fix(collections): Use correct file batch ID - #1073
Conversation
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughOpenAI ingestion now separates batch creation from polling and uses the creation batch ID for failed-file lookup. Collection batches retry timeouts, uploads validate content, and XLS/XLSX formats are registered. ChangesVector batch ingestion
Document upload validation
Collection batch retries
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant execute_batch_job
participant OpenAIProvider
participant OpenAIVectorStoreCrud
participant OpenAI
execute_batch_job->>OpenAIProvider: upload files and create batch
OpenAIProvider->>OpenAIVectorStoreCrud: update(vector_store_id, file_ids)
OpenAIVectorStoreCrud->>OpenAI: create file batch
OpenAI-->>OpenAIVectorStoreCrud: batch_id
OpenAIVectorStoreCrud->>OpenAI: retrieve batch until terminal
OpenAI-->>OpenAIVectorStoreCrud: status and file counts
Possibly related PRs
Suggested labels: Suggested reviewers: 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✨ Finishing Touches 💡 1📝 Generate docstrings 💡
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
github-actions
Bot
changed the title
OpenAPI changes ⚪ No API surface changesNote This PR does not modify the API contract.
|
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
🧹 Nitpick comments (2)
backend/app/tests/crud/rag/test_open_ai.py (2)
72-72: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winRemove the self-evident comment.
assert_not_called()already expresses this expectation; retain comments only when they explain rationale. As per coding guidelines, “Write comments and docstrings to explain why something is done, not what the code already shows.”🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/tests/crud/rag/test_open_ai.py` at line 72, Remove the self-evident “list_files should not have been called on the happy path” comment near the assert_not_called() expectation in the test, leaving the assertion unchanged.Source: Coding guidelines
44-44: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winAdd required type annotations to changed test signatures.
Annotate non-
selffixture parameters and return values consistently.
backend/app/tests/crud/rag/test_open_ai.py#L44-L44: annotatemock_client: MagicMock.backend/app/tests/crud/rag/test_open_ai.py#L75-L75: annotate injected fixtures and add-> None.backend/app/tests/crud/rag/test_open_ai.py#L106-L108: annotate injected fixtures and add-> None.backend/app/tests/crud/rag/test_open_ai.py#L122-L124: annotate injected fixtures and add-> None.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/tests/crud/rag/test_open_ai.py` at line 44, Add the requested annotations in backend/app/tests/crud/rag/test_open_ai.py: type _wire_batch’s mock_client parameter as MagicMock, and annotate all non-self injected fixture parameters plus return values with -> None in the test functions at lines 75, 106-108, and 122-124. Apply the annotations consistently across all four affected sites.Source: Coding guidelines
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@backend/app/tests/crud/rag/test_open_ai.py`:
- Line 72: Remove the self-evident “list_files should not have been called on
the happy path” comment near the assert_not_called() expectation in the test,
leaving the assertion unchanged.
- Line 44: Add the requested annotations in
backend/app/tests/crud/rag/test_open_ai.py: type _wire_batch’s mock_client
parameter as MagicMock, and annotate all non-self injected fixture parameters
plus return values with -> None in the test functions at lines 75, 106-108, and
122-124. Apply the annotations consistently across all four affected sites.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 79f5c90e-9ea2-43f6-8dc9-ff715cb54c65
📒 Files selected for processing (6)
backend/app/crud/rag/open_ai.pybackend/app/tests/crud/rag/test_open_ai.pybackend/app/tests/services/collections/providers/test_openai_provider.pybackend/app/tests/utils/llm_provider.pybackend/app/tests/utils/openai.pydocs/wiki/modules/knowledge-base.md
…in document upload process
There was a problem hiding this comment.
🧹 Nitpick comments (2)
backend/app/services/documents/helpers.py (2)
1-6: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winModernize deprecated
typingimports.
