Test audio image transcription #1911
Conversation
Please make sure all the checkboxes are checked:
|
|
| GitGuardian id | GitGuardian status | Secret | Commit | Filename | |
|---|---|---|---|---|---|
| 9573981 | Triggered | Generic Password | 13c034e | .github/workflows/examples_tests.yml | View secret |
🛠 Guidelines to remediate hardcoded secrets
- Understand the implications of revoking this secret by investigating where it is used in your code.
- Replace and store your secret safely. Learn here the best practices.
- Revoke and rotate this secret.
- If possible, rewrite git history. Rewriting git history is not a trivial act. You might completely break other contributing developers' workflow and you risk accidentally deleting legitimate data.
To avoid such incidents in the future consider
- following these best practices for managing and storing secrets including API keys and other credentials
- install secret detection on pre-commit to catch secret before it leaves your machine and ease remediation.
🦉 GitGuardian detects secrets in your source code to help developers and security teams secure the modern development process. You are seeing this because you or someone else with access to this repository has authorized GitGuardian to scan your pull request.
WalkthroughRefactored LLM adapter architecture by removing the synchronous Changes
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes
Possibly related PRs
Suggested reviewers
Pre-merge checks and finishing touches❌ Failed checks (3 warnings, 1 inconclusive)
✅ Passed checks (1 passed)
✨ Finishing touches
🧪 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 |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (4)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py (1)
97-99: Unused import ofGenericAPIAdapterin OLLAMA branch.
GenericAPIAdapteris imported butOllamaAPIAdapteris instantiated. This import appears to be dead code.elif provider == LLMProvider.OLLAMA: if llm_config.llm_api_key is None and raise_api_key_error: raise LLMAPIKeyNotSetError() - from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.generic_llm_api.adapter import ( - GenericAPIAdapter, - ) - return OllamaAPIAdapter(cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.py (1)
79-91: Hardcodedmax_tokens=4096ignores configuredmax_completion_tokens.The constructor accepts
max_completion_tokensand passes it to the base class (line 41), but line 81 uses a hardcoded value of4096. This means the configured token limit is ignored, potentially causing unexpected behavior.return await self.aclient( model=self.model, - max_tokens=4096, + max_tokens=self.max_completion_tokens, max_retries=2, messages=[cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py (2)
179-181: Commented-out code should be removed or reinstated.The
api_base=self.fallback_endpointline is commented out but the base classGenericAPIAdapteruses it in its fallback path. This inconsistency could cause the fallback to behave differently than expected.api_key=self.fallback_api_key, - # api_base=self.fallback_endpoint, + api_base=self.fallback_endpoint, response_model=response_model,
227-236: Replacelitellm.transcription()with async alternative to prevent event loop blocking.The synchronous
litellm.transcription()blocks the event loop in async context; useawait litellm.atranscription(), or wrap the call inasyncio.to_thread()if needed.
🧹 Nitpick comments (10)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/llm_interface.py (1)
9-14: Docstring mentions multimodal processing but interface lacks corresponding methods.The docstring claims the interface defines methods for "multimodal processing" (line 10), but only
acreate_structured_outputis documented and implemented. If multimodal capabilities are intended, consider adding method signatures for transcription and image processing to align with the documented scope, or update the docstring to reflect the actual interface.class LLMInterface(Protocol): """ - Define an interface for LLM models with methods for structured output, multimodal processing, and prompt display. + Define an interface for LLM models with methods for structured output generation. Methods: - acreate_structured_output(text_input: str, system_prompt: str, response_model: Type[BaseModel]) """cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.py (1)
1-10: Consider using@dataclassfor cleaner data container definition.The class uses class-level type annotations (lines 5-6) which aren't enforced at runtime and don't serve as instance attributes. Using
@dataclasswould be more idiomatic and eliminate the boilerplate__init__.Additionally, a module docstring would help document this type's purpose per coding guidelines.
+"""Types for LLM transcription return values.""" + +from dataclasses import dataclass from pydantic import BaseModel +@dataclass class TranscriptionReturnType: + """Container for transcription results with raw text and structured payload.""" + text: str payload: BaseModel - - def __init__(self, text: str, payload: BaseModel): - self.text = text - self.payload = payloadcognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py (2)
115-120: Inconsistent argument passing style: consider using keyword arguments.
