Skip to content
View nripankadas07's full-sized avatar

Block or report nripankadas07

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
nripankadas07/README.md

Nripanka Das

The No-Dependency AI + Systems Lab: small, readable, tested AI and systems tools for developers who want code they can run, understand, audit, and extend.

This profile is organized as a compact lab: a few flagship projects that carry the main technical ideas, plus a catalog of small parser, systems, AI, DevEx, and TypeScript libraries. The bias is toward source you can read in one sitting, examples that run from a fresh clone, and documentation that explains the tradeoffs.

Six Flagships To Pin

GitHub only gives the first impression room for a handful of repositories. These six are the front door.

Project Technical Thesis First Demo
patchgym Repo-specific coding-agent evaluation: mine real Git history into hidden-test tasks and grade whether agents actually fixed the code. patchgym demo
agent-framework Agents from scratch, but actually usable: a tiny inspectable runtime with tools, memory, traces, and safe no-key examples. python examples/no_api_key_agent.py
rag-pipeline RAG from first principles: chunking, retrieval, citations, evaluation, and reports without hiding the retrieval loop. python examples/local_rag_demo.py
prompt-eval Unit tests for prompts: regression checks that can run in CI without API keys. python examples/no_api_key_regression.py
safejson JSON parsing as a security boundary: duplicate-key detection, size/depth limits, and adversarial tests. python examples/security_boundary.py
decimal-ts Exact decimal arithmetic for money-style calculations in TypeScript, backed by BigInt instead of floats. npm install && npm run demo

Start Here

Each flagship is meant to be useful even if you only read it, and runnable in less than five minutes from a clean checkout.

git clone https://github.com/nripankadas07/patchgym
cd patchgym
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e ".[dev]"
patchgym demo
python -m pytest -q
git clone https://github.com/nripankadas07/decimal-ts
cd decimal-ts
npm install
npm run demo
npm test

The same pattern holds across the lab: clone, run the demo, run the tests, read the limitations before using the library as a dependency.

The Lab Catalog

AI, Evaluation, And Local-First Tooling

patchgym, agent-framework, rag-pipeline, prompt-eval, ai-toolkit, token-counter, schema-gen.

Security, Configuration, And Operational Tools

safejson, dep-audit, digestlite, env-vault, envdiff, config-loader, feature-flags, rate-limiter, retryback, retry-lib, task-queue, api-mocker, log-parser.

Parsers, Codecs, And Data Formats

tomlmini, bencode, jsonptr, jsonpatch-lite, iso8601, iniparse, urltemplate, querystring, dotenv-mini, csvquote, csvinfer, csvtable, csvtail, mdtable, mimedet, mimedb, hexdump, hexstr, morse, pigeon.

Matching, Text, URLs, And Shell Semantics

globmatch, wordwrap, stringcase, slugify-x, tinytemplate, tagged-template-ts, expand, pathmask, pathmatch-ts, shell-quote, shellexpand, urlnorm, phonenumber-mini, uuidgen, base62-ts.

Data Structures, Algorithms, And Math

decimal-ts, bitvec, trie, trie-ts, path-trie, ringbuf, rangeset, tinycache, randpick, numbertheory, chunkby, flatdict, dotpath, levendist.

Time, Numbers, Units, And Color

chronoparse, crontab-lite, timeago, durfmt, humanint, numparse, numfmt, unitcalc, semverlite, cidrcalc, colortool, colourdist.

TypeScript Building Blocks

result-ts, decoder-ts, argv-strict, argv-zod, parseopts-ts, tsparser, tsmemo, lru-ts, eventbus-ts, emitter-ts, tokenring-ts.

Product, CLI, And Lab Notes

markdownlint, diffstat, strtable, startup-dashboard, csv-explorer, cli-timer, json-differ, mlproject, nripankadas07.

Quality Bar

Every active repository is expected to have:

  • A specific README that says why the project exists and where it stops.
  • MIT license, security policy, contribution guide, code of conduct, changelog, roadmap, and quality notes.
  • Tests or an honest docs-only status.
  • Source-checkout installation instructions until package publication is real.
  • CI where a build or test surface exists.
  • Issue templates and a pull request template.
  • No fake package badges, fake benchmark numbers, or fake social proof.

For parsers and evaluators, correctness means adversarial inputs, conformance checks where possible, explicit limits, and typed failure modes. For TypeScript packages, correctness means typechecking, tests, build output, and package metadata that matches what is actually shipped.

Technical Essays

The flagship repositories include launch-ready technical notes:

Roadmap

  1. Keep the six flagships pinned and release-ready.
  2. Promote 15-25 core libraries with stronger examples, conformance tests, and packaging metadata.
  3. Consolidate redundant utilities instead of growing the repo count for its own sake.
  4. Mark weak or unclear projects as Needs Repair or Archive Candidate before deciding whether to keep them public.

Contact

Open an issue on the relevant repository for bugs, design questions, or focused collaboration. For profile-level context, use nripankadas07/nripankadas07.

Pinned Loading

  1. agent-framework agent-framework Public

    Tiny inspectable agent runtime with tools, memory, tracing, and safe no-key examples.

    Python

  2. rag-pipeline rag-pipeline Public

    Local-first RAG pipeline with chunking, retrieval, evaluation, and reports.

    Python

  3. bencode bencode Public

    Strict, dependency-free BitTorrent-style bencode encoder/decoder. Round-trips are exact: rejects every non-canonical encoding.

    Python

  4. decoder-ts decoder-ts Public

    Type-safe runtime JSON validators built from composable decoder combinators.

    TypeScript

  5. globmatch globmatch Public

    Zero-dependency fnmatch-compatible Python glob matcher with extglob, POSIX character classes, and ** globstar.

    Python

  6. randpick randpick Public

    Weighted random-sampling helpers — Vose's alias method, cumulative-bisect, Efraimidis-Spirakis A-Res. Zero dependencies.

    Python