Callable/Iterableshould come fromcollections.abc, andTupleis deprecated in favor of builtintuple(PEP 585). Ruff flags both (UP035).♻️ Proposed fix
-from typing import Callable, Optional, Tuple, Iterable, Union +from collections.abc import Callable, Iterable +from typing import Optional, UnionNote: other
Tuple[...]usages later in the file (e.g._read_edges,validate_document_content... wait, that one already uses lowercaseTuplereturn) should be updated totuple[...]for consistency once the import changes.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/services/documents/helpers.py` around lines 1 - 6, Update the imports in this module so Callable and Iterable come from collections.abc, remove Tuple from typing, and replace every Tuple[...] annotation in the file with the builtin tuple[...] syntax; preserve Optional and Union unless independently required.Source: Linters/SAST tools
53-57: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winDuplicated magic byte signatures between
MISLABELLED_BINARY_SIGNATURESandFORMAT_SPECS;"doc"has noFormatSpec.
b"PK\x03\x04"andb"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1"are inlined in both structures. Extracting them (and the zip-signature triple used for both docx and xlsx) into named constants would remove the duplication.Separately,
"doc"is absent fromFORMAT_SPECSeven though it shares the OLE2 signature documented for"xls"(comment at line 246 explicitly calls out ".doc"), andregistry.pymaps.doc→"doc". As written,.docuploads get zero content sniffing while.xlsuploads (same binary container format) do.♻️ Proposed fix
+OLE2_SIGNATURE = b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1" +ZIP_SIGNATURES = (b"PK\x03\x04", b"PK\x05\x06", b"PK\x07\x08") + MISLABELLED_BINARY_SIGNATURES: dict[bytes, str] = { - b"PK\x03\x04": "an Office/zip file (xlsx, docx)", - b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1": "a legacy Office file (xls, doc)", + ZIP_SIGNATURES[0]: "an Office/zip file (xlsx, docx)", + OLE2_SIGNATURE: "a legacy Office file (xls, doc)", b"%PDF-": "a PDF", } ... "docx": FormatSpec( - signatures=(b"PK\x03\x04", b"PK\x05\x06", b"PK\x07\x08"), checker=_check_docx + signatures=ZIP_SIGNATURES, checker=_check_docx ), "xlsx": FormatSpec( - signatures=(b"PK\x03\x04", b"PK\x05\x06", b"PK\x07\x08"), checker=_check_xlsx + signatures=ZIP_SIGNATURES, checker=_check_xlsx ), - "xls": FormatSpec(signatures=(b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1",)), + "xls": FormatSpec(signatures=(OLE2_SIGNATURE,)), + "doc": FormatSpec(signatures=(OLE2_SIGNATURE,)),Also applies to: 237-250
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/services/documents/helpers.py` around lines 53 - 57, Extract the shared PK/zip and OLE2 byte signatures into named constants, then reuse them in MISLABELLED_BINARY_SIGNATURES and the relevant FORMAT_SPECS entries, including the existing zip-signature triple for docx and xlsx. Add a “doc” FORMAT_SPECS entry matching the “xls” OLE2 signature and appropriate shared sniffing behavior so .doc uploads are handled like .xls.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@backend/app/services/documents/helpers.py`:
- Around line 1-6: Update the imports in this module so Callable and Iterable
come from collections.abc, remove Tuple from typing, and replace every
Tuple[...] annotation in the file with the builtin tuple[...] syntax; preserve
Optional and Union unless independently required.
- Around line 53-57: Extract the shared PK/zip and OLE2 byte signatures into
named constants, then reuse them in MISLABELLED_BINARY_SIGNATURES and the
relevant FORMAT_SPECS entries, including the existing zip-signature triple for
docx and xlsx. Add a “doc” FORMAT_SPECS entry matching the “xls” OLE2 signature
and appropriate shared sniffing behavior so .doc uploads are handled like .xls.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 6e6e3f5f-23a5-4f47-96b3-cfc57b2d0caa
📒 Files selected for processing (3)
backend/app/api/routes/documents.pybackend/app/services/doctransform/registry.pybackend/app/services/documents/helpers.py
… processing and error management
There was a problem hiding this comment.
Actionable comments posted: 5
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
backend/app/services/documents/helpers.py (1)
123-128: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winReject OOXML packages that lack the expected part.