AnthropicAdapteruses positional arguments (lines 116-118) while other adapters likeOpenAIAdapter,GeminiAdapter, andMistralAdapteruse keyword arguments. Using keyword arguments consistently improves readability and reduces risk of positional mismatches if constructor signatures change.return AnthropicAdapter( - llm_config.llm_api_key, - llm_config.llm_model, - max_completion_tokens, + api_key=llm_config.llm_api_key, + model=llm_config.llm_model, + max_completion_tokens=max_completion_tokens, instructor_mode=llm_config.llm_instructor_mode.lower(), )
130-139: Same inconsistency:GenericAPIAdapteruses positional arguments.For consistency with other adapters and maintainability, prefer keyword arguments here as well.
return GenericAPIAdapter( - llm_config.llm_api_key, - llm_config.llm_model, - max_completion_tokens, + api_key=llm_config.llm_api_key, + model=llm_config.llm_model, + max_completion_tokens=max_completion_tokens, "Custom", instructor_mode=llm_config.llm_instructor_mode.lower(),cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py (4)
7-7: Unused import:Optional
Optionalis imported but not used anywhere in this file.-from typing import Type, Optional +from typing import Type
198-220: Type annotation missing and inefficient MIME type validation order.
- The
inputparameter should be typed asstr.- MIME type validation occurs after the file has already been read and base64-encoded. If validation fails, work was wasted and the error message is misleading since we did successfully open the file.
- async def create_transcript(self, input) -> TranscriptionReturnType: + async def create_transcript(self, input: str) -> TranscriptionReturnType: """ Generate an audio transcript from a user query. ... """ + mime_type, _ = mimetypes.guess_type(input) + if not mime_type or not mime_type.startswith("audio/"): + raise ValueError( + f"Could not determine MIME type for audio file: {input}. Is the extension correct?" + ) async with open_data_file(input, mode="rb") as audio_file: encoded_string = base64.b64encode(audio_file.read()).decode("utf-8") - mime_type, _ = mimetypes.guess_type(input) - if not mime_type or not mime_type.startswith("audio/"): - raise ValueError( - f"Could not determine MIME type for audio file: {input}. Is the extension correct?" - ) response = await litellm.acompletion(
252-274: Type annotation and MIME validation order; return type mismatch.
- The
inputparameter should be typed asstr.- Same issue as
create_transcript: validate MIME type before reading the file.- The return type annotation says
BaseModel, but the method returns a rawlitellmresponse object, not aBaseModelinstance. Consider usingAnyor defining a proper return type.- async def transcribe_image(self, input) -> BaseModel: + async def transcribe_image(self, input: str): """ Generate a transcription of an image from a user query. ... Returns: -------- - - BaseModel: A structured output generated by the model, returned as an instance of - BaseModel. + The raw response from the LLM API containing the image description. """ + mime_type, _ = mimetypes.guess_type(input) + if not mime_type or not mime_type.startswith("image/"): + raise ValueError( + f"Could not determine MIME type for image file: {input}. Is the extension correct?" + ) async with open_data_file(input, mode="rb") as image_file: encoded_image = base64.b64encode(image_file.read()).decode("utf-8") - mime_type, _ = mimetypes.guess_type(input) - if not mime_type or not mime_type.startswith("image/"): - raise ValueError( - f"Could not determine MIME type for image file: {input}. Is the extension correct?" - )
297-297: Hardcodedmax_completion_tokens=300may be too restrictive.For detailed image descriptions, 300 tokens could truncate the response. Consider making this configurable or using
self.max_completion_tokensfor consistency with other methods.- max_completion_tokens=300, + max_completion_tokens=self.max_completion_tokens,cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py (1)
41-46: Docstring listscreate_structured_outputbut method may be removed.The docstring mentions
create_structured_outputas a public method, but based on the AI summary, the synchronous version was removed. Update the docstring to reflect the current API surface.Public methods: - acreate_structured_output - - create_structured_output - create_transcript - transcribe_image - show_promptcognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py (1)
138-138: Return typeOptional[TranscriptionReturnType]is misleading.The method always returns a
TranscriptionReturnTypeor raises an exception; it never returnsNone. RemoveOptionalfor accurate typing.- async def create_transcript(self, input) -> Optional[TranscriptionReturnType]: + async def create_transcript(self, input: str) -> TranscriptionReturnType:
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (10)
