fix(low-code): send HttpComponentsResolver partition router request options and support AsyncRetriever

[Baz (bazarnov)]

Summary

create_http_components_resolver builds the resolver retriever with only the pre-#707 stream_slicer= kwarg. Since #707 (v7.0.0), create_simple_retriever takes the partition router as partition_router and uses it as the request options provider, so:

1. A request_option on the resolver retriever's partition router is silently dropped.
2. An AsyncRetriever resolver retriever, which the schema allows, fails discovery with KeyError: 'requester' (the use_cache override). Past that, the resolver iterates the partition router instead of the job slices and would treat the slice log message as component values.

Changes

• create_http_components_resolver builds the partition router once and passes it as stream_slicer and partition_router, like create_default_stream. The resolver iterates the same instance, or the AsyncRetriever's job slicer.
• HttpComponentsResolver.resolve_components skips AirbyteMessage items (the AsyncRetriever slice log), as DynamicSchemaLoader does since fix(low-code): support AsyncRetriever in DynamicSchemaLoader #974.
• ConcurrentDeclarativeSource and the legacy ManifestDeclarativeSource set use_cache only on a resolver retriever that has a requester. Async job requesters stay uncached, so polling never replays a stale job status.

Compatibility

• Monorepo: 3 connectors use HttpComponentsResolver (source-airtable, source-google-sheets, source-hubspot), all with a SimpleRetriever and no request_option on its partition router. Their resolver requests (method, URL, headers, body) are identical on main and on this branch, and the resolved streams are the same.
• Manifests that declare a request_option on the resolver retriever's partition router send it again, as before v7.0.0. If the requester also sets the same key, combine_mappings now raises on the duplicate, and ignore_stream_slicer_parameters_on_paginated_requests now applies.
• An AsyncRetriever resolver is not cached, so each resolution creates a job. check and the Connector Builder resolve twice per process, so they create two jobs, as DynamicSchemaLoader with an AsyncRetriever does (fix(low-code): support AsyncRetriever in DynamicSchemaLoader #974).

Tests

In resolvers/test_http_components_resolver.py, end to end through ConcurrentDeclarativeSource and HttpMocker:

• test_dynamic_streams_with_http_components_resolver_partition_router_request_option: the resolver request carries ?p=p1.
• test_dynamic_streams_with_http_components_resolver_async_retriever: creates, polls and downloads a job. The resolved streams are item_1 and item_2, with none from the slice log.

Both fail on main. A mutation of each change on the ConcurrentDeclarativeSource path fails at least one of them. The legacy use_cache guard is the same one-line change and has no test of its own. unit_tests/sources/declarative/{resolvers,parsers,retrievers,schema} and unit_tests/legacy: 717 passed.

Release note

Dynamic-stream resolvers send their partition router's request_option again, as before 7.0.0. If a resolver's requester hard-codes the same key in request_parameters, request_headers or request_body_json, remove it from the requester.

Fixes #1181

[github-actions]

👋 Greetings, Airbyte Team Member!

Here are some helpful tips and reminders for your convenience.

💡 Show Tips and Tricks

Testing This CDK Version

You can test this version of the CDK using the following:

# Run the CLI from this branch:
uvx 'git+https://github.com/airbytehq/airbyte-python-cdk.git@bazarnov/fix-http-components-resolver-partition-router#egg=airbyte-python-cdk[dev]' --help

# Update a connector to use the CDK from this branch ref:
cd airbyte-integrations/connectors/source-example
poe use-cdk-branch bazarnov/fix-http-components-resolver-partition-router

PR Slash Commands

Airbyte Maintainers can execute the following slash commands on your PR:

• /autofix - Fixes most formatting and linting issues
• /poetry-lock - Updates poetry.lock file
• /test - Runs connector tests with the updated CDK
• /prerelease - Triggers a prerelease publish with default arguments
• /poe build - Regenerate git-committed build artifacts, such as the pydantic models which are generated from the manifest JSON schema in YAML.
• /poe <command> - Runs any poe command in the CDK environment

📚 Show Repo Guidance

Helpful Resources

• CDK API Reference

📝 Edit this welcome message.

[github-actions]

PyTest Results (Fast)

4 972 tests  +2   4 960 ✅ +2   10m 51s ⏱️ + 2m 21s
    1 suites ±0      12 💤 ±0 
    1 files   ±0       0 ❌ ±0 

Results for commit 17571bf. ± Comparison against base commit c756dce.

♻️ This comment has been updated with latest results.

[github-actions]

PyTest Results (Full)

4 975 tests  +2   4 963 ✅ +2   15m 55s ⏱️ -1s
    1 suites ±0      12 💤 ±0 
    1 files   ±0       0 ❌ ±0 

Results for commit 17571bf. ± Comparison against base commit c756dce.

♻️ This comment has been updated with latest results.

[Daryna Ishchenko (darynaishchenko)]

Approving. The wiring matches create_default_stream, and I probed 8 resolver shapes at this head vs main: the three monorepo users (airtable, google-sheets, hubspot — Substream/no router, no request_option) send identical requests and resolve the same streams; router request_options are sent per partition again as in 6.x; an AsyncRetriever resolver creates/polls/downloads a job where main raises KeyError: 'requester'; failed/timeout jobs during discover raise a traced system_error and free their tracker slots. Both new tests fail on main, and 4 of 5 production hunks are killed by a mutation (the legacy guard is the known exception).

One inline note on the release surface for the key-collision case — it is a scan plus a sentence, not a code change.