A PPTX or generic OOXML ZIP has
[Content_Types].xmlbut neitherword/norxl/, so it currently passes validation as both DOCX and XLSX. Requireexpected_partto be present; only useother_partto improve the mismatch message.Proposed fix
- if other_part in sample and expected_part not in sample: + if expected_part not in sample: raise DocumentValidationError( filename, - f"parsing error - file is a '{other_part.decode()}' Office package, " - "not the format its extension claims", + ( + f"parsing error - file is a '{other_part.decode()}' Office package, " + "not the format its extension claims" + if other_part in sample + else "parsing error - Office package does not match its extension" + ), )🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/services/documents/helpers.py` around lines 123 - 128, Update the validation condition around the Office package checks so a package is rejected whenever expected_part is absent, including cases where neither expected_part nor other_part exists. Preserve the existing DocumentValidationError and use other_part only when available to identify the detected package type in the mismatch message; otherwise report the missing expected part through the existing validation path.
🧹 Nitpick comments (4)
backend/app/services/collections/create_collection.py (1)
43-43: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winUse narrow types for retry inputs.
backend/app/services/collections/create_collection.py#L43-L43: annotatetask_instancewith a narrow bound-task protocol/type plusNone.backend/app/tests/services/collections/test_create_collection.py#L721-L721: replaceBaseExceptionwithTimeout | SoftTimeLimitExceeded.backend/app/tests/services/collections/test_create_collection.py#L779-L779: replaceBaseExceptionwithReject | TimeoutError.As per coding guidelines,
**/*.py: “Use Python 3.11+ and provide narrow type hints for every function parameter and return value.”🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/services/collections/create_collection.py` at line 43, Use a narrow bound-task protocol/type plus None for the task_instance parameter in _can_retry_batch. In backend/app/tests/services/collections/test_create_collection.py at lines 721 and 779, replace BaseException annotations with Timeout | SoftTimeLimitExceeded and Reject | TimeoutError respectively.Source: Coding guidelines
backend/app/tests/services/collections/test_create_collection.py (1)
720-720: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winMark intentionally unused patched storage mocks.
Ruff reports all four injected
mock_get_storageparameters as unused. Rename them to_mock_get_storage(or patch withnew=so no argument is injected) to keep the test suite lint-clean.
backend/app/tests/services/collections/test_create_collection.py#L720-L720: mark the unused storage mock intentional.backend/app/tests/services/collections/test_create_collection.py#L778-L778: mark the unused storage mock intentional.backend/app/tests/services/collections/test_create_collection.py#L830-L830: mark the unused storage mock intentional.backend/app/tests/services/collections/test_create_collection.py#L881-L881: mark the unused storage mock intentional.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/tests/services/collections/test_create_collection.py` at line 720, Rename each unused injected mock_get_storage parameter to _mock_get_storage in backend/app/tests/services/collections/test_create_collection.py at lines 720-720, 778-778, 830-830, and 881-881, preserving the existing patched-storage behavior while marking the mocks intentional and keeping Ruff lint-clean.Source: Linters/SAST tools
backend/app/services/documents/helpers.py (1)
231-238: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winExtract the shared ZIP signatures.
Both specs duplicate the same signatures. Define a
ZIP_SIGNATURESconstant and reuse it. As per coding guidelines, “Do not use magic values; extract repeated literals into constants, enums, or settings.”🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/services/documents/helpers.py` around lines 231 - 238, Define a shared ZIP_SIGNATURES constant containing the repeated ZIP signature tuple, then update the docx and xlsx FormatSpec definitions to reference it instead of duplicating the literal signatures. Keep needs_tail and each checker unchanged.Source: Coding guidelines
backend/app/services/collections/providers/registry.py (1)
8-8: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winKeep provider configuration out of the CRUD module.
backend/app/services/collections/providers/registry.pynow imports shared client settings frombackend/app/crud/rag/open_ai.py, creating a service-to-CRUD dependency. Move these constants to a neutral settings/provider-configuration module and import them from both locations.As per coding guidelines, use the repository's layered architecture: models represent database entities, CRUD handles database operations, API routes expose REST endpoints, services contain business logic, and Celery tasks handle asynchronous work.
