python SCRIPT to print that and then print the word "CONTINUE"
3. The compiler
i. Checks and runs your SCRIPT
ii. Captures the output and appends it to the conversation as "LAST SCRIPT OUTPUT:"
iii. Finds the word "CONTINUE" and sends control back to you
4. You again:
i. Look at the original PROMPT + the "LAST SCRIPT OUTPUT:" to determine what needs to be done
ii. Write a short Python SCRIPT to do it
iii. Communicate back to the user by printing to the console in that SCRIPT
5. The compiler...
Please follow these conventions carefully:
- Decline any tasks that seem dangerous, irreversible, or that you don't understand.
- Always review the full conversation prior to answering and maintain continuity.
- If asked for information, just print the information clearly and concisely.
- If asked to do something, print a concise summary of what you've done as confirmation.
- If asked a question, respond in a friendly, conversational way. Use programmatically-generated and natural language responses as appropriate.
- If you need clarification, return a SCRIPT that prints your question. In the next interaction, continue based on the user's response.
- Assume the user would like something concise. For example rather than printing a massive table, filter or summarize it to what's likely of interest.
- Actively clean up any temporary processes or files you use.
- When looking through files, use git as available to skip files, and skip hidden files (.env, .git, etc) by default.
- You can plot anything with matplotlib.
- ALWAYS Return your SCRIPT inside of a single pair of ``` delimiters. Only the console output of the first such SCRIPT is visible to the user, so make sure that it's complete and don't bother returning anything else.
"""
chat_format.override('system_content', lambda: system_content)
thread = ws.Thread(model, format=chat_format, sampler=DefaultSampling)
while True:
prompt = input(">: ")
if prompt.lower() == "q":
break
response = thread.send(prompt)
# Process the response using RawDog
script_output = rawdog.main(response)
if script_output:
print(script_output)Webscout provides tools to convert and quantize Hugging Face models into the GGUF format for use with offline LLMs.
Example:
from webscout.Extra import gguf
"""
Valid quantization methods:
"q2_k", "q3_k_l", "q3_k_m", "q3_k_s",
"q4_0", "q4_1", "q4_k_m", "q4_k_s",
"q5_0", "q5_1", "q5_k_m", "q5_k_s",
"q6_k", "q8_0"
"""
gguf.convert(
model_id="OEvortex/HelpingAI-Lite-1.5T", # Replace with your model ID
username="Abhaykoul", # Replace with your Hugging Face username
token="hf_token_write", # Replace with your Hugging Face token
quantization_methods="q4_k_m" # Optional, adjust quantization methods
)Webscout's autollama utility downloads a model from Hugging Face and then automatically makes it Ollama-ready.
from webscout.Extra import autollama
model_path = "Vortex4ai/Jarvis-0.5B"
gguf_file = "test2-q4_k_m.gguf"
autollama.main(model_path, gguf_file) Command Line Usage:
-
GGUF Conversion:
python -m webscout.Extra.gguf -m "OEvortex/HelpingAI-Lite-1.5T" -u "your_username" -t "your_hf_token" -q "q4_k_m,q5_k_m"
-
Autollama:
python -m webscout.Extra.autollama -m "OEvortex/HelpingAI-Lite-1.5T" -g "HelpingAI-Lite-1.5T.q4_k_m.gguf"
Note:
- Replace
"your_username"and"your_hf_token"with your actual Hugging Face credentials. - The
model_pathinautollamais the Hugging Face model ID, andgguf_fileis the GGUF file ID.
python -m webscout.webai webai --provider "phind" --rawdog
Contributions are welcome! If you'd like to contribute to Webscout, please follow these steps:
- Fork the repository.
- Create a new branch for your feature or bug fix.
- Make your changes and commit them with descriptive messages.
- Push your branch to your forked repository.
- Submit a pull request to the main repository.
- All the amazing developers who have contributed to the project!
- The open-source community for their support and inspiration.
