clarifai 11.7.5__tar.gz → 11.7.5rc1__tar.gz

{clarifai-11.7.5/clarifai.egg-info → clarifai-11.7.5rc1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: clarifai
-Version: 11.7.5
+Version: 11.7.5rc1
 Home-page: https://github.com/Clarifai/clarifai-python
 Author: Clarifai
 Author-email: support@clarifai.com

clarifai-11.7.5rc1/clarifai/__init__.py

@@ -0,0 +1 @@
+__version__ = "11.7.5rc1"

{clarifai-11.7.5 → clarifai-11.7.5rc1}/clarifai/cli/base.py

@@ -304,8 +304,8 @@ def view(ctx, output_format):
 def run(ctx, script, context=None):
     """Execute a script with the current context's environment"""
     context = ctx.obj.current if not context else context
-    cmd = f'CLARIFAI_USER_ID={context.user_id} CLARIFAI_API_BASE={context.base_url} CLARIFAI_PAT={context.pat} '
-    cmd += ' '.join([f'{k}={v}' for k, v in context.env.items()])
+    cmd = f'CLARIFAI_USER_ID={context.user_id} CLARIFAI_API_BASE={context.api_base} CLARIFAI_PAT={context.pat} '
+    cmd += ' '.join([f'{k}={v}' for k, v in context.to_serializable_dict().items()])
     cmd += f' {script}'
     os.system(cmd)
 

{clarifai-11.7.5 → clarifai-11.7.5rc1}/clarifai/cli/model.py

@@ -1,3 +1,4 @@
+import json
 import os
 import shutil
 import tempfile
@@ -25,7 +26,12 @@ from clarifai.utils.constants import (
     DEFAULT_OLLAMA_MODEL_REPO_BRANCH,
 )
 from clarifai.utils.logging import logger
-from clarifai.utils.misc import GitHubDownloader, clone_github_repo, format_github_repo_url
+from clarifai.utils.misc import (
+    GitHubDownloader,
+    clone_github_repo,
+    format_github_repo_url,
+    get_list_of_files_to_download,
+)
 
 
 @cli.group(
@@ -150,16 +156,27 @@ def init(
         branch = DEFAULT_OLLAMA_MODEL_REPO_BRANCH
 
     if github_url:
+        downloader = GitHubDownloader(
+            max_retries=3,
+            github_token=github_pat,
+        )
+        if toolkit:
+            owner, repo, _, folder_path = downloader.parse_github_url(url=github_url)
+        else:
+            owner, repo, branch, folder_path = downloader.parse_github_url(url=github_url)
+        logger.info(
+            f"Parsed GitHub repository: owner={owner}, repo={repo}, branch={branch}, folder_path={folder_path}"
+        )
+        files_to_download = get_list_of_files_to_download(
+            downloader, owner, repo, folder_path, branch, []
+        )
+        for i, file in enumerate(files_to_download):
+            files_to_download[i] = f"{i + 1}. {file}"
+        files_to_download = '\n'.join(files_to_download)
+        logger.info(f"Files to be downloaded are:\n{files_to_download}")
+        input("Press Enter to continue...")
         if not toolkit:
-            owner, repo, branch, folder_path = GitHubDownloader().parse_github_url(url=github_url)
-            logger.info(
-                f"Parsed GitHub repository: owner={owner}, repo={repo}, branch={branch}, folder_path={folder_path}"
-            )
             if folder_path != "":
-                downloader = GitHubDownloader(
-                    max_retries=3,
-                    github_token=github_pat,
-                )
                 try:
                     downloader.download_github_folder(
                         url=github_url,
@@ -168,6 +185,10 @@ def init(
                     )
                     logger.info(f"Successfully downloaded folder contents to {model_path}")
                     logger.info("Model initialization complete with GitHub folder download")
+                    logger.info("Next steps:")
+                    logger.info("1. Review the model configuration")
+                    logger.info("2. Install any required dependencies manually")
+                    logger.info("3. Test the model locally using 'clarifai model local-test'")
                     return
 
                 except Exception as e:
@@ -235,6 +256,9 @@ def init(
 
     # Fall back to template-based initialization if no GitHub repo or if GitHub repo failed
     if not github_url:
+        logger.info("Initializing model with default templates...")
+        input("Press Enter to continue...")
+
         from clarifai.cli.templates.model_templates import (
             get_config_template,
             get_model_template,
@@ -725,24 +749,39 @@ def local_runner(ctx, model_path, pool_size, verbose):
 