Also applies to: 67-71
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/services/collections/providers/registry.py` at line 8, Move OPENAI_MAX_RETRIES and OPENAI_TIMEOUT_SECONDS out of app.crud.rag.open_ai into a neutral settings or provider-configuration module. Update both providers.registry and the CRUD OpenAI implementation to import these constants from the new module, removing the service-to-CRUD dependency while preserving their existing values and behavior.Source: Coding guidelines
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@backend/app/crud/rag/open_ai.py`:
- Around line 140-141: Update _poll_file_batch() so only a batch with status
"completed" is returned as successful; explicitly handle terminal statuses such
as "failed" and "cancelled" by raising or otherwise failing the upload,
including cancelled counts in the failure details where applicable. Do not let
non-completed terminal batches reach the successful-log path.
- Around line 136-140: Update _poll_file_batch so each _retrieve_file_batch call
cannot run past the remaining BATCH_POLL_TIMEOUT_SECONDS deadline. Compute the
remaining time before retrieval and pass it through the client/request timeout
while disabling or limiting retries for this polling operation; preserve the
existing status handling and stop polling when the deadline is exhausted.
In `@backend/app/services/documents/helpers.py`:
- Line 240: Update the XLS FormatSpec near OLE2_SIGNATURE to use an XLS-specific
checker that inspects the OLE compound document for a Workbook or Book stream
before accepting it. Keep generic OLE2 validation for other formats, but reject
legacy .doc and unrelated OLE containers in the XLS path.
In `@docs/architecture/kaapi-knowledge-base-ARCHITECTURE.md`:
- Around line 292-302: Update the sequence diagram to match the revised Phase
1/Phase 2 lifecycle: show setup creating and recording the vector store before
persisting metadata, enqueue batch 1 with the resolved vector_store_id, and
remove any setup step that uploads all documents. Show each batch performing its
bounded document upload and vector-store attach operation.
In `@docs/wiki/modules/knowledge-base.md`:
- Line 39: Update the knowledge-base documentation sentence to use standard
Python terminology: replace “contextvar” with “ContextVar” or “context
variable.” Preserve the existing explanation about caller coupling and polling
deadlines.
---
Outside diff comments:
In `@backend/app/services/documents/helpers.py`:
- Around line 123-128: Update the validation condition around the Office package
checks so a package is rejected whenever expected_part is absent, including
cases where neither expected_part nor other_part exists. Preserve the existing
DocumentValidationError and use other_part only when available to identify the
detected package type in the mismatch message; otherwise report the missing
expected part through the existing validation path.
---
Nitpick comments:
In `@backend/app/services/collections/create_collection.py`:
- Line 43: Use a narrow bound-task protocol/type plus None for the task_instance
parameter in _can_retry_batch. In
backend/app/tests/services/collections/test_create_collection.py at lines 721
and 779, replace BaseException annotations with Timeout | SoftTimeLimitExceeded
and Reject | TimeoutError respectively.
In `@backend/app/services/collections/providers/registry.py`:
- Line 8: Move OPENAI_MAX_RETRIES and OPENAI_TIMEOUT_SECONDS out of
app.crud.rag.open_ai into a neutral settings or provider-configuration module.
Update both providers.registry and the CRUD OpenAI implementation to import
these constants from the new module, removing the service-to-CRUD dependency
while preserving their existing values and behavior.
In `@backend/app/services/documents/helpers.py`:
- Around line 231-238: Define a shared ZIP_SIGNATURES constant containing the
repeated ZIP signature tuple, then update the docx and xlsx FormatSpec
definitions to reference it instead of duplicating the literal signatures. Keep
needs_tail and each checker unchanged.
