Merge pull request #1167 from julep-ai/x/ffmpeg-fix

fix(integrations): updated ffmpeg file input to support list of strings

agents-api/agents_api/autogen/Tools.py

@@ -1199,7 +1199,7 @@ class FfmpegSearchArguments(BaseModel):
     """
     The bash command string
     """
-    file: str | None = None
+    file: str | list[str] | None = None
     """
     The base64 string of the file
     """
@@ -1217,7 +1217,7 @@ class FfmpegSearchArgumentsUpdate(BaseModel):
     """
     The bash command string
     """
-    file: str | None = None
+    file: str | list[str] | None = None
     """
     The base64 string of the file
     """

agents-api/agents_api/workflows/task_execution/__init__.py

@@ -393,6 +393,11 @@ async def _handle_YieldStep(
 
         yield_transition_type, yield_next_target = self.outcome.transition_to
         workflow.logger.info(f"Yield step: Transitioning to {yield_transition_type}")
+
+        if self.context is None:
+            msg = "Context must not be None"
+            raise ApplicationError(msg)
+
         await transition(
             self.context,
             output=output,

agents-api/agents_api/workflows/task_execution/helpers.py

@@ -79,6 +79,10 @@ async def continue_as_child(
             info.workflow_type, *args, **kwargs
         )
 
+    if execution_input.execution is None:
+        msg = "Execution input execution cannot be None"
+        raise ApplicationError(msg)
+
     execution_id = execution_input.execution.id
     execution_id_key = SearchAttributeKey.for_keyword("CustomStringField")
 

documentation/docs/integrations/media_file_processing/cloudinary.mdx

@@ -78,7 +78,7 @@ To get started with the Cloudinary integration, follow these steps to configure
       - **_tool_**: Refers to the tool defined earlier (`cloudinary_tool`).
       - **_arguments_**: Specifies the input parameters for the tool:
         <Accordion title="media_upload">
-          - **_file_**: The URL of the file to upload.
+          - **_file_**: The URL of the file to upload. More details can be found in the [Cloudinary documentation](https://cloudinary.com/documentation/image_upload_api_reference#upload).
           - **_public_id_**: (optional) Optional public ID for the uploaded file. Defaults to None.
           - **_upload_params_**: (optional) Optional transformations for the upload. Defaults to None.
           - **_return_base64_**: (optional) Whether to return the file in base64 encoding. Defaults to False.

documentation/docs/integrations/media_file_processing/ffmpeg.mdx

@@ -30,6 +30,13 @@ To get started with the FFmpeg integration, follow these steps to configure and
           cmd: "ffmpeg -i input.mp4 -vn -acodec copy output.aac"
           file: "base64_encoded_file"
     ```
+
+    <Info>
+      The `base64_encoded_file` is the base64 encoded file to process which in this case is the `input.mp4` file.
+      The `file` argument can accept either a single base64 encoded string or a list of base64 encoded strings.
+      However, even when passing a list of files, the FFmpeg command can only use a single input file (single `-i` flag).
+      Multiple input files with multiple `-i` flags are not supported.
+    </Info>
   </Step>
 </Steps>
 
@@ -53,7 +60,7 @@ To get started with the FFmpeg integration, follow these steps to configure and
       - **_tool_**: Refers to the tool defined earlier (`ffmpeg_tool`).
       - **_arguments_**: Specifies the input parameters for the tool:
         - **_cmd_**: The FFmpeg command to execute.
-        - **_file_**: The base64 encoded file to process.
+        - **_file_**: The base64 encoded file to process. Can be a single base64 encoded string or a list of base64 encoded strings.
   </Accordion>
 </AccordionGroup>
 

integrations-service/integrations/autogen/Tools.py

@@ -1199,7 +1199,7 @@ class FfmpegSearchArguments(BaseModel):
     """
     The bash command string
     """
-    file: str | None = None
+    file: str | list[str] | None = None
     """
     The base64 string of the file
     """
@@ -1217,7 +1217,7 @@ class FfmpegSearchArgumentsUpdate(BaseModel):
     """
     The bash command string
     """
-    file: str | None = None
+    file: str | list[str] | None = None
     """
     The base64 string of the file
     """

integrations-service/integrations/utils/integrations/ffmpeg.py

@@ -65,22 +65,35 @@ async def bash_cmd(arguments: FfmpegSearchArguments) -> FfmpegSearchOutput:
 
         # Decode base64 input
         try:
-            input_data = base64.b64decode(arguments.file)
+            if isinstance(arguments.file, str):
+                input_data = [base64.b64decode(arguments.file)]  # Convert single str to list
+            elif isinstance(arguments.file, list):
+                input_data = [base64.b64decode(file) for file in arguments.file]
+            else:
+                msg = "Invalid file input"
+                raise ValueError(msg)
         except Exception:
             return FfmpegSearchOutput(
                 fileoutput="Error: Invalid base64 input", result=False, mime_type=None
             )
 
         # Validate input format
-        is_valid, input_mime = await validate_format(input_data)
-
-        if not is_valid:
+        if not isinstance(input_data, list):
             return FfmpegSearchOutput(
-                fileoutput="Error: Unsupported input file format",
+                fileoutput="Error: Input data must be a list",
                 result=False,
                 mime_type=None,
             )
 
+        for data in input_data:
+            is_valid, input_mime = await validate_format(data)
+            if not is_valid:
+                return FfmpegSearchOutput(
+                    fileoutput="Error: Unsupported input file format",
+                    result=False,
+                    mime_type=None,
+                )
+
         # Create temporary directory
         temp_dir = tempfile.mkdtemp()
 
@@ -104,8 +117,16 @@ async def bash_cmd(arguments: FfmpegSearchArguments) -> FfmpegSearchOutput:
             stderr=asyncio.subprocess.PIPE,
         )
 
+        # Write each decoded image to FFmpeg's stdin
+        for image_data in input_data:
+            process.stdin.write(image_data)  # Send image to stdin
+            await process.stdin.drain()  # Ensure smooth streaming
+
+        # Close stdin to signal end of input
+        process.stdin.close()
+
         # Process FFmpeg output
-        stdout, stderr = await process.communicate(input=input_data)
+        stdout, stderr = await process.communicate()
         success = process.returncode == 0
 
         if success and os.path.exists(output_path):