prysmai 0.1.0

Prysm AI — Observability SDK for LLM applications. One-line integration for OpenAI, Anthropic, and any OpenAI-compatible provider.

Prysm AI — Python SDK

Observability for LLM applications. One line of code.

Prysm wraps your existing OpenAI client and routes every request through the Prysm proxy, capturing latency, token usage, cost, errors, and full request/response payloads — with zero changes to your application logic.

---

Installation

pip install prysm

Quick Start

import openai
from prysm import monitor

# Your existing OpenAI client
client = openai.OpenAI(api_key="sk-...")

# Wrap it with Prysm — that's it
monitored = monitor(client, prysm_key="sk-prysm-...")

# Every call is now tracked
response = monitored.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Explain quantum computing"}],
)

print(response.choices[0].message.content)

Your dashboard at app.prysmai.io now shows the request with full metrics: latency, tokens, cost, model, and the complete request/response.

---

How It Works

The SDK creates a new OpenAI client that points at the Prysm proxy instead of the OpenAI API directly. The proxy:

1. Authenticates the request using your sk-prysm-* API key
2. Forwards the request to OpenAI (or any configured provider) using your project's stored credentials
3. Captures the full request, response, timing, token counts, and cost
4. Returns the response to your application — unchanged

Your application code stays exactly the same. The only difference is the client instance.

Your App  →  Prysm Proxy  →  OpenAI API
              ↓
         Metrics stored
         (latency, tokens,
          cost, errors)

---

API Reference

monitor(client, prysm_key, base_url, timeout)

The primary entry point. Wraps an existing OpenAI client.

Parameter	Type	Default	Description
client	openai.OpenAI or openai.AsyncOpenAI	required	Your existing OpenAI client
prysm_key	str	PRYSM_API_KEY env var	Your Prysm API key (sk-prysm-...)
base_url	str	https://proxy.prysmai.io/v1	Prysm proxy URL
timeout	float	120.0	Request timeout in seconds

Returns: A new OpenAI client of the same type (sync or async) routed through Prysm.

# Sync
monitored = monitor(openai.OpenAI(api_key="sk-..."), prysm_key="sk-prysm-...")

# Async
monitored = monitor(openai.AsyncOpenAI(api_key="sk-..."), prysm_key="sk-prysm-...")

PrysmClient(prysm_key, base_url, timeout)

Lower-level client for more control.

from prysm import PrysmClient

prysm = PrysmClient(prysm_key="sk-prysm-...")

# Create sync client
client = prysm.openai()

# Create async client
async_client = prysm.async_openai()

prysm_context — Request Metadata

Attach metadata (user ID, session ID, custom tags) to every request for filtering and grouping in your dashboard.

from prysm import prysm_context

# Set globally
prysm_context.set(user_id="user_123", session_id="sess_abc")

# Or use scoped context
with prysm_context(user_id="user_456", metadata={"env": "production"}):
    response = monitored.chat.completions.create(...)
    # This request is tagged with user_456

# Outside the block, context reverts to user_123

Method	Description
prysm_context.set(user_id, session_id, metadata)	Set global context for all subsequent requests
prysm_context.get()	Get the current context object
prysm_context.clear()	Reset context to defaults
prysm_context(user_id, session_id, metadata)	Use as a context manager for scoped metadata

---

Environment Variables

The SDK reads these environment variables as fallbacks:

Variable	Description
PRYSM_API_KEY	Your Prysm API key (used if prysm_key is not passed)
PRYSM_BASE_URL	Custom proxy URL (used if base_url is not passed)

export PRYSM_API_KEY="sk-prysm-your-key-here"

from prysm import monitor
import openai

# No need to pass prysm_key — reads from env
monitored = monitor(openai.OpenAI(api_key="sk-..."))

---

Streaming

Streaming works exactly as you'd expect — no changes needed:

stream = monitored.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Write a haiku about AI"}],
    stream=True,
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")

The proxy captures Time to First Token (TTFT), total latency, and full streamed content.

---

Async Support

Full async support with the same API:

import asyncio
import openai
from prysm import monitor

async def main():
    client = openai.AsyncOpenAI(api_key="sk-...")
    monitored = monitor(client, prysm_key="sk-prysm-...")

    response = await monitored.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": "Hello async!"}],
    )
    print(response.choices[0].message.content)

asyncio.run(main())

---

Self-Hosted Proxy

If you're running the Prysm proxy on your own infrastructure:

monitored = monitor(
    client,
    prysm_key="sk-prysm-...",
    base_url="http://localhost:3000/v1",  # Your self-hosted proxy
)

---

What Gets Captured

Every request through the SDK is logged with:

Metric	Description
Model	Which model was called (gpt-4o, gpt-4o-mini, etc.)
Latency	Total request duration in milliseconds
TTFT	Time to first token (streaming requests)
Prompt tokens	Input token count
Completion tokens	Output token count
Cost	Calculated cost based on model pricing
Status	Success, error, or timeout
Request body	Full messages array and parameters
Response body	Complete model response
User ID	From prysm_context (if set)
Session ID	From prysm_context (if set)
Custom metadata	Any key-value pairs from prysm_context

---

Error Handling

The SDK preserves OpenAI's error types. If the upstream API returns an error, you get the same exception you'd get without Prysm:

try:
    response = monitored.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": "test"}],
    )
except openai.AuthenticationError:
    print("Invalid API key")
except openai.RateLimitError:
    print("Rate limited")
except openai.APIError as e:
    print(f"API error: {e}")

Prysm-specific errors (invalid Prysm key, proxy unreachable) also surface as standard OpenAI exceptions so your existing error handling works unchanged.

---

Requirements

• Python 3.9+
• openai >= 1.0.0
• httpx >= 0.24.0

---

Development

git clone https://github.com/osasisorae/prysmai.git
cd prysmai/sdk

# Install with dev dependencies
pip install -e ".[dev]"

# Run tests
pytest tests/ -v

Test Coverage

The SDK includes 41 tests covering:

• Client initialization and validation
• Environment variable fallbacks
• Sync and async client creation
• monitor() function behavior
• Context management (global, scoped, nested)
• Header injection via custom transports
• Full integration tests with mock HTTP server
• Error propagation (401, 500)

---

License

MIT — see LICENSE for details.

---

Built by Prysm AI — See inside your AI.