📡 LLM Radar

Real-time observability dashboard for LLM applications. Track every prompt, token count, cost, and latency across OpenAI and Anthropic — with one line of code.

from llm_radar import LLMRadar
radar = LLMRadar(app)  # that's it

Dashboard → http://localhost:8000/__llm_radar

Installation

pip install llm-radar

With provider SDKs:

pip install "llm-radar[openai]"       # + openai
pip install "llm-radar[anthropic]"    # + anthropic
pip install "llm-radar[all]"          # + both

Quick Start

OpenAI

from fastapi import FastAPI
from llm_radar import LLMRadar
import openai

app = FastAPI()
radar = LLMRadar(app)          # intercepts all openai calls automatically

client = openai.OpenAI()

@app.get("/chat")
async def chat(message: str):
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": message}],
    )
    return {"reply": response.choices[0].message.content}

Anthropic

from fastapi import FastAPI
from llm_radar import LLMRadar
import anthropic

app = FastAPI()
radar = LLMRadar(app)

client = anthropic.Anthropic()

@app.get("/summarize")
async def summarize(text: str):
    response = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=256,
        messages=[{"role": "user", "content": text}],
    )
    return {"summary": response.content[0].text}

Use with fastapi-radar

llm-radar works alongside fastapi-radar — one app, two dashboards.

from fastapi import FastAPI
from fastapi_radar import Radar
from llm_radar import LLMRadarPlugin

app = FastAPI()

radar = Radar(app)               # HTTP + SQL monitoring → /__radar/
llm   = LLMRadarPlugin(app)      # LLM monitoring       → /__llm_radar

What Gets Tracked

Configuration

radar = LLMRadar(
    app,
    dashboard_path="/__llm_radar",   # Custom path
    max_calls=1000,                   # Max records to keep
    retention_hours=24,               # Data retention window
    db_path="/var/data/llm",          # Custom DB location
    auth_dependency=my_auth_fn,       # Optional FastAPI dependency
    track_openai=True,
    track_anthropic=True,
)

Securing the Dashboard

from fastapi import Depends, HTTPException
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials

security = HTTPBearer()

def verify_token(creds: HTTPAuthorizationCredentials = Depends(security)):
    if creds.credentials != "your-secret-token":
        raise HTTPException(status_code=401, detail="Unauthorized")

radar = LLMRadar(app, auth_dependency=verify_token)

Supported Models & Pricing

Auto-detects cost for:

• OpenAI: gpt-4o, gpt-4o-mini, gpt-4-turbo, gpt-3.5-turbo, o1, o3-mini, o4-mini
• Anthropic: claude-opus-4, claude-sonnet-4, claude-haiku-4, claude-3.5-sonnet, claude-3-opus

Unrecognized models record 0 cost (no crash).

Contributing

git clone https://github.com/ganeshmandakapu/llm-radar
cd llm-radar
pip install -e ".[dev]"

License

MIT — Ganesh Mandakapu