⚡ amfi-stream

Streaming-first ingestion for AMFI mutual fund data.

Turn raw AMFI files into clean, schema-safe, analytics-ready Arrow tables — in parallel, without hacks.

---

🚀 The problem (you already know this)

AMFI data is:

• inconsistent
• semi-structured
• painful to clean
• not pipeline-friendly

Every existing tool assumes:

“just fetch and parse it”

That breaks at scale.

---

⚡ The shift

Don’t query AMFI. Ingest it properly.

---

🧩 What amfi-stream does

Raw AMFI Data
(NAV + Scheme Files)
        ↓
⚡ amfi-stream
(stream + sanitize + normalize)
        ↓
Arrow Tables (typed, clean)
        ↓
Polars / DuckDB / Pandas / Spark

---

✨ Why people switch

• ⚡ Streaming instead of batch downloads
• 🧼 Automatic normalization (no manual cleaning)
• 🧱 Strong schema via Apache Arrow
• 🧵 Parallel ingestion engine
• 📊 Directly usable in analytics tools
• 🐼 No Pandas dependency

---

🆚 Alternatives (quick reality check)

Tool	Model	Why it breaks
mfapi.in	API calls	One request per fund → slow
navpipe	SDK	Needs pre-known fund list
mftool	Scraper	Fragile, breaks silently
AMFI site	Raw files	No structure

amfi-stream:

✔ Dataset-level ingestion
✔ Streaming + parallel
✔ Schema enforced
✔ Built for pipelines

---

⚡ Quick start

from amfi_stream import (
    AMFIPipeline,
    stream_latest_nav,
    stream_scheme_master,
    stream_historical_nav
)

jobs = [
    stream_scheme_master(),
    stream_latest_nav(),
    stream_historical_nav("1-May-2025", "1-May-2026")
]

with AMFIPipeline(max_workers=4) as pipeline:
    result = pipeline.run(jobs)

print(result.latest_nav)

---

📦 Output

AMFIResult(
    scheme_master=pa.Table | None,
    latest_nav=pa.Table | None,
    historical_nav=pa.Table | None,
)

Typed. Predictable. Analytics-ready.

---

🏗 Architecture

URLs
 ↓
Streaming Engine
 ↓
Sanitizer
 ↓
Parser
 ↓
Arrow Tables
 ↓
Normalizers
 ↓
Pipeline Output

---

🔥 Design principles

• Streaming > batch
• Schema > guesswork
• Arrow > DataFrame conversions
• Deterministic > fragile parsing
• Minimal > bloated

---

🔮 Coming soon

• Derived analytics-ready columns
• Enhanced schema layers
• Faster historical ingestion

---

🤝 Contributing

If you’ve ever fought AMFI data, you already know why this exists.

Open areas:

• Performance tuning
• Enhanced schema creation
• Benchmark comparison
• Tests
• Documentation and docstrings

---

⭐ If this helped you

Give it a star — it helps more people discover a better way to handle AMFI data.

---

📜 License

Apache 2.0