     # Now we need to create a version for the model if no version exists. Only need one version that
     # mentions it's a local runner.
-    model_versions = [v for v in model.list_versions()]
+    model_versions = list(model.list_versions())
     method_signatures = builder.get_method_signatures(mocking=False)
+
+    create_new_version = False
     if len(model_versions) == 0:
         logger.warning("No model versions found. Creating a new version for local runner.")
+        create_new_version = True
+    else:
+        # Try to patch the latest version, and fallback to creating a new one if that fails.
+        latest_version = model_versions[0]
+        logger.warning(f"Attempting to patch latest version: {latest_version.model_version.id}")
+
+        try:
+            patched_model = model.patch_version(
+                version_id=latest_version.model_version.id,
+                pretrained_model_config={"local_dev": True},
+                method_signatures=method_signatures,
+            )
+            patched_model.load_info()
+            version = patched_model.model_version
+            logger.info(f"Successfully patched version {version.id}")
+            ctx.obj.current.CLARIFAI_MODEL_VERSION_ID = version.id
+            ctx.obj.to_yaml()  # save to yaml file.
+        except Exception as e:
+            logger.warning(f"Failed to patch model version: {e}. Creating a new version instead.")
+            create_new_version = True
+
+    if create_new_version:
         version = model.create_version(
             pretrained_model_config={"local_dev": True}, method_signatures=method_signatures
         ).model_version
         ctx.obj.current.CLARIFAI_MODEL_VERSION_ID = version.id
         ctx.obj.to_yaml()
-    else:
-        model.patch_version(
-            version_id=model_versions[0].model_version.id,
-            pretrained_model_config={"local_dev": True},
-            method_signatures=method_signatures,
-        )
-        version = model_versions[0].model_version
-        ctx.obj.current.CLARIFAI_MODEL_VERSION_ID = version.id
-        ctx.obj.to_yaml()  # save to yaml file.
 
     logger.info(f"Current model version {version.id}")
 
@@ -907,6 +946,114 @@ def local_runner(ctx, model_path, pool_size, verbose):
     )
 
