Skip to content

Pāṇini

A LLM-powered linguistic feature extraction & analysis framework.

Pāṇini allows you to describe a language's morphology as Rust types, write extraction directives, and let the pipeline handle the rest: prompt assembly, JSON schema generation, LLM orchestration, response parsing, and validation.

  • Linguistic Agnosticism

    No universal schema imposed. Each language defines exactly the features it needs.

  • Type-Safe Guarantee

    Morphology validated at compile-time. Auto-generated JSON schemas. Every LLM response is verified against the target schema, with automatic retry and self-correction logic to ensure result integrity.

  • Multi-Language Access

    High-performance Rust core, accessible via CLI or Python package (panini-lang).

  • Corpus Analysis

    Transform raw extractions into statistical coverage reports and lexical inventories.

🚀 Panini in 1 Minute

1. Describe Your Language

Model your grammar using standard Rust types. Your types become the "source of truth" for the AI.

pub enum PolishMorphology {
    Noun { lemma: String, gender: Gender, case: Case },
    Verb { lemma: String, aspect: Aspect, tense: Tense },
}

Generated JSON Schema (Fragment): 

{
  "anyOf": [
    {
      "type": "object",
      "properties": {
        "pos": { "const": "noun" },
        "lemma": { "type": "string" },
        "gender": { "$ref": "#/$defs/Gender" },
        "case": { "$ref": "#/$defs/Case" }
      }
    }
  ]
}

2. Extract!

Use the CLI or Python package to analyze any text.

import panini
result = panini.extract(
    language="pol",
    text="Studentka czyta książkę.",
    targets=["studentka"]
)

Sample Extraction Result: 

{
  "morphology": {
    "target_features": [
      {
        "word": "studentka",
        "morphology": {
          "pos": "noun",
          "lemma": "studentka",
          "gender": "feminine",
          "case": "nominative"
        }
      }
    ]
  }
}

3. Analyze the Lexicon

Generate a statistical analysis of the corpus to track coverage and frequency.

Extraction Output Sample (Polish): 

{
  "word": "studentka",
  "morphology": {
    "pos": "noun",
    "lemma": "studentka",
    "gender": "feminine",
    "number": "singular",
    "case": "nominative"
  }
}

Aggregation Report Sample: 

[NOUN] total: 15
  |- case [3/7]: nominative(8), accusative(5), genitive(2)
  |- gender [2/3]: feminine(10), masculine(5)

📚 Where to Start?

About the Name

Pāṇini is named after Pāṇini, the ancient Indian grammarian and author of the Aṣṭādhyāyī, the first systematic and formal description of the Sanskrit language.

Note: Pāṇini Framework is a technical tool for feature extraction and does not strictly follow the specific "Paninian Framework" notation used in traditional NLP (e.g., Pāṇinian Syntactico-Semantic Relation Labels).