Saturday, November 8, 2025

MCP - Model Context Protocol

November 08, 2025 0

 

MCP - Model Context Protocol:

MCP defined a LLM to access an external data, tools and context in a a structure way. MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems and data.

Overview of MCP:

AI application such as Claude or chatGPT can connect to data sources, tools [search engine] and workflow [prompts] through MCP and perform a tasks.

MCP like an interface which communicated to MCP client and discover their requirement and offer available services for their requirement. 
MCP Framework:
  • MCP SDK - It is a foundation for all the MCP development. It will use for Production and standard projects. It can be integrate into any tools or transport (STDIO, SSE)
  • FASTMCP 1.0 - It became a legacy support and integrated into MCP python SDK.
  • FASTMCP 2.0 - This is a latest and modern feature tools kits for advanced MCP workflows.
  • Others Frameworks - Java SDK and third party libs in other languages.
Agent workflows inside of Memory:


RAG - Retrieval Augmented Generation
It converts a data into numerical representation where each piece of data has information about how it relates to others.
Retrieval - when user ask a question or search, RAG turns question or search into own numerical representation (Embedding) and find a data which is similar meanings.
Augmentation - The top search result are then added into prompt and send to back to LLM
Generation - The search results give the LLM some local context and consider as response.
Embedding:
Embedding represent text as set of numerical data along with tensors (different dimensions)
Each dimension will store some information about text meaning or syntactical meaning.
Each words or sentence with similar meaning are stored near by vector space.
Models will learn to place a similar words or sentences close  together in the embedded space.
Common pre-trained models such as BERT and RoBERTs are  used for generating an embedding inside of vector space.
We can able to use an embedded for NLP tasks like semantic search, text classification and sentimental analysis.
Agentic RAG:
It is integrate an AI agents to enhance the RAG approach. It will breakdown from complex queries into manageable parts and using API tools where need to augment processing and better result.


Tags # AI # LLM # MlOps Continue Reading
at No comments:
Tags , ,

Implementation of AI agent

November 08, 2025 0


                                 

Installation of Ollama:
Ollama is an open source tool which will helps us to run a NLP [Natural Language Processing] through locally.
Step1) Downloading the Ollawa tool for your suitable operating system and installed it.





Tags # AI # LLM # MlOps Continue Reading
at No comments:
Tags , ,

Friday, July 25, 2025

Data Science - Basics

July 25, 2025 0




Data science:

Data has classified as three types:
Unstructured data: the data is collected in a random way such as social media data, Phone calls, Weather.
Structured data: It is collected a data in a structural and specific data. 
DB and APP data: Customer and invoice data.
ML has a four life cycles:
Build, Deploy, Train, Manage
Jupyter Lab:
* It is a webbased interface for notebook sessions
* Ability to work with integrate documents and activity including:
* Jupyter Notebook
* Text editors
* Terminals
Features of Jupyter Lab:
Menu Bar : Top level menu that expose actions available in JupyterLab
Launcher: Provide easy access to your notbooks, console, text editor and environment explorer
Left Sidebar: File browser and command Palette
Conda environment:
It is an Open source and environment management system.
Features of Conda Environment:
* Install and updated their dependencies
* Maintain a different software group
* Change over between environments
* Develop a netbook and deploy a modules
Data Science use of Conda environment:
Provide a below specific framework and build a module:
*PyTorch
*TensorFlow
General Machine learning Algorithm:
use case:
* Data manipulation
* Supervised machine learning 
* AutoML functionality
* Machine learning explainability
Top Libraries used for GML:
* category encoders
* lightgbm
* scikit-learn
* TesorFlow
Natural Language Process [NLP]
Use case:
* Text extraction
* Part of speech tagging
* Key pharse extraction
Top Libraries of NLP:
* nltk
* transformers
* eli5
* Lime
ONNX:
Opensource software for the Data science.
Use cases:
* Portability and interoperability between ML frameworks
* ONNX runtime liberary allows us to run a module on different platforms.
Top Libraries of ONNX:
* onnx
* onnxconverter-common
* onnxmltools
* onnxruntime
PyTorch:
It is an opensource of machine learning framework. It is used mulitple of deep learning algorithm. 
Use cases:
* Computer vision, NLP and general machine learning
* Deep neural networks and algorithms for deep learning.
Top Libraries of PyTorch:
* Panda
* daal4py
TensorFlow:
Use cases:
* Machine learning
* Deep neural networks
* Flexible architecture run on CPUs, GPUs and TPUs
Top Libraries of TensorFlow:
* Panda
* scikit-learn
* Tensorboard
* TensorFlow
Tags # Datascience Continue Reading
at No comments:
Tags