cognee/infrastructure/llm/LLMGateway.py(0 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.py(2 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.py(3 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py(7 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py(3 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/llm_interface.py(1 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py(5 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/ollama/adapter.py(0 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py(5 hunks)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.py(1 hunks)
💤 Files with no reviewable changes (2)
- cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/ollama/adapter.py
- cognee/infrastructure/llm/LLMGateway.py
🧰 Additional context used
📓 Path-based instructions (3)
**/*.py
📄 CodeRabbit inference engine (AGENTS.md)
**/*.py: Use 4-space indentation in Python code
Use snake_case for Python module and function names
Use PascalCase for Python class names
Use ruff format before committing Python code
Use ruff check for import hygiene and style enforcement with line-length 100 configured in pyproject.toml
Prefer explicit, structured error handling in Python code
Files:
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/llm_interface.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py
⚙️ CodeRabbit configuration file
**/*.py: When reviewing Python code for this project:
- Prioritize portability over clarity, especially when dealing with cross-Python compatibility. However, with the priority in mind, do still consider improvements to clarity when relevant.
- As a general guideline, consider the code style advocated in the PEP 8 standard (excluding the use of spaces for indentation) and evaluate suggested changes for code style compliance.
- As a style convention, consider the code style advocated in CEP-8 and evaluate suggested changes for code style compliance.
- As a general guideline, try to provide any relevant, official, and supporting documentation links to any tool's suggestions in review comments. This guideline is important for posterity.
- As a general rule, undocumented function definitions and class definitions in the project's Python code are assumed incomplete. Please consider suggesting a short summary of the code for any of these incomplete definitions as docstrings when reviewing.
Files:
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/llm_interface.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py
cognee/**/*.py
📄 CodeRabbit inference engine (AGENTS.md)
Use shared logging utilities from cognee.shared.logging_utils in Python code
Files:
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/llm_interface.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py
cognee/{modules,infrastructure,tasks}/**/*.py
📄 CodeRabbit inference engine (AGENTS.md)
Co-locate feature-specific helpers under their respective package (modules/, infrastructure/, or tasks/)
Files:
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/llm_interface.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.pycognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/get_llm_client.py
🧬 Code graph analysis (4)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.py (4)
cognee/infrastructure/llm/exceptions.py (1)
ContentPolicyFilterError(4-5)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py (1)
GenericAPIAdapter(38-300)cognee/shared/logging_utils.py (1)
get_logger(212-224)cognee/modules/observability/get_observe.py (1)
get_observe(5-25)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py (3)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py (2)
GenericAPIAdapter(38-300)create_transcript(198-242)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.py (1)
TranscriptionReturnType(4-10)cognee/infrastructure/llm/LLMGateway.py (1)
create_transcript(41-47)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py (3)
cognee/infrastructure/files/utils/open_data_file.py (1)
open_data_file(11-54)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py (2)
GenericAPIAdapter(38-300)create_transcript(198-242)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.py (1)
TranscriptionReturnType(4-10)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py (4)
cognee/infrastructure/files/utils/open_data_file.py (1)
open_data_file(11-54)cognee/modules/observability/get_observe.py (1)
get_observe(5-25)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/types.py (1)
TranscriptionReturnType(4-10)cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py (1)
create_transcript(208-237)
🔇 Additional comments (7)
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.py (1)
35-49: LGTM on the adapter refactoring to useGenericAPIAdapter.The inheritance change, constructor delegation, and API client initialization using
self.api_keyfrom the base class are correctly implemented. The observability decorator integration aligns with the broader PR changes.cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.py (3)
62-78: Well-structured base class initialization with extended parameters.The
super().__init__call correctly passes all relevant configuration includingendpoint,api_version,transcription_model, and fallback credentials. The integration withGenericAPIAdapteris properly implemented.