 
+def _parse_json_param(param_value, param_name):
+    """Parse JSON parameter with error handling.
+
+    Args:
+        param_value: The JSON string to parse
+        param_name: Name of the parameter for error messages
+
+    Returns:
+        dict: Parsed JSON dictionary
+
+    Raises:
+        ValueError: If JSON parsing fails
+    """
+    if not param_value or param_value == '{}':
+        return {}
+    try:
+        return json.loads(param_value)
+    except json.JSONDecodeError as e:
+        logger.error(f"ValueError: Invalid JSON in --{param_name} parameter: {e}")
+        raise click.Abort()
+
+
+def _process_multimodal_inputs(inputs_dict):
+    """Process inputs to convert URLs and file paths to appropriate data types.
+
+    Args:
+        inputs_dict: Dictionary of input parameters
+
+    Returns:
+        dict: Processed inputs with Image/Video/Audio objects where appropriate
+    """
+    from clarifai.runners.utils.data_types import Audio, Image, Video
+
+    for key, value in list(inputs_dict.items()):
+        if not isinstance(value, str):
+            continue
+
+        if value.startswith(("http://", "https://")):
+            # Convert URL strings to appropriate data types
+            if "image" in key.lower():
+                inputs_dict[key] = Image(url=value)
+            elif "video" in key.lower():
+                inputs_dict[key] = Video(url=value)
+            elif "audio" in key.lower():
+                inputs_dict[key] = Audio(url=value)
+        elif os.path.isfile(value):
+            # Convert file paths to appropriate data types
+            try:
+                with open(value, "rb") as f:
+                    file_bytes = f.read()
+                if "image" in key.lower():
+                    inputs_dict[key] = Image(bytes=file_bytes)
+                elif "video" in key.lower():
+                    inputs_dict[key] = Video(bytes=file_bytes)
+                elif "audio" in key.lower():
+                    inputs_dict[key] = Audio(bytes=file_bytes)
+            except IOError as e:
+                logger.error(f"ValueError: Failed to read file {value}: {e}")
+                raise click.Abort()
+
+    return inputs_dict
+
+
+def _validate_model_params(model_id, user_id, app_id, model_url):
+    """Validate model identification parameters.
+
+    Args:
+        model_id: Model ID
+        user_id: User ID
+        app_id: App ID
+        model_url: Model URL
+
+    Raises:
+        ValueError: If validation fails
+    """
+    # Check if we have either (model_id, user_id, app_id) or model_url
+    has_triple = all([model_id, user_id, app_id])
+    has_url = bool(model_url)
+
+    if not (has_triple or has_url):
+        logger.error(
+            "ValueError: Either --model_id & --user_id & --app_id or --model_url must be provided."
+        )
+        raise click.Abort()
+
+
+def _validate_compute_params(compute_cluster_id, nodepool_id, deployment_id):
+    """Validate compute cluster parameters.
+
+    Args:
+        compute_cluster_id: Compute cluster ID
+        nodepool_id: Nodepool ID
+        deployment_id: Deployment ID
+
+    Raises:
+        ValueError: If validation fails
+    """
+    if any([compute_cluster_id, nodepool_id, deployment_id]):
+        has_cluster_nodepool = bool(compute_cluster_id) and bool(nodepool_id)
+        has_deployment = bool(deployment_id)
+
+        if not (has_cluster_nodepool or has_deployment):
+            logger.error(
+                "ValueError: Either --compute_cluster_id & --nodepool_id or --deployment_id must be provided."
+            )
+            raise click.Abort()
+
+
 @model.command(help="Perform a prediction using the model.")
 @click.option(
     '--config',
@@ -918,10 +1065,6 @@ def local_runner(ctx, model_path, pool_size, verbose):
 @click.option('--user_id', required=False, help='User ID of the model used to predict.')
 @click.option('--app_id', required=False, help='App ID of the model used to predict.')
 @click.option('--model_url', required=False, help='Model URL of the model used to predict.')
-@click.option('--file_path', required=False, help='File path of file for the model to predict')
-@click.option('--url', required=False, help='URL to the file for the model to predict')
-@click.option('--bytes', required=False, help='Bytes to the file for the model to predict')
-@click.option('--input_type', required=False, help='Type of input')
 @click.option(
     '-cc_id',
     '--compute_cluster_id',
@@ -933,9 +1076,17 @@ def local_runner(ctx, model_path, pool_size, verbose):
     '-dpl_id', '--deployment_id', required=False, help='Deployment ID to use for the model'
 )
 @click.option(
-    '--inference_params', required=False, default='{}', help='Inference parameters to override'
+    '-dpl_usr_id',
+    '--deployment_user_id',
+    required=False,
+    help='User ID to use for runner selector (organization or user). If not provided, defaults to PAT owner user_id.',
 )
-@click.option('--output_config', required=False, default='{}', help='Output config to override')
+@click.option(
+    '--inputs',
+    required=False,
+    help='JSON string of input parameters for pythonic models (e.g., \'{"prompt": "Hello", "max_tokens": 100}\')',
+)
+@click.option('--method', required=False, default='predict', help='Method to call on the model.')
 @click.pass_context
 def predict(
     ctx,
@@ -944,133 +1095,104 @@ def predict(
     user_id,
     app_id,
     model_url,
-    file_path,
-    url,
-    bytes,
-    input_type,
     compute_cluster_id,
     nodepool_id,
     deployment_id,
-    inference_params,
-    output_config,
+    deployment_user_id,
+    inputs,
+    method,
 ):
-    """Predict using the given model"""
-    import json
+    """Predict using a Clarifai model.
+
+    \b
+    Model Identification:
+        Use either --model_url OR the combination of --model_id, --user_id, and --app_id
+
+    \b
+    Input Methods:
+        --inputs: JSON string with parameters (e.g., '{"prompt": "Hello", "max_tokens": 100}')
+        --method: Method to call on the model (default is 'predict')
+
+    \b
+    Compute Options:
+        Use either --deployment_id OR both --compute_cluster_id and --nodepool_id
 
+    \b
+    Examples:
+        Text model:
+            clarifai model predict --model_url <url> --inputs '{"prompt": "Hello world"}'
+
+        With compute cluster:
+            clarifai model predict --model_id <id> --user_id <uid> --app_id <aid> \\
+                                  --compute_cluster_id <cc_id> --nodepool_id <np_id> \\
+                                  --inputs '{"prompt": "Hello"}'
+    """
     from clarifai.client.model import Model
+    from clarifai.urls.helper import ClarifaiUrlHelper
     from clarifai.utils.cli import from_yaml, validate_context
 
