REASONBLOCKS_API_KEY from your server secret store with
your existing Anthropic or OpenAI SDK. The same connection supports recording and
approved serving; you do not need to deploy a different client after training.
Keep your provider client and credential. Native Bedrock keeps IAM
authentication in the existing boto3 client.
See Quickstart to create the connection.
Application files
Directly configured Anthropic and OpenAI clients need no ReasonBlocks helper package. Keep their source URL and headers in application configuration. ForConnection(source_id="SOURCE_ID"), deploy the application code and
rbtrace==1.3.1; no generated config file is needed. Set REASONBLOCKS_API_KEY
at runtime. Pin the package version used in your tests.
Existing 1.3.0 deployments remain compatible; upgrading to 1.3.1 adds native
Bedrock dispatch diagnostics and requires no connection or credential changes.
The rest of this section applies to existing generated helpers.
Deploy the generated reasonblocks_setup.py and its adjacent
.reasonblocks/config.json with the application. Include the hidden config
directory explicitly in wheel package data or image copy rules. The generated
.reasonblocks/SETUP.md is optional at runtime.
Install rbtrace==1.3.1 from PyPI into the application’s Python environment
and record the version in its dependency manifest. That release contains the
CLI and the helper’s automatic run labelling.
Use an importable helper location appropriate to your package. Verify the real
entrypoint from a working directory outside the repository. See
helper placement.
Runtime secrets
For direct clients orConnection, supply REASONBLOCKS_API_KEY through the deployment’s secret environment.
For an existing generated helper, use REASONBLOCKS_CAPTURE_KEY.
The existing RB_CAPTURE_KEY alias is accepted; if both names are set, their
values must agree. Keep the provider credential in its normal secret location.
Do not put either key in generated files or the image.
The capture service receives model requests and the provider credential needed
to forward them. Native Bedrock capture keeps AWS authentication local. Review the
workflow content you send. New server keys have no scheduled expiration by default;
optional expirations and existing keys still expire as shown in API Keys.
Legacy source-key rotation provides a 24-hour overlap, bounded by the old key’s expiry.
Tasks and verification
Wire the helper into the process that makes the actual model calls. Create one run ID per top-level user turn, from the first model call through tool execution and the final answer. Start a new ID for the next turn, including in workers that handle many tasks. See Integrating your agent. Rundoctor in the deployed Python environment with the configured helper path.
It is a local check, not a remote credential or capture test. Confirm source
records after a real workflow before claiming the deployment is capturing correctly.
With Connection, call rb.status() or check the live status in Manage connection.
Review recorded requests separately from training examples and exclusions.
Normal provider charges apply to those calls.