134-182: Content policy error handling with fallback is comprehensive.The error handling catches content policy violations, checks for valid fallback configuration before attempting retry, and properly re-raises non-content-policy
InstructorRetryExceptionerrors. The nested try/except for the fallback path mirrors the primary error handling appropriately.
118-127: No action required. The message ordering in the messages array does not affect Gemini's response quality when using the instructor library adapter, as the adapter automatically converts and properly formats messages for the Gemini API.cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py (1)
77-88: LGTM: Proper delegation to base class constructor.The
super().__init__()call correctly passes all required parameters toGenericAPIAdapter, including the newtranscription_modeland fallback configuration.cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py (2)
51-59: LGTM: Proper base class initialization.The constructor correctly delegates to
GenericAPIAdapter.__init__()with appropriate parameters including the new transcription model options.
70-70: Good addition of observability decorator.Adding
@observe(as_type="generation")aligns with the consistent observability pattern across all adapters in this refactor.
| Public methods: | ||
| - acreate_structured_output(text_input: str, system_prompt: str, response_model: | ||
| Type[BaseModel]) -> BaseModel | ||
| - acreate_structured_output(text_input: str, system_prompt: str, response_model: Type[BaseModel]) -> BaseModel | ||
| - create_transcript(input) -> BaseModel: Transcribe audio files to text | ||
| - transcribe_image(input) -> BaseModel: Inherited from GenericAPIAdapter | ||
| """ |
There was a problem hiding this comment.
Docstring inaccuracy: create_transcript return type and method ownership.
The docstring states create_transcript(input) -> BaseModel but per the GenericAPIAdapter context, this method returns TranscriptionReturnType. Also, both create_transcript and transcribe_image are inherited from the base class, not defined in this adapter.
Public methods:
- acreate_structured_output(text_input: str, system_prompt: str, response_model: Type[BaseModel]) -> BaseModel
- - create_transcript(input) -> BaseModel: Transcribe audio files to text
- - transcribe_image(input) -> BaseModel: Inherited from GenericAPIAdapter
+
+ Inherited from GenericAPIAdapter:
+ - create_transcript(input) -> TranscriptionReturnType: Transcribe audio files to text
+ - transcribe_image(input) -> BaseModel: Transcribe image content
"""| transcription_model = self.transcription_model | ||
| if self.transcription_model.startswith("mistral"): | ||
| transcription_model = self.transcription_model.split("/")[-1] | ||
| file_name = input.split("/")[-1] |
There was a problem hiding this comment.
Use os.path.basename() instead of string split for path handling.
input.split("/")[-1] is not portable across operating systems. Use os.path.basename() for cross-platform compatibility. As per coding guidelines, prioritize portability.
+import os
...
- file_name = input.split("/")[-1]
+ file_name = os.path.basename(input)📝 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.
| file_name = input.split("/")[-1] | |
| import os | |
| ... | |
| file_name = os.path.basename(input) |
🤖 Prompt for AI Agents
In
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py
around line 156, replace the non-portable path handling "input.split('/')[-1]"
with os.path.basename(input) to ensure cross-platform correctness; if os is not
already imported at the top of the module, add "import os" and update the
variable assignment accordingly so the filename is derived using
os.path.basename.