     validate_context(ctx)
+
+    # Load configuration from file if provided
     if config:
-        config = from_yaml(config)
-        (
-            model_id,
-            user_id,
-            app_id,
-            model_url,
-            file_path,
-            url,
-            bytes,
-            input_type,
-            compute_cluster_id,
-            nodepool_id,
-            deployment_id,
-            inference_params,
-            output_config,
-        ) = (
-            config.get(k, v)
-            for k, v in [
-                ('model_id', model_id),
-                ('user_id', user_id),
-                ('app_id', app_id),
-                ('model_url', model_url),
-                ('file_path', file_path),
-                ('url', url),
-                ('bytes', bytes),
-                ('input_type', input_type),
-                ('compute_cluster_id', compute_cluster_id),
-                ('nodepool_id', nodepool_id),
-                ('deployment_id', deployment_id),
-                ('inference_params', inference_params),
-                ('output_config', output_config),
-            ]
-        )
-    if (
-        sum(
-            [
-                opt[1]
-                for opt in [(model_id, 1), (user_id, 1), (app_id, 1), (model_url, 3)]
-                if opt[0]
-            ]
-        )
-        != 3
-    ):
-        raise ValueError(
-            "Either --model_id & --user_id & --app_id or --model_url must be provided."
-        )
-    if compute_cluster_id or nodepool_id or deployment_id:
-        if (
-            sum(
-                [
-                    opt[1]
-                    for opt in [(compute_cluster_id, 0.5), (nodepool_id, 0.5), (deployment_id, 1)]
-                    if opt[0]
-                ]
-            )
-            != 1
-        ):
-            raise ValueError(
-                "Either --compute_cluster_id & --nodepool_id or --deployment_id must be provided."
-            )
-    if model_url:
-        model = Model(
-            url=model_url,
-            pat=ctx.obj['pat'],
-            base_url=ctx.obj['base_url'],
-            compute_cluster_id=compute_cluster_id,
-            nodepool_id=nodepool_id,
-            deployment_id=deployment_id,
+        config_data = from_yaml(config)
+        # Override None values with config data
+        model_id = model_id or config_data.get('model_id')
+        user_id = user_id or config_data.get('user_id')
+        app_id = app_id or config_data.get('app_id')
+        model_url = model_url or config_data.get('model_url')
+        compute_cluster_id = compute_cluster_id or config_data.get('compute_cluster_id')
+        nodepool_id = nodepool_id or config_data.get('nodepool_id')
+        deployment_id = deployment_id or config_data.get('deployment_id')
+        deployment_user_id = deployment_user_id or config_data.get('deployment_user_id')
+        inputs = inputs or config_data.get('inputs')
+        method = method or config_data.get('method', 'predict')
+
+    # Validate parameters
+    _validate_model_params(model_id, user_id, app_id, model_url)
+    _validate_compute_params(compute_cluster_id, nodepool_id, deployment_id)
+
+    # Generate model URL if not provided
+    if not model_url:
+        model_url = ClarifaiUrlHelper.clarifai_url(
+            user_id=user_id, app_id=app_id, resource_type="models", resource_id=model_id
         )
+    logger.debug(f"Using model at URL: {model_url}")
+
+    # Create model instance
+    model = Model(
+        url=model_url,
+        pat=ctx.obj.current.pat,
+        base_url=ctx.obj.current.api_base,
+        compute_cluster_id=compute_cluster_id,
+        nodepool_id=nodepool_id,
+        deployment_id=deployment_id,
+        deployment_user_id=deployment_user_id,
+    )
+
+    model_methods = model.client.available_methods()
+    stream_method = (
+        model.client.method_signature(method).split()[-1][:-1].lower().startswith('iter')
+    )
+
+    # Determine prediction method and execute
+    if inputs and (method in model_methods):
+        # Pythonic model prediction with JSON inputs
+        inputs_dict = _parse_json_param(inputs, "inputs")
+        inputs_dict = _process_multimodal_inputs(inputs_dict)
+        model_prediction = getattr(model, method)(**inputs_dict)
     else:
-        model = Model(
-            model_id=model_id,
-            user_id=user_id,
-            app_id=app_id,
-            pat=ctx.obj['pat'],
-            base_url=ctx.obj['base_url'],
-            compute_cluster_id=compute_cluster_id,
-            nodepool_id=nodepool_id,
-            deployment_id=deployment_id,
+        logger.error(
+            f"ValueError: The model does not support the '{method}' method. Please check the model's capabilities."
         )
+        raise click.Abort()
 
-    if inference_params:
-        inference_params = json.loads(inference_params)
-    if output_config:
-        output_config = json.loads(output_config)
-
-    if file_path:
-        model_prediction = model.predict_by_filepath(
-            filepath=file_path,
-            input_type=input_type,
-            inference_params=inference_params,
-            output_config=output_config,
-        )
-    elif url:
-        model_prediction = model.predict_by_url(
-            url=url,
-            input_type=input_type,
-            inference_params=inference_params,
-            output_config=output_config,
-        )
-    elif bytes:
-        bytes = str.encode(bytes)
-        model_prediction = model.predict_by_bytes(
-            input_bytes=bytes,
-            input_type=input_type,
-            inference_params=inference_params,
-            output_config=output_config,
-        )  ## TO DO: Add support for input_id
-    click.echo(model_prediction)
+    if stream_method:
+        for chunk in model_prediction:
+            click.echo(chunk, nl=False)
+        click.echo()  # Ensure a newline after the stream ends
+    else:
+        click.echo(model_prediction)
 