In `@backend/app/tests/services/collections/test_create_collection.py`:
- Line 720: Rename each unused injected mock_get_storage parameter to
_mock_get_storage in
backend/app/tests/services/collections/test_create_collection.py at lines
720-720, 778-778, 830-830, and 881-881, preserving the existing patched-storage
behavior while marking the mocks intentional and keeping Ruff lint-clean.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 08d37652-b4a3-4011-9097-4498923146b3
📒 Files selected for processing (9)
backend/app/crud/rag/open_ai.pybackend/app/services/collections/create_collection.pybackend/app/services/collections/providers/registry.pybackend/app/services/documents/helpers.pybackend/app/tests/crud/rag/test_open_ai.pybackend/app/tests/services/collections/providers/test_openai_provider.pybackend/app/tests/services/collections/test_create_collection.pydocs/architecture/kaapi-knowledge-base-ARCHITECTURE.mddocs/wiki/modules/knowledge-base.md
🚧 Files skipped from review as they are similar to previous changes (2)
- backend/app/tests/services/collections/providers/test_openai_provider.py
- backend/app/tests/crud/rag/test_open_ai.py
| deadline = time.monotonic() + BATCH_POLL_TIMEOUT_SECONDS | ||
|
|
||
| while True: | ||
| batch = self._retrieve_file_batch(batch_id, vector_store_id) | ||
| if batch.status != "in_progress": |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | 🏗️ Heavy lift
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "## locate file"
fd -a 'open_ai\.py$' . || true
echo "## inspect relevant file outline/section"
if [ -f backend/app/crud/rag/open_ai.py ]; then
wc -l backend/app/crud/rag/open_ai.py
ast-grep outline backend/app/crud/rag/open_ai.py || true
echo "## lines 1-240"
sed -n '1,240p' backend/app/crud/rag/open_ai.py | nl -ba
echo "## timeout/retry references"
rg -n "timeout|retries|deadline|BATCH_POLL_TIMEOUT_SECONDS|_retrieve_file_batch|AsyncOpenAI|OpenAI" backend/app/crud/rag/open_ai.py || true
fi
echo "## dependency/version manifests"
for f in requirements.txt requirements-dev.txt pyproject.toml setup.cfg setup.py uv.lock poetry.lock Pipfile.lock; do
[ -f "$f" ] && { echo "--- $f"; rg -n "openai|httpx" "$f" || true; }
done
echo "## other timeout/retry usages"
rg -n "AsyncOpenAI|OpenAI\(|timeout\s*=|retries\s*=" backend || true
echo "## git diff stat and relevant diff"
git diff --stat || true
git diff -- backend/app/crud/rag/open_ai.py || trueRepository: ProjectTech4DevAI/kaapi-backend
Length of output: 1362
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "## lines 1-240"
sed -n '1,240p' backend/app/crud/rag/open_ai.py
echo "## timeout/retry references"
rg -n "timeout|retries|deadline|BATCH_POLL_TIMEOUT_SECONDS|_retrieve_file_batch|AsyncOpenAI|OpenAI" backend/app/crud/rag/open_ai.py || true
echo "## dependency/version manifests"
for f in requirements.txt requirements-dev.txt pyproject.toml setup.cfg setup.py uv.lock poetry.lock Pipfile.lock; do
[ -f "$f" ] && { echo "--- $f"; rg -n "openai|httpx" "$f" || true; }
done
echo "## other timeout/retry usages"
rg -n "AsyncOpenAI|OpenAI\(|timeout\s*=|retries\s*=" backend || true
echo "## focused tests for poll/retrieval client"
rg -n "OpenAIVectorStoreCrud|_poll_file_batch|_retrieve_file_batch|OPENAI_MAX_RETRIES|OPENAI_TIMEOUT_SECONDS|vector_store_file_batches|retrieve" backend/app/tests backend/app/crud backend/app/services || trueRepository: ProjectTech4DevAI/kaapi-backend
Length of output: 35059
🌐 Web query:
OpenAI Python SDK configuration 90 second timeout max_retries=2 default retry behavior
💡 Result:
To configure the OpenAI Python SDK with a 90-second timeout and 2 retries, you can set these parameters when initializing the OpenAI client [1][2]. Configuration Example: from openai import OpenAI import httpx # Configure timeout and retries client = OpenAI( timeout=90.0, # Set total timeout to 90 seconds [1][3] max_retries=2, # Set maximum number of retries [1][3]) Default Behavior and Details: - Timeout: The default timeout is 10 minutes (600 seconds) [4]. Setting timeout=90.0 overrides this global default [1][2]. You can also pass an httpx.Timeout object for more granular control (e.g., specifying connect, read, or write timeouts) [1][5]. - Retries: The default number of retries is already 2 [1][6]. These retries occur automatically for specific error types, including connection errors, 408 Request Timeout, 409 Conflict, 429 Rate Limit, and >=500 Internal Server Errors [1][3][6]. - Per-Request Overrides: If you need to apply different settings to a specific call rather than the entire client, you can use the with_options method [1][7]: response = client.with_options(timeout=90.0, max_retries=2).chat.completions.create(...) Note: Requests that time out are subject to the same retry logic and will be retried up to the configured max_retries limit [1][2][5].