Sunday, July 20, 2025

AI Basics

July 20, 2025 0

 


Supervised Learning:
It ill provides an output if we give an input into the applications called supervised Learning.
Machine Learning Vs Data Science:
Machine Learning:
Field of study that's gives computers the ability to learn without explicitly programs.
Ex. YouTube Advertisement and online shopping. It will generate an advertisement and shopping related notification for the user interest.
Data Science:
It will extract the knowledge and insights from data.
Ex. Share market data. It will analysis an insight of data and decided a probability of output.
Deep Learning:
It will take a multiple input and decided output like human brain by using a Artificial Neural Network.
Open-source frameworks for Machine Learning tool:
  • PyTorch
  • TensorFlow
  • Hugging Face
  • PaddlePaddle
  • Scikit-learn
  • R
How is Alexa works?
Steps to process the command:
1) Trigger word detection to activate the device [ Hi Alexa]
2) Speech recognition - "what is the weather in Delhi today" [convert audio file into text]
3) Intent recognition - purpose of the user "weather in Delhi"
4) Execute weather query and given output to user.
Generative AI:
AI system that can produce a high-quality content like text, image and audio. It used the machine learning model and learned a data and generate an output content.




 
Tags # AI Continue Reading
at No comments:
Tags

Monday, June 23, 2025

Ansible Modules

June 23, 2025 0

 

inlinefile module:

Adding /Modify or delete a line inside of file.
Main parameter:
path - full path of the file
line - text
insertbefore / insertafter - EOF/regular expression
validate - Validation of command
state - Present/absent
mode/owner/group - permission
setype/seuser/selevel - SElinux setup
Ping Module:
It is validated the host reachability of remote host.
ansible.builtin.ping is a module name.
Reboot Module:
We can reboot a remote host through reboot module.
Main parameter of this reboot module as below:
reboot_timeout - 600
msg - text of reboot notification
reboot_command - define a reboot command depends up on OS
pre_reboot_delay - 0
Post_reboot_delay - 0
test_command - 'whoami'
boot_time_command - "cat /proc/sys/kernel/boot_id"

Copy Module:
Copy a file from one location to other location.
Main Parameters:
dest - Remote file path
src - local file path
fail_on_missing - yes / no
validate_checksum - yes /no
flat - yes/no
Service Module:
We can enable and disable the system services through this module.
ansible.builtin.service_facts
Main parameters:
name : Service name
state :  started, stopped, restarted, reloaded
enabled: yes/no
arguments/args : extra args
Package installation module:
Main parameters:
name - Name of the package
state - present/installed/absent/removed/latest
Create a file:
Main parameter:
path - file path
state - absent/directory/hard/link/touch
Module for a file permission change:
Main parameters:
Path - file path
owner - user
group - group
mode - rwx mode
state - file state [absent/directory/hard/link/touch]
setype/seuser/selevel - SElinux
Module for Download a file through Internet
Main parameters:
url - download URL
dest - destionation path
force - no/yes
checksum - checksum:URL
force_basic_auth/url_username/url_password/use_gssapi - HTTP basic auth/GSSAPI kerberos
headers - custom HTTP headers
http_agent - ansible-http-get
owner/group/mode - permission
setype/seuser/selevel - SElinux

Example:
---
 - name: Download an ansible package
   hosts: all
   become: false
   gather_facts: false
   vars: 
     myurl: "https://releases.ansible.com/ansible/ansible-2.9.25.tar.gz"
mycrc: "sha256:https://releases.ansible.com/ansible/ansible-2.9.25.tar.gz"
mydest: "/home/test/ansible-2.9.25.tar.gz"
   tasks:
     - name: downloading an ansible file
   ansible.builtin.get_url:
     url: "{{ myurl }}" 
desk: "{{ mydest }}"
checksum: "{{ mycrc }}"
mode: '0644'
owner: devops
group: wheel
Module for backing up the file
Main Parameters:
src - source path
dest - destionation path
archive - mirrors the rsync archive flag, enables recursive, links, perms, times, owner, group, flags 
rsync_opts - no/yes

Changed the line inside of file:
---
- name: search demo
  hosts: all
  vars:
    myfile: "/etc/ssh/sshd_config"
    myline: 'PasswordAuthentication no'
  become: true
  tasks:
    - name: string found
      ansible.builtin.lineinfile:
        name: "{{ myfile }]"
        line: "{{ myline }}"
        state: present
      check_mode: true
      register: conf
      failed_when:(conf is changed) or (conf is failed)

Tags # Ansible # Automation # DevOps Continue Reading
at No comments:
Tags , ,

Monday, May 12, 2025

Prometheus - Write a Query through PromQL - Part 2

May 12, 2025 0

 

