Sanskrit Lexicon / CDSL Workspace

A multi-repo workspace for Sanskrit digital humanities, lexicography, grammar, pedagogy, and Russian-language scholarship.

Maintained by Dr. Mārcis Gasūns · sanskrit-lexicon organization


What this is

The Documents/GitHub/ root is not itself a git repository — it is a coordination layer that holds shared instruction files and ~196 standalone git repos cloned side by side. Each subdirectory is its own repo with its own README, dependencies, build system, and CI.

This workspace serves several interconnected missions:

  1. Sanskrit dictionary digitization — the Cologne Digital Sanskrit Lexicon (CDSL) ecosystem: source data, correction pipelines, comparative microstructural analysis, TEI/OntoLex standards, PHP and Python APIs, and spell-checking across dictionaries (MW, PW, PWG, PWK, AP90, etc.).
  2. Russian translations of Sanskrit dictionaries — translator-first dictionary projects (kosha) and Russian-translation tracks inside SanskritLexicography/.
  3. An online Sanskrit learning platform — Systema-Sanscriticum, a Laravel application deployed at samskrte.ru (with a sibling at samskrtam.ru).
  4. Sanskrit grammar and pedagogy — grammar materials (Bühler, Kochergina, Knauer), wave-notation verse visualizer (SanskritKaraoke), spoken/Telegram Sanskrit corpora.
  5. Indological scholarly archives — conference program archives (IndologyScholars), the Russian poetic translation of the Ramayana (RussianRamayana).
  6. Russian writing-style tools — RuWritingStyles, a PyPI-published CLI pipeline with eval harness and Web Studio.
  7. Agent/LLM harness tooling — Claude Code hooks/skills, Codex twins, a 16-step hyperresearch research pipeline, and cross-repo management infrastructure.

Coordination layer

RepoRole
github-spine/Org-level spine — holds shared CLAUDE.md, AGENTS.md, SHARED_CODE.md, SKILLS_INDEX.md, and template packs (PR, CI, CoC). Deployed to the workspace root so every agent session picks them up automatically.
Uprava/Cross-repo management hub — GTD rollup, project interlinks, findings, epistemic registries (assumptions, contradictions, gaps, dead ends, recipes, staleness, glossary), handoff store, articles dashboard, and the tools fleet. Private repo.
claude-config/Version-controlled backup of ~/.claude — hooks, commands, agents, skills, settings. Hooks enforce Markdown and convention rules deterministically.

The root-level CLAUDE.md is the org-level instruction file loaded by Claude Code for every repo under this workspace. It describes the hyperresearch pipeline and scholarly API integration.


Installation

There is no workspace-wide installer. Each subdirectory repo has its own setup.

Prerequisites

Cloning a specific project

git clone https://github.com/gasyoun/<repo-name>.git
# or from the sanskrit-lexicon org:
git clone https://github.com/sanskrit-lexicon/<repo-name>.git

Then follow that repo's own README.md for its specific installation steps.

Python projects (typical pattern)

cd <repo-name>
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt   # or: pip install -e . if pyproject.toml exists

Laravel project (Systema-Sanscriticum)

cd Systema-Sanscriticum
composer install
cp .env.example .env
php artisan key:generate

How to run

Sanskrit learning platform (Systema-Sanscriticum)

cd Systema-Sanscriticum
php artisan serve

Deployed in production at samskrte.ru.

Dictionary API (kosha)

cd kosha
python app/main.py

Static sites (GitHub Pages)

Live sites:

hyperresearch pipeline

# The CLI is installed via uv, not on PATH by default:
/Users/mac/.local/share/uv/tools/hyperresearch/bin/hyperresearch --help

# Check vault health
/Users/mac/.local/share/uv/tools/hyperresearch/bin/hyperresearch status -j

Running tests

cd <repo-name>
pytest                    # if pytest.ini or pyproject.toml configures it
python -m pytest tests/   # explicit

Usage examples

Cross-repo coordination via Uprava

cd Uprava
cat GTD_NEXT_ACTIONS.md              # what to do next
python tools/repo_count.py --count   # count repos in workspace
python tools/mint_handoff.py         # mint a new handoff ID
python tools/build_dashboard_data.py # refresh articles dashboard
python tools/pyfloor.py              # check Python interpreter floor

Searching the research vault

/Users/mac/.local/share/uv/tools/hyperresearch/bin/hyperresearch search "slp1 transliteration" --json
/Users/mac/.local/share/uv/tools/hyperresearch/bin/hyperresearch scholar search "sanskrit lexicography" --limit 25 -j
/Users/mac/.local/share/uv/tools/hyperresearch/bin/hyperresearch note list --json

Key configuration notes

FileWhat it does
CLAUDE.mdOrg-level Claude Code instruction file. Loaded automatically for every repo. Describes the hyperresearch pipeline, scholarly APIs, untrusted-content policy.
.mcp.jsonMCP server config. Defines a spider HTTP MCP server with bearer auth via SPIDER_API_KEY.
.zcodeignoreZCode IDE file-search ignore rules.
.hyperresearch/config.tomlhyperresearch pipeline configuration.

Environment variables: SPIDER_API_KEY is required for the spider MCP server. Individual repos have their own .env.example files.

Python interpreter floor: Python 3.9+ is the minimum, enforced by Uprava/tools/pyfloor.py and Uprava/tools/audit_pyfloor.py --check in CI.


Common questions

Is this a monorepo?

No. The root Documents/GitHub/ is not a git repository. Each subdirectory is a standalone git repo hosted on GitHub under gasyoun or the sanskrit-lexicon organization.

How do I find what to work on?

Read Uprava/GTD_NEXT_ACTIONS.md — the cross-repo GTD rollup with human actions (@DO / @DECIDE / @WAITING) and engineering resume-points.

How are agent instructions shared across repos?

The github-spine/ repo holds CLAUDE.md and AGENTS.md, deployed to the workspace root. Claude Code and Codex pick them up automatically for every repo. The always-on CLAUDE.md stays under a 3,000-token budget; the full skill catalogue lives in github-spine/SKILLS_INDEX.md.

Where does the Sanskrit learning platform live?

Systema-Sanscriticum/ — a Laravel application deployed at samskrte.ru.


Known limitations


Maintained by Dr. Mārcis Gasūns (gasyoun) · sanskrit-lexicon organization