Citations:
- 1: https://github.com/openai/openai-python/blob/2384e276969d5908eebff2ca12bdfcd69b3cfa93/README.md
- 2: https://github.com/openai/openai-python/blob/main/README.md
- 3: https://developers.openai.com/api/reference/python/
- 4: https://github.com/openai/openai-python/blob/722d3fff/src/openai/_constants.py
- 5: https://mintlify.wiki/openai/openai-python/concepts/timeouts
- 6: https://mintlify.wiki/openai/openai-python/concepts/retries
- 7: https://mintlify.wiki/openai/openai-python/concepts/client
Make the batch-poll deadline enforceable.
_retrieve_file_batch() uses the client’s 90s timeout and default retries, but _poll_file_batch() only checks the 240s deadline after _retrieve_file_batch() returns. If the SDK retries timed-out retrievals, polling can exceed the advertised wall-clock budget. Bound each retrieve/retry by the remaining deadline, or disable/limit retries for polling specifically.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@backend/app/crud/rag/open_ai.py` around lines 136 - 140, Update
_poll_file_batch so each _retrieve_file_batch call cannot run past the remaining
BATCH_POLL_TIMEOUT_SECONDS deadline. Compute the remaining time before retrieval
and pass it through the client/request timeout while disabling or limiting
retries for this polling operation; preserve the existing status handling and
stop polling when the deadline is exhausted.
Source: MCP tools
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (2)
backend/app/tests/services/documents/test_helpers.py (2)
265-279: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winAnnotate the test-method parameters.
Line 265 and Line 278 leave both
selfandmonkeypatchuntyped; the same applies to the other newly added test methods. Useself: Selfandmonkeypatch: pytest.MonkeyPatchthroughout the new test classes.As per coding guidelines, “Use Python 3.11+ and provide narrow type hints for every function parameter and return value.”
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/tests/services/documents/test_helpers.py` around lines 265 - 279, Annotate every newly added test method in the test classes with narrow parameter types: use self: Self and monkeypatch: pytest.MonkeyPatch, while preserving the existing return annotations. Apply this consistently to test_valid_csv_returns_format_and_transformer, test_invalid_content_raises_http_400_with_client_message, and the other new test methods.Source: Coding guidelines
205-219: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDerive the large-payload size from the configured parse threshold.
300000is duplicated and only exercises the closure-only branch while it remains aboveJSON_FULL_PARSE_MAX_BYTES. Import that constant and build both fixtures from it so these tests continue to cover the intended branch if the limit changes.Proposed fix
from app.services.documents.helpers import ( + JSON_FULL_PARSE_MAX_BYTES, SNIFF_HEAD_BYTES, SNIFF_TAIL_BYTES, @@ - content = b"[" + b" " * 300000 + b"]" + content = b"[" + b" " * JSON_FULL_PARSE_MAX_BYTES + b"]" @@ - content = b"[" + b" " * 300000 + content = b"[" + b" " * JSON_FULL_PARSE_MAX_BYTESAs per coding guidelines, “Do not use magic values; extract repeated literals into constants, enums, or settings.”
