ezzabuzaid
@ezzabuzaid
Empirical Engineer. Creating tools is my hobby.
32
Published Tools
78
Total Stars
6,368
Weekly Downloads
108
GitHub Followers
79
Public Repos
100/100
Avg Security
Published Tools
31 Skills1 Agentacross 11 categoriesezzabuzaid
Create a runtime HTTP client or AI SDK tools directly from an OpenAPI document, without generating source files.
ezzabuzaid
Load OpenAPI or Postman documents and normalize them into the intermediate representation shared by SDK-IT generators.
ezzabuzaid
A framework for building multi-agent AI systems with TypeScript. Create agents that use tools, coordinate through handoffs, and work together to solve complex tasks.
...moreezzabuzaid
AI SDK-native conversation compaction with explicit budgets and provenance.
ezzabuzaid
Development-only Zukhruf DevTool UI. `devtool()` returns a mountable Hono app containing only the bundled browser assets and their SPA fallback; consumers do not import React or CSS, and the DevTool owns no listener, port, runtime object, credentials, or
...moreezzabuzaid
AI-powered natural language to SQL. Ask questions in plain English, get executable queries.
ezzabuzaid
A general-purpose LLM evaluation framework with dataset loading, scoring, run persistence, model comparison, and console reporting.
...moreezzabuzaid
A domain-agnostic context system for LLM applications, including fragment rendering, persistence, chat orchestration, and sandbox tooling.
...moreezzabuzaid
A home for unstable, in-progress building blocks that are not yet part of the stable `@deepagents/*` surface. APIs here may change without notice.
...moreezzabuzaid
A collection of document embeddings in SQLite. It embeds only new or changed text through an AI SDK embedding model, and finds the nearest documents of a document by exact cosine similarity inside a publish-date window.
...moreezzabuzaid
A pipeline that copies the data of apps and services into stores that SQL can read: sources, streams, checkpoints, watches and derived datasets.
...moreezzabuzaid
A rule of an elt computed dataset: it gives each loaded row a vector of its text, made by an AI SDK embedding model, so a search can find rows by meaning.
...moreezzabuzaid
The SQLite destination of elt: each stream is a raw table in one SQLite file, with a reader view over it and a catalog of each table, view and column.
...moreezzabuzaid
The Postgres destination of elt: each stream is a typed table in _warehouse, with a reader view in marts that the host names, a history of each row for readings as of a position, and the views and computed datasets that a pipeline derives.
...moreezzabuzaid
A rule of an elt computed dataset: it cuts the text of each loaded row into chunks that each count at most a number of tokens of the embedding model's tokenizer, so each chunk can get its own vector.
...moreezzabuzaid
Disposable fixtures for integration tests: Docker containers and database servers, SQLite and DuckDB databases, HTTP servers and streams. A supervisor removes what a killed test run left behind.
...moreezzabuzaid
A coprocess: a long-lived child process that the parent talks to over stdin and stdout, with JSON values on lines, and stops with SIGTERM and then SIGKILL after a grace period.
...moreezzabuzaid
A client for the Docker Engine API on Node.js. It selects the engine that the Docker CLI selects, and it sends HTTP requests to that engine through a Unix socket, a Windows named pipe or SSH, with no Docker CLI and no third-party dependencies.
...moreezzabuzaid
Ranks a stream of articles for a reader's profile through a scorer port (a decision model, or embedding similarity calibrated on the reader's answers), with thresholds that turn the probabilities into yes or no.
...moreezzabuzaid
A TypeScript port of the MTEB evaluation engine (Massive Text Embedding Benchmark, version 2.26.28): tasks, evaluators, metrics, search, results and the result cache. Retrieval and reranking first; the encoder is an AI SDK embedding model.
...moreezzabuzaid
The Nx part of the monorepo standard: inference plugins for format, test, lint, opentofu and prisma targets, and generators for the root files (preset, init), apps (node-app, react-app, pair, ui-package), a db package (prisma) and delivery (dokploy).
...moreezzabuzaid
Postgres's LISTEN and NOTIFY for one SQLite file: a notification is a row written in the sender's transaction, and each listening program of the computer gets it after the commit, woken over its own Unix socket or named pipe.
...moreezzabuzaid
Vector types, their math, and the SQL form of each vector in SQLite, sqlite-vec and SQL Server, which a program composes into its own SQL. After the vector types of pgvector and the TensorPrimitives of .NET.
...moreezzabuzaid
An AI SDK embedding middleware that calls the model only for new text and gives back exactly the numbers it gave, kept in memory, SQLite or SQL Server. After .NET's CachingEmbeddingGenerator.
...moreezzabuzaid
Puts reports of one real event into one event. It shortlists the nearest reports by embedding similarity, scores each pair with a frozen logistic model, and clusters the 7-day window again with group average on each new publish day.
...moreezzabuzaid
Reads one SQL Server database: discovery, exact reads, Change Tracking and change signals. It knows nothing of elt.
ezzabuzaid
An elt source for one SQL Server database: each user table that the login can read becomes a stream.
ezzabuzaid
<p align="center">A type-safe SDK generator that converts OpenAPI specifications into TypeScript client.</p>
ezzabuzaid
Hono framework integration for SDK-IT that provides type-safe request validation and semantic response handling.
ezzabuzaid
<p align="center">Command-line interface for SDK-IT that simplifies generating type-safe client SDKs from OpenAPI specifications</p>
...moreezzabuzaid
A graceful shutdown for a Node.js program: SIGINT, SIGTERM, SIGHUP and crashes abort one AbortSignal, the program runs its cleanup, and then the process dies by the termination signal, so a shell, Docker or Kubernetes sees an interrupted process.
...moreezzabuzaid
One pool of long-lived workers, such as native helpers, for each computer: every program sends its jobs to one store, and an elected owner runs them, while startups, sizings and evictions decide when workers start, how many run and when they close.
...more