Python Integration

Pāṇini is available on PyPI as panini-lang. It's the recommended interface for researchers, linguists, and developers who prefer the Python ecosystem.

🔧 Installation

pip install panini-lang

🚀 Basic Usage

The panini module provides a high-level interface to the Rust extraction engine.

Synchronous Extraction

import panini

# Extraction for a single word
result = panini.extract(
    provider="google",
    model="gemini-1.5-flash",
    api_key="...",
    language="pol",
    text="Studentka czyta książkę.",
    targets=["studentka"]
)

# Iterating over morphological features
for feat in result['morphology']['target_features']:
    m = feat['morphology']
    print(f"Word: {feat['word']}, Lemma: {m['lemma']}, POS: {m['pos']}")

Sample Output:

Word: studentka, Lemma: studentka, POS: noun

Asynchronous Extraction

import asyncio
import panini

async def analyze():
    res = await panini.async_extract(
        provider="openai",
        model="gpt-4o",
        api_key="...",
        language="ara",
        text="ذهب الولد إلى المدرسة",
        targets=["ذهب"]
    )
    return res

asyncio.run(analyze())

Sample Async Result:

{
  "morphology": {
    "target_features": [
      {
        "word": "ذهب",
        "morphology": { "root": "ذ-ه-ب", "wazn": "fa'ala", "tense": "past" }
      }
    ]
  }
}

🧬 Extraction Parameters

The extract() and async_extract() functions share the same parameters:

Parameter	Type	Default	Description
`provider`	`str`	(Required)	`"openai"`, `"anthropic"`, or `"google"`.
`model`	`str`	(Required)	Model ID (e.g., `"gpt-4o"`).
`api_key`	`str`	(Required)	Provider-specific API key.
`language`	`str`	(Required)	ISO 639-3 target language code.
`text`	`str`	(Required)	The sentence or text to analyze.
`targets`	`list[str]`	(Required)	Specific words to focus extraction on.
`components`	`list[str]`	`None`	(Optional) List of component keys to run.
`temperature`	`float`	`0.2`	Sampling temperature.

🎨 Prompt Customization

You can pass the prompts parameter as a path to a YAML file or as a Python dictionary.

# 1. Get default prompts
custom_prompts = panini.get_default_prompts()

# 2. Modify a specific section
custom_prompts["system_role"] = "You are an expert linguist."

# 3. Use in extraction
res = panini.extract(..., prompts=custom_prompts)

Custom Prompt Effect: The LLM will prioritize the new system role (e.g., "Expert Linguist"), leading to more detailed pedagogical explanations in the results.

🌍 Supported Languages

To see which languages are currently registered in the engine:

print(panini.supported_languages())
# => ["pol", "tur", "ara", "fra", ...]

Rust/Python Bridge Performance

The framework uses pythonize to convert Rust types directly into Python objects without intermediate JSON stringification, ensuring optimal performance.