Metadata-Version: 2.4
Name: tawjih
Version: 0.13.1
Summary: Direct a language model from trees of multi-sided knowledge files, with traceable paths, evaluation and human-approved AI assistance (English and Arabic).
Author: Mohammed Khalid Almodiny
License-Expression: CC-BY-NC-SA-4.0
Project-URL: Homepage, https://github.com/AlmodinyMk/LLM-tawjih-Model-building
Project-URL: Documentation, https://github.com/AlmodinyMk/LLM-tawjih-Model-building#readme
Project-URL: Arabic, https://github.com/AlmodinyMk/LLM-tawjih-Model-building/blob/main/README.ar.md
Project-URL: Paper, https://doi.org/10.5281/zenodo.24332793
Keywords: llm,prompt engineering,knowledge organization,provenance,evaluation,mcp,arabic,lora
Classifier: Development Status :: 3 - Alpha
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: Natural Language :: English
Classifier: Natural Language :: Arabic
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3 :: Only
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Text Processing
Requires-Python: >=3.11
Description-Content-Type: text/markdown
License-File: LICENSE
Dynamic: license-file

# tawjih — توجيه أوامر الآلة

A personal, file-based way to direct a language model. Knowledge is kept in
trees of small files; each file has a human side (the concept, for people),
a machine side (instructions for a model and optional Python code), a
reading side (passages from books, with their source) and optional worked
examples. A *path* strings files from any trees into a workflow; when it
runs, only the machine side is sent.

Python 3.11 or later, standard library only, Linux and Windows.
**Experimental:** the test suite passes, but no real AI provider has been
used with it yet.

```
pip install tawjih
tawjih init
tawjih new "Mathematics"
tawjih new "Algebra" --parent 1
tawjih tree
```

- A path language in English and Arabic: `TO`, `AND`, `IF … ELSE`, `AMONG`,
  `THROUGH`, `LINK`, `CALL`, with per-step `EACH`, `APPROVE`, `USING`,
  `RETRY … UNTIL`, `PICK`.
- Contracts between steps, staleness when sources change, a cache to resume.
- Evaluation: cases from the examples, a path compared with the model asked
  directly, baselines, datasets with SHA-256 manifests, lineage with `trace`.
- Governance: locks, review, staged proposals, sealed releases, schedules.
- `tawjih mcp`: an MCP server for AI assistants, where a person approves
  every write and every run.
- `tawjih do`: management statements in English or Arabic.

Full documentation, the example project, the Linux windows and the optional
training bridge are in the repository:
<https://github.com/AlmodinyMk/LLM-tawjih-Model-building>
([العربية](https://github.com/AlmodinyMk/LLM-tawjih-Model-building/blob/main/README.ar.md)).

![How Tawjih is built](https://raw.githubusercontent.com/AlmodinyMk/LLM-tawjih-Model-building/main/docs/images/architecture-en.png)

Licensed CC BY-NC-SA 4.0: noncommercial use with credit and the same
license for adaptations; commercial use needs permission from the author.
