Prabhu Agarwal
11/26/2025, 2:12 PMsource_declarative_manifest.componentskapa.ai
11/26/2025, 2:12 PMsource_declarative_manifest.components) that either: 1. Is not present / not built correctly in your connector image, or 2. Has a missing dependency, so the module import fails. From the GitHub issue you attached, the same error appeared when a custom components file used pendulum, which was not included in the CDK / connector dependencies, so the component module failed to load. [[Prestashop issue](https://github.com/airbytehq/airbyte/issues/55693)] For Connector Builder custom components specifically: - Custom components are Python classes you ship with your connector and reference by fully qualified class name in the YAML. [[Custom components](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/custom-components)] - They extend the Low‑Code CDK; you must switch to the YAML editor and correctly reference them. [[How to use](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/custom-components#how-to-use-custom-components)] Given your symptoms: - The async reporting stream (with custom component) works. - Other streams (not using the custom component directly) fail with Could not load module source_declarative_manifest.components. Most likely: 1. Your manifest for those streams still references a component in source_declarative_manifest.components (e.g. a shared paginator, retrier, authenticator, etc.), and 2. That module either: - Doesn’t exist in the built image, or - Imports something not available in the environment (similar to the pendulum issue). [[Prestashop issue](https://github.com/airbytehq/airbyte/issues/55693)] What to check: 1. In the YAML (Builder → YAML mode), search for any class_name / type references pointing to source_declarative_manifest.components... in the non‑reporting streams. 2. Make sure those classes actually exist in your Python package and are importable. 3. Remove or refactor any usage that relies on libraries not available in the connector runtime (as was done in the Prestashop fix by rewriting components to avoid pendulum). [[Prestashop issue](https://github.com/airbytehq/airbyte/issues/55693)] 4. If you’re running Self‑Managed and using custom components, ensure the unsafe code flag is enabled, or the server may block loading custom Python: set AIRBYTE_ENABLE_UNSAFE_CODE=true on workload-launcher and connector-builder-server via values.yaml and redeploy. [[Enable custom components](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/custom-components#how-to-enable-custom-components)] If after confirming the class references and dependencies you still see the same error, the knowledge sources don’t provide more detailed troubleshooting steps. In that case, the next action would be to download the connector logs from the Builder UI and inspect the full Python traceback to see exactly which import inside source_declarative_manifest.components is failing.Prabhu Agarwal
11/26/2025, 2:24 PMkapa.ai
11/26/2025, 2:25 PMkapa.ai
11/26/2025, 2:25 PMsource_declarative_manifest.components (the default module name used when you have custom components in a declarative/manifest-style source), and it’s impacting all streams, not only the one that uses the custom component. From the docs and related issues, there are a few key points to check: 1. Custom components affect the whole connector, not just one stream When you enable custom components in a declarative connector, the Python module containing those components must be valid and importable for the entire connector, even for streams that don’t use those custom classes directly. If that module fails to import (e.g., syntax error, missing dependency), you’ll see “Could not load module `source_declarative_manifest.components`” and all streams can fail. [[Custom components](https://docs.airbyte.com/platform/connector-development/config-based/advanced-topics/custom-components)] 2. Check that the module is importable and has no missing dependencies In a similar GitHub issue, the connector failed with the same message because the components file imported pendulum, which wasn’t available in the runtime; fixing the file to avoid that dependency resolved the error. [[Prestashop issue](https://github.com/airbytehq/airbyte/issues/55693)] Concretely: - Ensure your custom Python code does not import libraries that are not bundled with the connector runtime. - Make sure there are no syntax errors or top‑level code that might throw at import time. 3. Verify all YAML references to custom components In Builder UI, you must switch to the YAML view and: - Confirm that any type: "Custom..." and class_name: "..." entries refer to the correct, fully qualified class name in the module you defined. [[How to use custom components](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/custom-components#how-to-use-custom-components)] - Check that other streams don’t accidentally reference a non‑existent class in source_declarative_manifest.components (for example, shared paginator / retriever / transformation names copied from the working stream). 4. Ensure unsafe code is enabled (self‑managed) If you’re on a self‑managed / Core deployment, custom components are gated behind AIRBYTE_ENABLE_UNSAFE_CODE. If that’s not enabled for workload-launcher and connector-builder-server, custom code may not be loaded correctly. [[Enable custom components](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/custom-components#how-to-enable-custom-components)] Given the knowledge sources, I can’t see deeper builder‑2.0–specific debugging steps. If after: - verifying the Python file imports cleanly, - removing/avoiding external libraries not shipped with the runtime, and - double‑checking all YAML class_name references, you still get the same error on non‑custom streams, the next step is to inspect the full stack trace from the Builder UI logs to see which exact import in source_declarative_manifest.components is failing. The docs and issues don’t provide more detailed guidance beyond that.