Skip to content

README.md

Latest commit

 

History

History
130 lines (88 loc) · 4.89 KB

README.md

File metadata and controls

130 lines (88 loc) · 4.89 KB
Raw

tessera-llm-proxy

PyPI version Python

Drop-in cost optimization for LLM applications. One line of code patches your existing OpenAI / Anthropic / Mistral / DeepSeek / Groq / Together / Fireworks / OpenRouter / Perplexity / Cerebras / xAI client to route through Tessera's measurement + auto-optimize proxy. You keep your provider account and keys; we route + cache + compress and measure savings on every request.

Free Sandbox tier: 60M tokens / month · no card required · no fee until you upgrade. Get your free key at tesseraai.io/dev.

Get a free API key

→ Get free key (email + ToS, 30 seconds, no card)

After signup you get:

  • Your tk_ API key (shown once)
  • A magic-link for the dashboard — see your token counter + savings counter live
  • 60M tokens/month at 30 req/min — generous for hobby + side projects

Install

pip install tessera-llm-proxy

One-line setup

Drop this at the top of your application's entry point (main.py, app.py, manage.py, wherever your app boots):

import tessera
tessera.activate("tk_your_tessera_key")

That's it. Your existing code runs unchanged — openai.OpenAI(), anthropic.Anthropic(), mistralai.Mistral(), and other supported SDK constructors are transparently patched to route through Tessera. Your provider keys (OpenAI sk-..., Anthropic sk-ant-..., etc.) stay in your environment as before; Tessera forwards them upstream untouched.

Environment variable form

If you'd rather not put the key in code:

export TESSERA_KEY=tk_your_tessera_key
import tessera
tessera.activate()  # reads TESSERA_KEY from environment

What gets patched

Calling tessera.activate(...) patches the following SDKs at import time (each is opt-in: only patched if the library is installed):

SDK Tessera route
openai (≥1.0) https://api.tesseraai.io/v1/openai
anthropic https://api.tesseraai.io/v1/anthropic
mistralai https://api.tesseraai.io/v1/mistral
cohere https://api.tesseraai.io/v1/cohere (Wave 2)
groq https://api.tesseraai.io/v1/groq

If you use a framework that wraps these SDKs (LangChain, LlamaIndex, CrewAI, AutoGen, Mastra, Pydantic AI, etc.), the patch applies transparently because those frameworks call the underlying SDK constructors which are what we patched.

Direct provider URLs

If you call providers that aren't covered by an official Python SDK (DeepSeek, Together, Fireworks, OpenRouter, Perplexity, Cerebras, xAI), construct an openai.OpenAI client manually with the matching Tessera URL:

from openai import OpenAI

# DeepSeek via Tessera
client = OpenAI(
    api_key="sk-deepseek-...",
    base_url=tessera.url("deepseek"),  # → https://api.tesseraai.io/v1/deepseek
    default_headers=tessera.headers(),
)

# Same pattern for: together, fireworks, openrouter, perplexity, cerebras, groq, xai

tessera.url(provider) and tessera.headers() are pure helpers — no globals, no patching. Use them when you want explicit, traceable wiring.

Verification

import tessera
tessera.activate("tk_...")

assert tessera.is_active(), "Tessera should be active after activate()"
print(tessera.status())  # → ProxyStatus(active=True, providers_patched=['openai', 'anthropic'], proxy_base='https://api.tesseraai.io/v1')

Deactivation

To restore the original SDK constructors (e.g. in test teardown):

tessera.deactivate()

Configuration

tessera.activate(
    key="tk_...",                              # or TESSERA_KEY env var
    proxy_base="https://api.tesseraai.io/v1",  # default; override for staging/dev
    feature_tag="checkout-summarizer",         # attaches to every request for per-feature attribution
)

feature_tag lets you split savings reporting per workload (e.g. one tag per logical feature in your app). You can also set per-request tags by passing extra_body={"tessera_feature_tag": "..."} on individual SDK calls.

How the pricing works

Flat monthly subscription, priced by your gross monthly tokens submitted (before optimization). You keep 100% of the measured savings — savings is your ROI proof, not the billing basis.

Tier Gross tokens / month Price / month
Free Sandbox ≤ 60M $0 (no card required)
Starter ≤ 1B $199
Growth ≤ 5B $999
Scale ≤ 20B $3,999
Enterprise 20B+ Custom
  • Kill-switch available anytime from your portal — pauses optimization, traffic still flows passthrough.

Full terms: https://tesseraai.io/terms

License

Apache-2.0