PromQL:
PromQL expressions has classified as follows:
Instant Vector
Range Vector
Scalar
String
Instant and Range vector has a basic syntax consist of metric name along with label matchers. A metric name in Prometheus is a label named __name__ under the value. 
example:
{__name__="prometheus_build_info"}
PromQL will query directly by name [prometheus_build_info]

Label matchers are defined using a label key, a label matching operator, and the label value. Some possible label-matching operators:
=: The label value matches the specified string
!=: The label value does not match the specified string
=~: The label value matches the regex in the specified string
!~: The label value does not match the regex in the specified string

Instant Vector are selectors that select the data at this instant using a recent available data for the time series which you are selected. 
Range Vector is retriving data over a given time range. This time range is specified as duration followed by unit.
The following are valid units for a duration:

ms: Milliseconds
s: Seconds
m: Minutes
h: Hours
d: Days (assumes a day is always 24 hours)
w: Weeks (assumes a week is always 7 days)
y: Years (assumes a year is always 365 days)

These range vector selectors are often more useful when aggregating data to perform an analysis to drive results such as the number of requests per second a service experienced over a particular time range.
Prometheus has two primary API endpoints that are used to retrive time series data /api/v1/query and /api/v1/query_range.
Note : Both instant vectors and range vectors are valid for the /query endpoint, but only instant vectors are valid for the /query_range endpoint.
Offsets:
It will allow us to pull the time series data from the past based on offset values.
Example:
my_cool_custom_metric offset 5m
Group Modifier and Vector Matching:
Prometheus has provided ways to join the labels of different matrics together when using the group_left and group_right query modifiers. This option allows for many-to-one and one-to-many matching of vectors.
Logical and set binary operators
PromQL used the boolean operators of and,or and unless query operators
Example:
vector1 and vector2
It requires that for every series in vector1, there must be an exactly matching label set in vector2 (besides the __name__ label). If there is no match for the label set of a series in vector1 inside of vector2, then that series is not included in the results.  

Tags # DevOps # Monitoring # Prometheus Continue Reading
at No comments:
Tags , ,

Sunday, May 4, 2025

Prometheus - Time Series Database - Part 1

May 04, 2025 0

 


Prometheus Installation:

We can install the Prometheus in different ways.
  •  Install through source
  • Docker container
  • Install through script
  • Install through package manager
Prometheus with a metrics comes from node_exporter process from remote hosts. Each metrics has associated with one or more-time series data on it.
Events are point-in-time of record of action occurring during that time.
Prometheus is using a pull-based model for extracting the data from systems it monitors whereas Graphite and Nagios are using push-based model which push the metrics to a remote system.
Components of Prometheus Stack:
  • Prometheus Alert manager for routing an alert coming from Prometheus 
  • Exporters: It has multiple exporter components depends upon a device such as Node Exporter.
  • Grafana: It used to visualize metrics from Prometheus through dashboards.
Prometheus has four main components such as time series database [TSDB], scrape manager, rule manager and Web UI.
Time Series Database:
   It is a special kind of database that is optimized for storing data points in a time series manner. The core part of TSDB are head blocks [Write-ahead log] and data format (block, chunks, and indices). The head block is an entry point for samples being scraped and stored in TSDB. 
The chunk gets appending into head block, until chunk hits it sample limit until stays in memory. It will add into WAL at first before added into memory, so we will preserve the data if server is in panic or reboot at automatically.
To increase resource efficiency, some wizardry is done to the chunk data so that it only stores the direct value of the first timestamp (t0) and first value (v0). Subsequent timestamps are set to the delta of the prior timestamp, so starting at t2, all timestamps are the delta of a delta. Similarly, subsequent values are compared to their prior value using a bitwise XOR operator. This just means that the difference between the samples is stored, and if they’re the same, then 0 is stored.
Scrape Manager:
  The scrape manager is a part of Prometheus that handles pulling metrics from applications and exporters by performing scraping and maintaining an internal list of what things should be scraped by Prometheus.
Rule Manager:
  The rule manager is part of Prometheus that handles evaluating alerts and recording the rules as per Prometheus. It is handling evaluating rule groups compromised of alerts and recording rules on regular intervals. 
Web UI/API:
  The Web UI/API is the portion of Prometheus that you can access via your browser. The Prometheus UI/API is a powerful REST API. It provides integration with other graphical tools such as Grafana. 
Alert Manager:
  It is responsible to send an alert to Slack, PagerDuty and other alerting destinations.  Alert Manager handles all of that owns through routing tree-based workflow. 





Tags # DevOps # Monitoring # Prometheus Continue Reading
at No comments:
Tags , ,