feat: Multi user context

[monoxgas]

Support tracking and exporting spans with api key overrides to support multi-user services

Key Changes

• Reworks our span exporter to check for an override attribute and create a dedicated export instance for those spans - routing them per api key/user.
• Added an dn.run(..., api_key=) argument and with dn.using_api_key() to override this context on demand.
• Ensured that get_run_context carries this override information so if we have a backend stack passing around user-local context, it's reflected in the exports.

Example

import dreadnode as dn
import threading

OTHER_API_KEY = "..."


with dn.run("testing", project="context"):
    dn.log_input("x", 5)
    dn.log_output("z", 15)


def continue_run_in_thread(context):
    with dn.continue_run(context):
        dn.log_input("a", 10)
        dn.log_output("b", 20)

with dn.using_api_key(OTHER_API_KEY):
    with dn.run("other", project="context") as run:
        dn.log_input("y", 10)
        dn.log_output("y", 10)

        context = dn.get_run_context()
        
        thread = threading.Thread(target=continue_run_in_thread, args=(context,))
        thread.start()
        thread.join()

---

Generated Summary:

PR Description

This pull request introduces significant modifications to the SDK's tracing mechanism, enhancing support for multi-user scenarios by allowing the use of API keys for span exports.

• Refactored the get_run_context method to simplify the extraction of run context data.
• Introduced a new context manager, using_api_key, to facilitate temporary overriding of the API key used for exporting spans during nested operations.
• Added api_key as an optional parameter to run and initialize methods, enabling context-specific tracing.
• Replaced the BatchSpanProcessor with a new RoutingSpanProcessor that allows spans to be routed based on user-specific tokens.
• Updated the RunContext structure to include export_auth_token to maintain API key information across different contexts.
• Cleaned up unused imports and comments related to OpenTelemetry integrations.

These changes enable more flexible tracing capabilities and improve the SDK's usability in multi-user applications, allowing for controlled export configurations based on individual API keys.

This summary was generated with ❤️ by rigging

[Copilot]

Pull Request Overview

Introduces multi-user context support to enable tracking and exporting spans with API key overrides, allowing services to handle multiple users with dedicated export instances per API key.

• Added RoutingSpanProcessor to route spans based on user-specific API keys instead of using a single BatchSpanProcessor
• Introduced using_api_key() context manager and api_key parameter to run() for per-user span export control
• Enhanced RunContext to include export authentication token for cross-thread/process context continuity

Reviewed Changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
dreadnode/tracing/span.py	Added export auth token context management and updated RunSpan to support API key overrides
dreadnode/tracing/processors.py	New RoutingSpanProcessor for multi-user span routing based on API keys
dreadnode/tracing/constants.py	Added constant for span resource attribute token
dreadnode/main.py	Replaced BatchSpanProcessor with RoutingSpanProcessor and added using_api_key context manager
dreadnode/init.py	Exported using_api_key function
docs/sdk/main.mdx	Updated documentation to reflect new API key functionality

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Changing RunContext from a required TypedDict to total=False makes all fields optional, which could break existing code that assumes required fields like 'run_id' are always present. Consider using separate required and optional field groups or create a new TypedDict for the optional fields.

Suggested change

	class RunContext(te.TypedDict, total=False):
	"""Context for transferring and continuing runs in other places."""
	run_id: str
	run_name: str
	project: str
	trace_context: dict[str, str]
	export_auth_token: str | None
	class RunContextBase(te.TypedDict):
	"""Required fields for RunContext."""
	run_id: str
	run_name: str
	project: str
	trace_context: dict[str, str]
	class RunContext(RunContextBase, total=False):
	"""Context for transferring and continuing runs in other places, with optional fields."""
	export_auth_token: str | None

[Copilot]

The noqa comment suggests this is flagged as a security issue. Consider using a constant or configuration value instead of a hardcoded string literal for the API key header name.

[Copilot]

The example code uses dreadnode.with_api_key() but the actual function name is using_api_key(). This inconsistency will cause the example to fail.

[Copilot]

The documentation example uses with_api_key() but the actual function name is using_api_key(). This inconsistency will cause the example to fail.