Cruxy Python SDK

Type-safe Python client for the Cruxy API — an OpenAI-compatible LLM gateway covering chat completions, embeddings, reranking, moderations, token counting, files, and device login.

Distributed on PyPI as cruxy-api; imported as cruxy.

from cruxy import Cruxy

with Cruxy(bearer_auth="cxy_live_...") as cruxy:
    res = cruxy.chat.create_chat_completion(
        model="kavi",
        messages=[{"role": "user", "content": "Hello!"}],
    )

Every operation is generated in both sync and async form (create_chat_completion / create_chat_completion_async).

Every request carries Cruxy-Version: 2026-06-13 unless you set one, so responses keep naming the tiers kavi, vaani and mira, whatever the gateway's default for an unversioned request. A version you pass always wins, per call or as a header:

cruxy.chat.create_chat_completion(model="kavi", messages=msgs, cruxy_version="2026-10-18")
cruxy.models.list_models(http_headers={"Cruxy-Version": "2026-10-18"})

From 2026-10-18 responses name the tiers canto, saga and epos.

This is a generated SDK

Everything under src/cruxy/ is generated by Speakeasy — do not hand-edit it; your changes will be erased on the next generation. Change an input and regenerate instead.

spec/openapi.snapshot.json  ──┐
                              ├─ speakeasy run ─► .speakeasy/out.openapi.yaml ─► src/cruxy/**
overlays/cruxy-overlay.yaml ──┘        (merged spec, committed)

Path	Hand-authored?	Role
spec/openapi.snapshot.json	yes (pinned)	Frozen copy of the upstream OpenAPI document
overlays/cruxy-overlay.yaml	yes	Patches defects in that document
.speakeasy/out.openapi.yaml	no	Merged spec; committed, diffed by CI
src/cruxy/**	no	Generated SDK

Generation reads the committed snapshot, not the live API, so an upstream spec change can never silently reshape a released client. CI enforces this with a keyless, deterministic drift gate; a separate manual workflow diffs the snapshot against the live spec and surfaces moves as an explicit update-snapshot PR. See CONTRIBUTING.md.

The overlay is transitional

overlays/cruxy-overlay.yaml exists only because the upstream document is generated by swag and carries defects the SDK cannot consume — untyped oneOf request/response bodies, a broken multipart upload body, missing operation IDs, and Go package name prefixes (httpx., llm.) on every schema.

Each block names the defect it works around and when it can go. As the api-repo fixes land, delete the corresponding block rather than carrying it forward. The overlay should shrink over time, not accumulate.

Streaming

Streaming chat is not part of the generated surface — Speakeasy's SSE support is gated to a paid tier, and its model (one operation, one data schema) cannot express this wire anyway. Non-streaming chat generates normally; the ChatV2Stream{Start,Delta,Finish,Error} frame models are retained via x-speakeasy-include so the hand-authored layer in src/cruxy/streaming.py has types to deserialize into. Details in CONTRIBUTING.md.

The wire is a sequence of named events — one start, then any number of delta, then exactly one terminal finish or error:

import os
from cruxy import ChatStreamFrameError, Cruxy, IncompleteChatStreamError

with Cruxy(bearer_auth=os.environ["CRUXY_API_KEY"]) as cruxy:
    stream = cruxy.chat.stream(
        model="auto",
        messages=[{"role": "user", "content": "Explain SSE in two sentences."}],
    )
    try:
        for delta in stream:            # only `delta` frames are yielded
            if delta.content:
                print(delta.content, end="", flush=True)
    except ChatStreamFrameError as err:
        # Terminal `error` frame: the completion failed AFTER the 200 was
        # committed. `code` is stable and `retryable` is a real signal.
        if err.retryable:
            ...
    except IncompleteChatStreamError:
        # Neither terminal frame arrived. A v2 stream ALWAYS terminates
        # explicitly, so this is a FAILED completion, not a short one.
        ...

    print(stream.routing.tier)          # from `start` — which tier served it
    print(stream.finish_reason, stream.usage.input_tokens)   # from `finish`

stream_async() is the async counterpart (await it, then async for).

Only delta frames are yielded. The others populate the stream object: start fills id / model / routing / request_id / memory_recall / rules_applied, and the terminal finish fills finish_reason / usage / memory_capture. Both iterators also reassemble tool_calls by index across frames, exposed as stream.tool_calls. An event name this SDK version does not know is ignored, so a server that adds frames does not break an older client. Runnable examples live in examples/.