| async with open_data_file(input, mode="rb") as f: | ||
| transcription_response = self.mistral_client.audio.transcriptions.complete( | ||
| model=transcription_model, | ||
| file={ | ||
| "content": f, | ||
| "file_name": file_name, | ||
| }, | ||
| ) | ||
|
|
||
| return TranscriptionReturnType(transcription_response.text, transcription_response) |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
find . -name "adapter.py" -path "*/mistral/*" | head -20Repository: topoteretes/cognee
Length of output: 159
🏁 Script executed:
fd "adapter.py" -p "*/mistral/*" -x cat -n {}Repository: topoteretes/cognee
Length of output: 160
🏁 Script executed:
rg "mistral_client" -A 5 -B 5Repository: topoteretes/cognee
Length of output: 3101
🏁 Script executed:
cat -n cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py | sed -n '140,175p'Repository: topoteretes/cognee
Length of output: 1338
🏁 Script executed:
rg "import.*Mistral|from.*mistral" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.pyRepository: topoteretes/cognee
Length of output: 91
🏁 Script executed:
rg "async def.*transcrib" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py -A 20Repository: topoteretes/cognee
Length of output: 44
🏁 Script executed:
rg "async def" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py | grep -i transcribRepository: topoteretes/cognee
Length of output: 44
🏁 Script executed:
cat -n cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py | sed -n '130,140p'Repository: topoteretes/cognee
Length of output: 588
🏁 Script executed:
rg "def.*transcrib" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py -B 5Repository: topoteretes/cognee
Length of output: 44
🏁 Script executed:
cat -n cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py | head -30Repository: topoteretes/cognee
Length of output: 1202
🏁 Script executed:
rg "Mistral|mistralai" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.pyRepository: topoteretes/cognee
Length of output: 305
🏁 Script executed:
grep -r "asyncio.to_thread\|run_in_executor" cognee/ | head -10Repository: topoteretes/cognee
Length of output: 600
🏁 Script executed:
cat pyproject.toml | grep -A 5 "mistral"Repository: topoteretes/cognee
Length of output: 373
🏁 Script executed:
rg "mistral_client\." cognee/ -B 2 -A 2Repository: topoteretes/cognee
Length of output: 794
🏁 Script executed:
rg "self\.aclient\|acompletion" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py -B 3 -A 3Repository: topoteretes/cognee
Length of output: 44
🏁 Script executed:
python3 << 'EOF'
import subprocess
result = subprocess.run(
["pip", "show", "mistralai"],
capture_output=True,
text=True
)
print(result.stdout)
print(result.stderr)
EOFRepository: topoteretes/cognee
Length of output: 148
🏁 Script executed:
python3 << 'EOF'
# Check if mistralai module has async audio transcription
try:
from mistralai import Mistral
import inspect
client = Mistral.__dict__
# Check for audio attribute
if hasattr(Mistral, 'audio'):
print("Mistral has 'audio' attribute")
audio_attr = getattr(Mistral, 'audio')
print(f"Type: {type(audio_attr)}")
else:
print("Mistral does not have 'audio' attribute in class definition")
# List some key methods/attributes
print("\nMistral class attributes/methods:")
for attr in dir(Mistral):
if not attr.startswith('_'):
print(f" - {attr}")
except Exception as e:
print(f"Error: {e}")
EOFRepository: topoteretes/cognee
Length of output: 148
🏁 Script executed:
rg "\.audio\.transcriptions\." cognee/ -B 5 -A 5Repository: topoteretes/cognee
Length of output: 3082
🏁 Script executed:
cat -n cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py | sed -n '1,80p'Repository: topoteretes/cognee
Length of output: 3169
🏁 Script executed:
rg "aclient" cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py -B 3 -A 3Repository: topoteretes/cognee
Length of output: 666
Wrap synchronous Mistral API call in asyncio.to_thread() to avoid blocking the event loop.
The create_transcript method is declared as async, but self.mistral_client.audio.transcriptions.complete() is a synchronous call that blocks the event loop. Replace the call at line 158 with:
transcription_response = await asyncio.to_thread(
self.mistral_client.audio.transcriptions.complete,
model=transcription_model,
file={
"content": f,
"file_name": file_name,
},
)Alternatively, check if the mistralai SDK provides an async audio transcription API.
🤖 Prompt for AI Agents
In
cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py
around lines 157 to 166, the synchronous call to
self.mistral_client.audio.transcriptions.complete(...) inside an async function
blocks the event loop; wrap that call with asyncio.to_thread(...) and await it
(or switch to an async SDK method if available) so the blocking work runs in a
thread; also ensure asyncio is imported at top of the file and keep passing the
same model and file arguments when calling via to_thread.
Description
Run CI/CD for audio/image transcription PR from contributor @rajeevrajeshuni
Acceptance Criteria
Type of Change
Screenshots/Videos (if applicable)
Pre-submission Checklist
DCO Affirmation
I affirm that all code in every commit of this pull request conforms to the terms of the Topoteretes Developer Certificate of Origin.
Summary by CodeRabbit
Release Notes
New Features
Breaking Changes
Refactor
✏️ Tip: You can customize this high-level summary in your review settings.