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/tests/services/documents/test_helpers.py` around lines 205 - 219, Update test_large_json_closed_correctly_passes and test_large_json_not_closed_raises to import and use JSON_FULL_PARSE_MAX_BYTES when constructing their payloads, adding enough bytes to remain above the configured full-parse threshold. Remove the duplicated 300000 literal while preserving the existing closed and unclosed tail behaviors.Source: Coding guidelines
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@backend/app/tests/crud/rag/test_open_ai.py`:
- Line 305: Update the new test methods
test_raises_interrupted_error_past_deadline and the method at the additional
referenced location with narrow type annotations for self if required, crud, and
mock_client, and add -> None return annotations. Follow the existing test typing
conventions and do not alter test behavior.
- Around line 332-336: Update get_mock_openai_client_with_vector_store() to
configure file_batches.retrieve.return_value to the same mock_file_batch used
for the batch contract, then replace the poll.return_value assertion in the test
with an assertion that file_batches.retrieve.return_value is batch.
- Around line 307-309: Update the monotonic side-effect sequence in the timeout
test to derive its post-deadline value from the production
BATCH_POLL_TIMEOUT_SECONDS constant instead of hard-coding 300, while preserving
the existing timing sequence and timeout behavior.
---
Nitpick comments:
In `@backend/app/tests/services/documents/test_helpers.py`:
- Around line 265-279: Annotate every newly added test method in the test
classes with narrow parameter types: use self: Self and monkeypatch:
pytest.MonkeyPatch, while preserving the existing return annotations. Apply this
consistently to test_valid_csv_returns_format_and_transformer,
test_invalid_content_raises_http_400_with_client_message, and the other new test
methods.
- Around line 205-219: Update test_large_json_closed_correctly_passes and
test_large_json_not_closed_raises to import and use JSON_FULL_PARSE_MAX_BYTES
when constructing their payloads, adding enough bytes to remain above the
configured full-parse threshold. Remove the duplicated 300000 literal while
preserving the existing closed and unclosed tail behaviors.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 8f4d2555-2352-4e72-88c6-112d1cc14252
📒 Files selected for processing (2)
backend/app/tests/crud/rag/test_open_ai.pybackend/app/tests/services/documents/test_helpers.py
| # monotonic: deadline setup, first below-deadline check, its sleep arg, | ||
| # then a value past the deadline that trips the timeout. | ||
| mock_time.monotonic.side_effect = [0, 0, 0, 300] |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Tie the timeout test to the configured deadline.
Line 309 hard-codes 300; if BATCH_POLL_TIMEOUT_SECONDS changes, this sequence may no longer reach the deadline. Use the production timeout constant instead.
As per coding guidelines, do not use magic values; extract them into constants or settings.
Proposed fix
- mock_time.monotonic.side_effect = [0, 0, 0, 300]
+ mock_time.monotonic.side_effect = [
+ 0,
+ 0,
+ 0,
+ BATCH_POLL_TIMEOUT_SECONDS,
+ ]📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| # monotonic: deadline setup, first below-deadline check, its sleep arg, | |
| # then a value past the deadline that trips the timeout. | |
| mock_time.monotonic.side_effect = [0, 0, 0, 300] | |
| # monotonic: deadline setup, first below-deadline check, its sleep arg, | |
| # then a value past the deadline that trips the timeout. | |
| mock_time.monotonic.side_effect = [ | |
| 0, | |
| 0, | |
| 0, | |
| BATCH_POLL_TIMEOUT_SECONDS, | |
| ] |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@backend/app/tests/crud/rag/test_open_ai.py` around lines 307 - 309, Update
the monotonic side-effect sequence in the timeout test to derive its
post-deadline value from the production BATCH_POLL_TIMEOUT_SECONDS constant
instead of hard-coding 300, while preserving the existing timing sequence and
timeout behavior.