A stream that ends without a terminal finish or error frame raises IncompleteChatStreamError — it never ends iteration silently. That is the single most important contract in this layer; treat the completion as failed and retry. There is no [DONE] sentinel on this wire, and a v1-style one is explicitly not a terminator.

Pass an idempotency_key when you intend to act on that advice. On a stream it buys retry-safe billing, not replay: the gateway derives the request id from the key on every chat request and the usage ledger is uniquely indexed on it, so a retried stream is billed once. It does not make the stream replayable — verbatim SSE replay is deliberately excluded server-side, so there is no stored response to serve and no 409 on a same-key/different-body reuse.

Retries are opt-in and apply to the stream-open phase only (a 429/5xx before any frame has been delivered). A stream that fails once frames are flowing is never re-sent, so a retry cannot duplicate content — that is structural, not policy.

---

Summary

Cruxy API: OpenAI-compatible LLM gateway for the Cruxy model tiers (kavi, vaani, mira), plus embeddings, moderations, and device login. Streaming chat is served as OpenAI-style Server-Sent Events. Cruxy-native fields are additive and namespaced under a cruxy.* key so a vanilla OpenAI client keeps parsing.

This spec is GENERATED from the Go handlers at build time (make openapi) and covers the PUBLIC surface only; admin/internal endpoints are excluded by design.

Table of Contents

• Cruxy Python SDK
  • This is a generated SDK
  • SDK Installation
  • IDE Support
  • SDK Example Usage
  • Authentication
  • Available Resources and Operations
  • File uploads
  • Retries
  • Error Handling
  • Server Selection
  • Custom HTTP Client
  • Resource Management
  • Debugging
• Development
  • Maturity
  • Contributions

SDK Installation

The SDK can be installed with uv, pip, or poetry package managers.

uv

uv is a fast Python package installer and resolver, designed as a drop-in replacement for pip and pip-tools. It's recommended for its speed and modern Python tooling capabilities.

uv add git+<UNSET>.git

PIP

PIP is the default package installer for Python, enabling easy installation and management of packages from PyPI via the command line.

pip install git+<UNSET>.git

Poetry

Poetry is a modern tool that simplifies dependency management and package publishing by using a single pyproject.toml file to handle project metadata and dependencies.

poetry add git+<UNSET>.git

Shell and script usage with uv

You can use this SDK in a Python shell with uv and the uvx command that comes with it like so:

uvx --from cruxy-api python

It's also possible to write a standalone Python script without needing to set up a whole project like so:

#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "cruxy-api",
# ]
# ///

from cruxy import Cruxy

sdk = Cruxy(
  # SDK arguments
)

# Rest of script here...

Once that is saved to a file, you can run it with uv run script.py where script.py can be replaced with the actual file name.

IDE Support

PyCharm

Generally, the SDK will work well with most IDEs out of the box. However, when using PyCharm, you can enjoy much better integration with Pydantic by installing an additional plugin.

• PyCharm Pydantic Plugin

SDK Example Usage

Example

# Synchronous Example
from cruxy import Cruxy
import os


with Cruxy(
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:

    res = c_client.chat.create_chat_completion(max_tokens=1024, messages=[
        {
            "content": "Hello",
            "role": "user",
            "tool_call_id": "call_abc123",
            "tool_calls": [
                {
                    "function": {
                        "arguments": "{\"city\":\"Paris\"}",
                        "name": "get_weather",
                    },
                    "id": "call_abc123",
                    "type": "function",
                },
            ],
        },
    ], model="kavi", reasoning_effort="medium", response_format={
        "json_schema": {
            "name": "person",
            "strict": True,
        },
        "type": "json_schema",
    }, stream=True, temperature=0.7, tools=[
        {
            "function": {
                "description": "Get the current weather for a city",
                "name": "get_weather",
            },
            "type": "function",
        },
    ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50")

    # Handle response
    print(res)

The same SDK client can also be used to make asynchronous requests by importing asyncio.

