fundkit 0.1.3

Async-first Python library for Mutual Fund data and analytics

FundKit

A modern, async-first Python library for Indian mutual fund data. FundKit fetches NAV and scheme information directly from AMFI - no third-party data vendors, no opaque middle layers.

If you're building dashboards, research pipelines, or backend services that need reliable fund data, FundKit gives you typed DataFrames with sensible caching out of the box.

import asyncio
from fundkit import NAVClient

async def main():
    async with NAVClient() as client:
        nav = await client.get_nav(128628)
        print(nav)

asyncio.run(main())

Installation

pip install fundkit
# or
uv add fundkit

Optional pandas support:

pip install fundkit[pandas]

What's available

The data module is ready to use today. The schema module provides typed Pydantic models for structured responses. Planned modules - portfolio, analytics, tax, sip, compare, and more - are still in progress.

Client	Purpose
NAVClient	Today's NAV - lookup by scheme code, name, AMC, or fund type
HistoricalNAVClient	NAV history for a date range
SchemeDetailsClient	Scheme metadata - category, launch date, minimum investment, and more

All clients return Polars DataFrames by default (except SchemeDetailsClient.get_scheme_details, which returns a typed SchemeDetails model). Pass df_format="pandas" if you prefer pandas. They also share helper methods for scheme validation and discovery - see the data module docs for the full list.

Usage

import asyncio
from datetime import date
from fundkit import NAVClient, HistoricalNAVClient, SchemeDetailsClient

async def main():
    async with NAVClient(verbose=True) as client:
        # Single or multiple scheme codes
        nav = await client.get_nav(128628)
        batch = await client.get_nav([119597, 120505, 108272])

        # Search by name, AMC, or fund type
        by_name = await client.get_nav_by_name("bluechip", case_sensitive=False)
        by_amc = await client.get_nav_by_amc("SBI")
        by_type = await client.get_nav_by_type("Open Ended Schemes")

        # Utilities
        valid = await client.is_valid_scheme_code(119597)
        schemes = await client.get_scheme_codes(query="bluechip", by="scheme_name")
        amcs = await client.get_amc_list()

    async with HistoricalNAVClient(verbose=True) as client:
        history = await client.get_history(
            124182,
            start_date=date(2023, 1, 1),
            end_date=date.today(),
            df_format="pandas",  # optional - defaults to polars
        )

    async with SchemeDetailsClient(verbose=True) as client:
        details = await client.get_scheme_details(128628)
        bulk = await client.get_scheme_details_bulk([128628, 119597])

asyncio.run(main())

Set verbose=True on any client to print cache hits, fetches, and other log messages to the console.

For architecture, caching details, output schemas, and a full method reference, see src/fundkit/data/README.md.

How FundKit compares

Most Python libraries in this space follow a similar pattern: a synchronous class that wraps a third-party API (like mfapi.in), returns dicts or pandas DataFrames, and re-fetches on every call.

That works fine for one-off scripts, but it gets painful in async web apps or when you're filtering thousands of schemes repeatedly.

FundKit takes a different approach:

	Typical MF data libraries	FundKit
Data source	Third-party API proxy	AMFI directly
Execution model	Synchronous (requests)	Async-first (httpx + asyncio)
Default output	Dict / JSON / pandas	Typed Polars DataFrame
Bulk filtering	Load everything, filter in Python	Vectorized Polars operations
Caching	None or in-memory only	Memory -> disk (parquet) -> network
Typed schemas	Untyped dicts	Consistent column names and dtypes

FundKit doesn't try to replace every feature these libraries offer (daily performance metrics, JSON export, MCP servers, etc.) - at least not yet. The focus right now is a solid, fast, cache-aware data layer you can build on top of.

Caching

FundKit caches data locally so repeated calls in the same day don't hammer AMFI.

Cache location is platform native:

Platform	Path
Linux	~/.cache/fundkit/
macOS	~/Library/Caches/fundkit/
Windows	%LOCALAPPDATA%\fundkit\

Data	TTL
Latest NAV	Same calendar day (memory + disk)
Historical NAV	Permanent, append-only per AMC
Scheme details	7 days (memory + disk)

Historical data is never re-fetched once cached - past NAV is immutable. Latest NAV refreshes automatically when the calendar day changes.

Why FundKit

• Async-first - built on httpx and asyncio; fits naturally into FastAPI, Django async views, or any modern async stack
• AMFI-native - one less dependency that can go down or change its API without warning
• Polars-first - filtering ~15k schemes is significantly faster than row-by-row Python; pandas export is one argument away
• Typed and structured - consistent column names (scheme_code, nav, date, …) whether you're querying one scheme or ten thousand