Source: Coding guidelines
… logic for better error handling
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
backend/app/crud/rag/open_ai.py (1)
98-98: 📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick winAnnotate
selfon the touched CRUD methods.
create,_create_file_batch,_retrieve_file_batch, and_poll_file_batchannotate parameters and returns but leaveselfuntyped. Also applies toread,update, anddelete, which should be typed consistently to match the required every-parameter typing policy.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/crud/rag/open_ai.py` at line 98, Update the CRUD methods create, _create_file_batch, _retrieve_file_batch, _poll_file_batch, read, update, and delete to add the required type annotation for self, while preserving their existing parameter and return annotations.Source: Coding guidelines
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@backend/app/crud/rag/open_ai.py`:
- Line 136: Centralize the OpenAI file-batch status values used by the CRUD flow
instead of hard-coding "in_progress" and "completed" in the status handling
around batch.status. Define reusable constants or an enum in the appropriate
OpenAI module, update this path and dependent tests/callers to reference them,
and preserve the existing protocol values.
---
Outside diff comments:
In `@backend/app/crud/rag/open_ai.py`:
- Line 98: Update the CRUD methods create, _create_file_batch,
_retrieve_file_batch, _poll_file_batch, read, update, and delete to add the
required type annotation for self, while preserving their existing parameter and
return annotations.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 8cc6c27b-e9d4-49de-892f-74ddf9bdb4a2
📒 Files selected for processing (7)
backend/app/crud/rag/open_ai.pybackend/app/services/documents/helpers.pybackend/app/tests/crud/rag/test_open_ai.pybackend/app/tests/services/documents/test_helpers.pybackend/app/tests/utils/openai.pydocs/architecture/kaapi-knowledge-base-ARCHITECTURE.mddocs/wiki/modules/knowledge-base.md
🚧 Files skipped from review as they are similar to previous changes (2)
- docs/architecture/kaapi-knowledge-base-ARCHITECTURE.md
- docs/wiki/modules/knowledge-base.md
| """Poll until indexing finishes; the Celery soft time limit is the deadline.""" | ||
| while True: | ||
| batch = self._retrieve_file_batch(batch_id, vector_store_id) | ||
| if batch.status != "in_progress": |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Locate and inspect the target file and relevant literals/usages without running repo code.
if [ -f backend/app/crud/rag/open_ai.py ]; then
echo "== target file size =="
wc -l backend/app/crud/rag/open_ai.py
echo
echo "== relevant file sections =="
sed -n '120,145p;270,290p' backend/app/crud/rag/open_ai.py
echo
echo "== batch.status/status value literals in target =="
python3 - <<'PY'
import re
from pathlib import Path
p=Path('backend/app/crud/rag/open_ai.py')
text=p.read_text()
for i,line in enumerate(text.splitlines(),1):
if '.status' in line or re.search(r'"(?:in_progress|completed|failed|pending|success|cancelled)"', line):
print(f"{i}: {line}")
PY
else
echo "backend/app/crud/rag/open_ai.py not found"
git ls-files | rg 'open_ai\.py$|rag' || true
fi
echo
echo "== tests/files mentioning open_ai batch status literals =="
git ls-files | rg 'test|spec|py$' | xargs -r grep -E 'batch\.status|in_progress|completed|success' | head -n 200 || trueRepository: ProjectTech4DevAI/kaapi-backend
Length of output: 22815
Centralize OpenAI file-batch status values.
"in_progress" and "completed" are hard-coded directly in the CRUD path. Extract batch-status constants or an enum here and use the same values in any tests/callers that depend on this protocol state.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@backend/app/crud/rag/open_ai.py` at line 136, Centralize the OpenAI
file-batch status values used by the CRUD flow instead of hard-coding
"in_progress" and "completed" in the status handling around batch.status. Define
reusable constants or an enum in the appropriate OpenAI module, update this path
and dependent tests/callers to reference them, and preserve the existing
protocol values.
Source: Coding guidelines
…ssertions for job status
1 participant
Issue
Closes #1076
Summary
create()function when listing failed files.Checklist
Before submitting a pull request, please ensure that you mark these task.
fastapi run --reload app/main.pyordocker compose upin the repository root and test.