# Asynchronous Example
import asyncio
from cruxy import Cruxy
import os

async def main():

    async with Cruxy(
        bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
    ) as c_client:

        res = await c_client.chat.create_chat_completion_async(max_tokens=1024, messages=[
            {
                "content": "Hello",
                "role": "user",
                "tool_call_id": "call_abc123",
                "tool_calls": [
                    {
                        "function": {
                            "arguments": "{\"city\":\"Paris\"}",
                            "name": "get_weather",
                        },
                        "id": "call_abc123",
                        "type": "function",
                    },
                ],
            },
        ], model="kavi", reasoning_effort="medium", response_format={
            "json_schema": {
                "name": "person",
                "strict": True,
            },
            "type": "json_schema",
        }, stream=True, temperature=0.7, tools=[
            {
                "function": {
                    "description": "Get the current weather for a city",
                    "name": "get_weather",
                },
                "type": "function",
            },
        ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50")

        # Handle response
        print(res)

asyncio.run(main())

Authentication

Per-Client Security Schemes

This SDK supports the following security scheme globally:

Name	Type	Scheme	Environment Variable
bearer_auth	http	HTTP Bearer	CRUXY_BEARER_AUTH

To authenticate with the API the bearer_auth parameter must be set when initializing the SDK client instance. For example:

from cruxy import Cruxy
import os


with Cruxy(
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:

    res = c_client.chat.create_chat_completion(max_tokens=1024, messages=[
        {
            "content": "Hello",
            "role": "user",
            "tool_call_id": "call_abc123",
            "tool_calls": [
                {
                    "function": {
                        "arguments": "{\"city\":\"Paris\"}",
                        "name": "get_weather",
                    },
                    "id": "call_abc123",
                    "type": "function",
                },
            ],
        },
    ], model="kavi", reasoning_effort="medium", response_format={
        "json_schema": {
            "name": "person",
            "strict": True,
        },
        "type": "json_schema",
    }, stream=True, temperature=0.7, tools=[
        {
            "function": {
                "description": "Get the current weather for a city",
                "name": "get_weather",
            },
            "type": "function",
        },
    ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50")

    # Handle response
    print(res)

Available Resources and Operations

Available methods

Auth

• get_current_identity - Current identity

Chat

• create_chat_completion - Chat completions (native v2)

DeviceLogin

• start_device_login - Start device login
• poll_device_token - Poll device login

Embeddings

• create_embeddings - Create embeddings

Feedback

• submit_feedback - Submit feedback

Files

• list_files - List files
• upload_file - Upload a file
• delete_file - Delete a file
• get_file - Get file metadata
• download_file_content - Download file content

Models

• list_models - List model tiers

Moderations

• create_moderation - Classify content (moderations)

Rerank

• rerank_documents - Rerank documents

Tokens

• count_tokens - Count input tokens

File uploads

Certain SDK methods accept file objects as part of a request body or multi-part request. It is possible and typically recommended to upload files as a stream rather than reading the entire contents into memory. This avoids excessive memory consumption and potentially crashing with out-of-memory errors when working with very large files. The following example demonstrates how to attach a file stream to a request.

from cruxy import Cruxy
import os


with Cruxy(
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:

    res = c_client.files.upload_file(file={
        "file_name": "example.file",
        "content": open("example.file", "rb"),
    })

    # Handle response
    print(res)

Retries

Some of the endpoints in this SDK support retries. If you use the SDK without any configuration, it will fall back to the default retry strategy provided by the API. However, the default retry strategy can be overridden on a per-operation basis, or across the entire SDK.

To change the default retry strategy for a single API call, simply provide a RetryConfig object to the call:

from cruxy import Cruxy
from cruxy.utils import BackoffStrategy, RetryConfig
import os


with Cruxy(
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:

    res = c_client.chat.create_chat_completion(max_tokens=1024, messages=[
        {
            "content": "Hello",
            "role": "user",
            "tool_call_id": "call_abc123",
            "tool_calls": [
                {
                    "function": {
                        "arguments": "{\"city\":\"Paris\"}",
                        "name": "get_weather",
                    },
                    "id": "call_abc123",
                    "type": "function",
                },
            ],
        },
    ], model="kavi", reasoning_effort="medium", response_format={
        "json_schema": {
            "name": "person",
            "strict": True,
        },
        "type": "json_schema",
    }, stream=True, temperature=0.7, tools=[
        {
            "function": {
                "description": "Get the current weather for a city",
                "name": "get_weather",
            },
            "type": "function",
        },
    ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50",
        RetryConfig("backoff", BackoffStrategy(1, 50, 1.1, 100), False))

