Migrating from TrustwiseSDK to TrustwiseClient

Note

TrustwiseSDK is deprecated and will be removed in a future release. TrustwiseClient is a superset that covers all metrics, guardrails, policies, and the full platform API.

Why Migrate

Deprecation — TrustwiseSDK now emits a FutureWarning on every instantiation and will be removed in a future major version.
Unified client — TrustwiseClient provides agents, guardrails, policies, evaluations, risk classification, red-team generation, and metrics through a single client.
Server-side guardrails — The old SDK uses local threshold checking; the new client leverages platform-managed guardrails with provider integrations (AWS Bedrock, Trustwise, etc.).

What Changes

Before (`TrustwiseSDK`)	After (`TrustwiseClient`)
`from trustwise.sdk import TrustwiseSDK`	`from trustwise.sdk.platform import TrustwiseClient`
`TrustwiseSDK(config)`	`TrustwiseClient(config)`
`sdk.metrics.faithfulness.evaluate(...)`	`client.metrics.faithfulness(...)`
`sdk.guardrails(thresholds={...})` (removed)	`client.guardrails.evaluate(id, ...)` (server-side)
N/A	`client.agents`, `client.policies`, `client.evaluations`, etc.

Both clients use the same TrustwiseConfig, so the configuration step is unchanged.

Step-by-Step

1. Update Imports

- from trustwise.sdk import TrustwiseSDK
+ from trustwise.sdk.platform import TrustwiseClient

2. Replace Instantiation

- sdk = TrustwiseSDK(config)
+ client = TrustwiseClient(config)

3. Update Metric Calls

The TrustwiseClient metrics API uses direct keyword arguments instead of the .evaluate() wrapper:

- result = sdk.metrics.faithfulness.evaluate(
-     query="...", response="...", context=[...]
- )
+ result = client.metrics.faithfulness(
+     query="...", response="...", context=[...]
+ )

4. Use Server-Side Guardrails

The old local guardrail system (sdk.guardrails(...)) has been removed. Use platform-managed guardrails instead:

# Create a guardrail on the platform (once)
guardrail = client.guardrails.create(
    name="My Guardrail",
    evaluators=[
        {"evaluator": "faithfulness", "threshold": 0.8},
        {"evaluator": "toxicity", "threshold": 0.3},
    ],
)

# Evaluate against it
result = client.guardrails.evaluate(
    guardrail["id"], input="user message", output="assistant reply"
)

Complete Before / After

Before:

import os
from trustwise.sdk import TrustwiseSDK
from trustwise.sdk.config import TrustwiseConfig

config = TrustwiseConfig(api_key=os.environ["TW_API_KEY"])
sdk = TrustwiseSDK(config)

result = sdk.metrics.faithfulness.evaluate(
    query="What is the capital of France?",
    response="The capital of France is Paris.",
    context=[{"chunk_text": "Paris is the capital of France.", "chunk_id": "1"}],
)
print(result.score)

After:

import os
from trustwise.sdk.config import TrustwiseConfig
from trustwise.sdk.platform import TrustwiseClient

config = TrustwiseConfig(api_key=os.environ["TW_API_KEY"])
client = TrustwiseClient(config)

result = client.metrics.faithfulness(
    query="What is the capital of France?",
    response="The capital of France is Paris.",
    context=[{"chunk_text": "Paris is the capital of France.", "chunk_id": "1"}],
)
print(result.score)

# Plus: access agents, policies, guardrails, evaluations, etc.
agents = client.agents.list(limit=5)

FAQ

Can I use both clients side-by-side?: Yes. They share the same TrustwiseConfig and can coexist during migration. However, TrustwiseSDK will emit a FutureWarning each time it is instantiated.
Is there an async version of TrustwiseClient?: Not yet. TrustwiseClient is synchronous. An async counterpart is planned for a future release.
When will TrustwiseSDK be removed?: A removal date has not been set yet. Watch the Changelog for deprecation timeline updates.