Open-source · Python · FastAPI · MIT

Trading Signal API
for Python Developers

Send a price series, get a structured BUY / SELL / HOLD signal — with confidence score, risk score, and explainable reasoning. No ML expertise required.

Works with any price series — crypto, stocks, commodities, synthetic data.

View on GitHub Run locally ↓

Deterministic — same input, same output Zero network calls at inference No external data dependency Runs fully local

Input → Output

Request

curl -X POST http://localhost:8000/v1/predict \
  -H "X-API-Key: dev-key-1234" \
  -H "Content-Type: application/json" \
  -d '{"prices": [100, 101.5, 99.8,
                  102.3, 103.1, 101.9]}'

Response

{
  "action": "HOLD",
  "confidence": 0.51,
  "expected_value": -0.01,
  "risk_score": 0.18,
  "signal_strength": 0.01,
  "reasoning_signals": [
    "positive momentum (+1.90%)",
    "price above MA5 (+0.55%)"
  ]
}

How it works

prices[ ] → feature extraction → logistic regression → signal + reasoning

Extracts momentum, volatility, and moving-average features from the last N prices
Model trains once on first startup and is persisted to disk — subsequent starts are instant

Why developers use this

Algorithmic trading bots

Call the API in a loop — no ML pipeline to build or maintain.

Financial dashboards

Surface structured signals alongside raw price data.

Strategy prototyping

Get a working signal endpoint in under 5 minutes.

Run locally in 60 seconds

Clone and install

git clone https://github.com/your-username/PIaaS
cd PIaaS && pip install -r requirements.txt

Start the server

uvicorn main:app --reload

Call the API — default key works immediately, no config needed

curl -X POST http://localhost:8000/v1/predict \
  -H "X-API-Key: dev-key-1234" \
  -H "Content-Type: application/json" \
  -d '{"prices": [100, 101.5, 99.8, 102.3, 103.1, 101.9]}'

Or use the Python SDK

from sdk.client import PIaaSClient

client = PIaaSClient("http://localhost:8000", api_key="dev-key-1234")
signal = client.predict([100, 101.5, 99.8, 102.3, 103.1, 101.9])

if signal["action"] == "BUY" and signal["confidence"] > 0.65:
    place_order(size=compute_position(signal["risk_score"]))

Response fields

Field	Type	Description
`action`	BUY · SELL · HOLD	Recommended decision
`confidence`	0.0 – 1.0	Model confidence in the prediction
`expected_value`	float	P(up) − P(down)
`risk_score`	0.0 – 1.0	Volatility-based risk estimate
`signal_strength`	0.0 – 1.0	Distance from neutral (0 = no signal, 1 = strong)
`reasoning_signals`	list[str]	Human-readable explanation of the decision

Why this is different

deterministic Identical inputs always produce identical outputs — no stochastic inference

local-first Zero network calls at inference time — runs entirely in your process

transparent Logistic regression, not a black box — every signal is traceable to a feature

self-contained No data broker, no API key, no rate limit from a third party

What this is not

Not financial advice — outputs are developer signals, not investment recommendations
Not a prediction guarantee — confidence scores reflect the model, not market certainty
Not a managed service — designed to run in your own infrastructure

Ready to add trading signals to your project?

No setup required beyond pip install. Works in under 60 seconds.

View on GitHub Run locally ↑