    # Handle response
    print(res)

If you'd like to override the default retry strategy for all operations that support retries, you can use the retry_config optional parameter when initializing the SDK:

from cruxy import Cruxy
from cruxy.utils import BackoffStrategy, RetryConfig
import os


with Cruxy(
    retry_config=RetryConfig("backoff", BackoffStrategy(1, 50, 1.1, 100), False),
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:

    res = c_client.chat.create_chat_completion(max_tokens=1024, messages=[
        {
            "content": "Hello",
            "role": "user",
            "tool_call_id": "call_abc123",
            "tool_calls": [
                {
                    "function": {
                        "arguments": "{\"city\":\"Paris\"}",
                        "name": "get_weather",
                    },
                    "id": "call_abc123",
                    "type": "function",
                },
            ],
        },
    ], model="kavi", reasoning_effort="medium", response_format={
        "json_schema": {
            "name": "person",
            "strict": True,
        },
        "type": "json_schema",
    }, stream=True, temperature=0.7, tools=[
        {
            "function": {
                "description": "Get the current weather for a city",
                "name": "get_weather",
            },
            "type": "function",
        },
    ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50")

    # Handle response
    print(res)

Error Handling

CruxyBaseError is the base class for all HTTP error responses. It has the following properties:

Property	Type	Description
err.message	str	Error message
err.status_code	int	HTTP response status code eg 404
err.headers	httpx.Headers	HTTP response headers
err.body	str	HTTP body. Can be empty string if no body is returned.
err.raw_response	httpx.Response	Raw HTTP response
err.data		Optional. Some errors may contain structured data. See Error Classes.

Example

from cruxy import Cruxy, errors
import os


with Cruxy(
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:
    res = None
    try:

        res = c_client.chat.create_chat_completion(max_tokens=1024, messages=[
            {
                "content": "Hello",
                "role": "user",
                "tool_call_id": "call_abc123",
                "tool_calls": [
                    {
                        "function": {
                            "arguments": "{\"city\":\"Paris\"}",
                            "name": "get_weather",
                        },
                        "id": "call_abc123",
                        "type": "function",
                    },
                ],
            },
        ], model="kavi", reasoning_effort="medium", response_format={
            "json_schema": {
                "name": "person",
                "strict": True,
            },
            "type": "json_schema",
        }, stream=True, temperature=0.7, tools=[
            {
                "function": {
                    "description": "Get the current weather for a city",
                    "name": "get_weather",
                },
                "type": "function",
            },
        ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50")

        # Handle response
        print(res)


    except errors.CruxyBaseError as e:
        # The base class for HTTP error responses
        print(e.message)
        print(e.status_code)
        print(e.body)
        print(e.headers)
        print(e.raw_response)

        # Depending on the method different errors may be thrown
        if isinstance(e, errors.CruxyError):
            print(e.data.code)  # Optional[models.CruxyErrorCode]
            print(e.data.error)  # Optional[str]
            print(e.data.request_id)  # Optional[str]

Error Classes

Primary errors:

• CruxyBaseError: The base class for HTTP error responses.
  • CruxyError: *

Less common errors (6)

Network errors:

• httpx.RequestError: Base class for request errors.
  • httpx.ConnectError: HTTP client was unable to make a request to a server.
  • httpx.TimeoutException: HTTP request timed out.

Inherit from CruxyBaseError:

• DeviceTokenPendingResponseError: authorization_pending | access_denied | expired_token | invalid_grant. Status code 400. Applicable to 1 of 15 methods.*
• ResponseValidationError: Type mismatch between the response data and the expected Pydantic model. Provides access to the Pydantic validation error via the cause attribute.

* Check the method documentation to see if the error is applicable.

Server Selection

Override Server URL Per-Client

The default server can be overridden globally by passing a URL to the server_url: str optional parameter when initializing the SDK client instance. For example:

from cruxy import Cruxy
import os


with Cruxy(
    server_url="https://api.cruxy.in",
    bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
) as c_client:

    res = c_client.chat.create_chat_completion(max_tokens=1024, messages=[
        {
            "content": "Hello",
            "role": "user",
            "tool_call_id": "call_abc123",
            "tool_calls": [
                {
                    "function": {
                        "arguments": "{\"city\":\"Paris\"}",
                        "name": "get_weather",
                    },
                    "id": "call_abc123",
                    "type": "function",
                },
            ],
        },
    ], model="kavi", reasoning_effort="medium", response_format={
        "json_schema": {
            "name": "person",
            "strict": True,
        },
        "type": "json_schema",
    }, stream=True, temperature=0.7, tools=[
        {
            "function": {
                "description": "Get the current weather for a city",
                "name": "get_weather",
            },
            "type": "function",
        },
    ], vault_id="6b1f9c1e-6d1a-4f0e-9f5a-2b7c1d3e4f50")