 
 @model.command(name="list")
@@ -1096,7 +1218,7 @@ def list_model(ctx, user_id, app_id):
     from clarifai.client import User
 
     try:
-        pat = ctx.obj.contexts["default"]["env"]["CLARIFAI_PAT"]
+        pat = ctx.obj.current.pat
     except Exception as e:
         pat = None
 

{clarifai-11.7.5 → clarifai-11.7.5rc1}/clarifai/cli/pipeline.py

@@ -229,12 +229,34 @@ def init(pipeline_path):
     # Create the pipeline directory if it doesn't exist
     os.makedirs(pipeline_path, exist_ok=True)
 
+    # Prompt for user inputs
+    click.echo("Welcome to Clarifai Pipeline Initialization!")
+    click.echo("Please provide the following information:")
+
+    user_id = click.prompt("User ID", type=str)
+    app_id = click.prompt("App ID", type=str)
+    pipeline_id = click.prompt("Pipeline ID", default="hello-world-pipeline", type=str)
+    num_steps = click.prompt("Number of pipeline steps", default=2, type=int)
+
+    # Get step names
+    step_names = []
+    default_names = ["stepA", "stepB", "stepC", "stepD", "stepE", "stepF"]
+
+    for i in range(num_steps):
+        default_name = default_names[i] if i < len(default_names) else f"step{i + 1}"
+        step_name = click.prompt(f"Name for step {i + 1}", default=default_name, type=str)
+        step_names.append(step_name)
+
+    click.echo(f"\nCreating pipeline '{pipeline_id}' with steps: {', '.join(step_names)}")
+
     # Create pipeline config.yaml
     config_path = os.path.join(pipeline_path, "config.yaml")
     if os.path.exists(config_path):
         logger.warning(f"File {config_path} already exists, skipping...")
     else:
-        config_template = get_pipeline_config_template()
+        config_template = get_pipeline_config_template(
+            pipeline_id=pipeline_id, user_id=user_id, app_id=app_id, step_names=step_names
+        )
         with open(config_path, 'w', encoding='utf-8') as f:
             f.write(config_template)
         logger.info(f"Created {config_path}")
@@ -249,8 +271,8 @@ def init(pipeline_path):
             f.write(readme_template)
         logger.info(f"Created {readme_path}")
 
-    # Create pipeline steps (stepA and stepB)
-    for step_id in ["stepA", "stepB"]:
+    # Create pipeline steps
+    for step_id in step_names:
         step_dir = os.path.join(pipeline_path, step_id)
         os.makedirs(step_dir, exist_ok=True)
 
@@ -263,7 +285,9 @@ def init(pipeline_path):
         if os.path.exists(step_config_path):
             logger.warning(f"File {step_config_path} already exists, skipping...")
         else:
-            step_config_template = get_pipeline_step_config_template(step_id)
+            step_config_template = get_pipeline_step_config_template(
+                step_id=step_id, user_id=user_id, app_id=app_id
+            )
             with open(step_config_path, 'w', encoding='utf-8') as f:
                 f.write(step_config_template)
             logger.info(f"Created {step_config_path}")
@@ -290,13 +314,9 @@ def init(pipeline_path):
 
     logger.info(f"Pipeline initialization complete in {pipeline_path}")
     logger.info("Next steps:")
-    logger.info("1. Search for '# TODO: please fill in' comments in the generated files")
-    logger.info("2. Update your user_id and app_id in all config.yaml files")
-    logger.info(
-        "3. Implement your pipeline step logic in stepA/1/pipeline_step.py and stepB/1/pipeline_step.py"
-    )
-    logger.info("4. Add dependencies to requirements.txt files as needed")
-    logger.info("5. Run 'clarifai pipeline upload config.yaml' to upload your pipeline")
+    logger.info("1. Implement your pipeline step logic in the generated pipeline_step.py files")
+    logger.info("2. Add dependencies to requirements.txt files as needed")
+    logger.info("3. Run 'clarifai pipeline upload config.yaml' to upload your pipeline")
 
 
 @pipeline.command(['ls'])