    # Handle response
    print(res)

Custom HTTP Client

The Python SDK makes API calls using the httpx HTTP library. In order to provide a convenient way to configure timeouts, cookies, proxies, custom headers, and other low-level configuration, you can initialize the SDK client with your own HTTP client instance. Depending on whether you are using the sync or async version of the SDK, you can pass an instance of HttpClient or AsyncHttpClient respectively, which are Protocol's ensuring that the client has the necessary methods to make API calls. This allows you to wrap the client with your own custom logic, such as adding custom headers, logging, or error handling, or you can just pass an instance of httpx.Client or httpx.AsyncClient directly.

For example, you could specify a header for every request that this sdk makes as follows:

from cruxy import Cruxy
import httpx

http_client = httpx.Client(headers={"x-custom-header": "someValue"})
s = Cruxy(client=http_client)

or you could wrap the client with your own custom logic:

from cruxy import Cruxy
from cruxy.httpclient import AsyncHttpClient
import httpx

class CustomClient(AsyncHttpClient):
    client: AsyncHttpClient

    def __init__(self, client: AsyncHttpClient):
        self.client = client

    async def send(
        self,
        request: httpx.Request,
        *,
        stream: bool = False,
        auth: Union[
            httpx._types.AuthTypes, httpx._client.UseClientDefault, None
        ] = httpx.USE_CLIENT_DEFAULT,
        follow_redirects: Union[
            bool, httpx._client.UseClientDefault
        ] = httpx.USE_CLIENT_DEFAULT,
    ) -> httpx.Response:
        request.headers["Client-Level-Header"] = "added by client"

        return await self.client.send(
            request, stream=stream, auth=auth, follow_redirects=follow_redirects
        )

    def build_request(
        self,
        method: str,
        url: httpx._types.URLTypes,
        *,
        content: Optional[httpx._types.RequestContent] = None,
        data: Optional[httpx._types.RequestData] = None,
        files: Optional[httpx._types.RequestFiles] = None,
        json: Optional[Any] = None,
        params: Optional[httpx._types.QueryParamTypes] = None,
        headers: Optional[httpx._types.HeaderTypes] = None,
        cookies: Optional[httpx._types.CookieTypes] = None,
        timeout: Union[
            httpx._types.TimeoutTypes, httpx._client.UseClientDefault
        ] = httpx.USE_CLIENT_DEFAULT,
        extensions: Optional[httpx._types.RequestExtensions] = None,
    ) -> httpx.Request:
        return self.client.build_request(
            method,
            url,
            content=content,
            data=data,
            files=files,
            json=json,
            params=params,
            headers=headers,
            cookies=cookies,
            timeout=timeout,
            extensions=extensions,
        )

s = Cruxy(async_client=CustomClient(httpx.AsyncClient()))

Resource Management

The Cruxy class implements the context manager protocol and registers a finalizer function to close the underlying sync and async HTTPX clients it uses under the hood. This will close HTTP connections, release memory and free up other resources held by the SDK. In short-lived Python programs and notebooks that make a few SDK method calls, resource management may not be a concern. However, in longer-lived programs, it is beneficial to create a single SDK instance via a context manager and reuse it across the application.

from cruxy import Cruxy
import os
def main():

    with Cruxy(
        bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
    ) as c_client:
        # Rest of application here...


# Or when using async:
async def amain():

    async with Cruxy(
        bearer_auth=os.getenv("CRUXY_BEARER_AUTH", ""),
    ) as c_client:
        # Rest of application here...

Debugging

You can setup your SDK to emit debug logs for SDK requests and responses.

You can pass your own logger class directly into your SDK.

from cruxy import Cruxy
import logging

logging.basicConfig(level=logging.DEBUG)
s = Cruxy(debug_logger=logging.getLogger("cruxy"))

You can also enable a default debug logger by setting an environment variable CRUXY_DEBUG to true.

Development

Maturity

This SDK is in beta, and there may be breaking changes between versions without a major version update. Therefore, we recommend pinning usage to a specific package version. This way, you can install the same version each time without breaking changes unless you are intentionally looking for the latest version.

Contributions

While we value open-source contributions to this SDK, this library is generated programmatically. Any manual changes added to internal files will be overwritten on the next generation. We look forward to hearing your feedback. Feel free to open a PR or an issue with a proof of concept and we'll do our best to include it in a future release.

SDK Created by Speakeasy