feat(api): add file_processor API skeleton

This change adds a file_processor API skeleton that provides a foundationfor converting files into structured content for vector store ingestionwith support for chunking strategies and optional embedding generation.

Signed-off-by: Alina Ryan <[email protected]>

Author: alinaryan

docs/docs/providers/file_processor/index.mdx

@@ -0,0 +1,10 @@
+---
+sidebar_label: File Processor
+title: File_Processor
+---
+
+# File_Processor
+
+## Overview
+
+This section contains documentation for all available providers for the **file_processor** API.

docs/docs/providers/file_processor/inline_reference.mdx

@@ -0,0 +1,17 @@
+---
+description: "Reference file processor implementation (placeholder for development)"
+sidebar_label: Reference
+title: inline::reference
+---
+
+# inline::reference
+
+## Description
+
+Reference file processor implementation (placeholder for development)
+
+## Sample Configuration
+
+```yaml
+{}
+```

src/llama_stack/apis/datatypes.py

@@ -127,6 +127,7 @@ class Api(Enum, metaclass=DynamicApiMeta):
     files = "files"
     prompts = "prompts"
     conversations = "conversations"
+    file_processor = "file_processor"
 
     # built-in API
     inspect = "inspect"

src/llama_stack/apis/file_processor/__init__.py

@@ -0,0 +1,7 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from .file_processor import *

src/llama_stack/apis/file_processor/file_processor.py

@@ -0,0 +1,96 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Any, Protocol, runtime_checkable
+
+from pydantic import BaseModel
+
+from llama_stack.apis.common.tracing import telemetry_traceable
+from llama_stack.apis.vector_io.vector_io import Chunk, VectorStoreChunkingStrategy
+from llama_stack.apis.version import LLAMA_STACK_API_V1ALPHA
+from llama_stack.schema_utils import json_schema_type, webmethod
+
+
+@json_schema_type
+class ProcessFileRequest(BaseModel):
+    """Request for processing a file into structured content."""
+
+    file_data: bytes
+    """Raw file data to process."""
+
+    filename: str
+    """Original filename for format detection and processing hints."""
+
+    options: dict[str, Any] | None = None
+    """Optional processing options. Provider-specific parameters."""
+
+    chunking_strategy: VectorStoreChunkingStrategy | None = None
+    """Optional chunking strategy for splitting content into chunks."""
+
+    include_embeddings: bool = False
+    """Whether to generate embeddings for chunks."""
+
+
+@json_schema_type
+class ProcessedContent(BaseModel):
+    """Result of file processing operation."""
+
+    content: str
+    """Extracted text content from the file."""
+
+    chunks: list[Chunk] | None = None
+    """Optional chunks if chunking strategy was provided."""
+
+    embeddings: list[list[float]] | None = None
+    """Optional embeddings for chunks if requested."""
+
+    metadata: dict[str, Any]
+    """Processing metadata including processor name, timing, and provider-specific data."""
+
+
+@telemetry_traceable
+@runtime_checkable
+class FileProcessor(Protocol):
+    """
+    File Processor API for converting files into structured, processable content.
+
+    This API provides a flexible interface for processing various file formats
+    (PDFs, documents, images, etc.) into text content that can be used for
+    vector store ingestion, RAG applications, or standalone content extraction.
+
+    The API supports:
+    - Multiple file formats through extensible provider architecture
+    - Configurable processing options per provider
+    - Integration with vector store chunking strategies
+    - Optional embedding generation for chunks
+    - Rich metadata about processing results
+
+    Future providers can extend this interface to support additional formats,
+    processing capabilities, and optimization strategies.
+    """
+
+    @webmethod(route="/file-processor/process", method="POST", level=LLAMA_STACK_API_V1ALPHA)
+    async def process_file(
+        self,
+        file_data: bytes,
+        filename: str,
+        options: dict[str, Any] | None = None,
+        chunking_strategy: VectorStoreChunkingStrategy | None = None,
+        include_embeddings: bool = False,
+    ) -> ProcessedContent:
+        """
+        Process a file into structured content with optional chunking and embeddings.
+
+        This method processes raw file data and converts it into text content for applications such as vector store ingestion.
+
+        :param file_data: Raw bytes of the file to process.
+        :param filename: Original filename for format detection.
+        :param options: Provider-specific processing options (e.g., OCR settings, output format).
+        :param chunking_strategy: Optional strategy for splitting content into chunks.
+        :param include_embeddings: Whether to generate embeddings for chunks.
+        :returns: ProcessedContent with extracted text, optional chunks, and metadata.
+        """
+        ...

src/llama_stack/core/resolver.py

@@ -16,6 +16,7 @@
 from llama_stack.apis.datasets import Datasets
 from llama_stack.apis.datatypes import ExternalApiSpec
 from llama_stack.apis.eval import Eval
+from llama_stack.apis.file_processor import FileProcessor
 from llama_stack.apis.files import Files
 from llama_stack.apis.inference import Inference, InferenceProvider
 from llama_stack.apis.inspect import Inspect
@@ -96,6 +97,7 @@ def api_protocol_map(external_apis: dict[Api, ExternalApiSpec] | None = None) ->
         Api.files: Files,
         Api.prompts: Prompts,
         Api.conversations: Conversations,
+        Api.file_processor: FileProcessor,
     }
 
     if external_apis:

src/llama_stack/distributions/ci-tests/build.yaml

@@ -29,6 +29,8 @@ distribution_spec:
     - provider_type: remote::weaviate
     files:
     - provider_type: inline::localfs
+    file_processor:
+    - provider_type: inline::reference
     safety:
     - provider_type: inline::llama-guard
     - provider_type: inline::code-scanner

src/llama_stack/distributions/ci-tests/run.yaml

@@ -5,6 +5,7 @@ apis:
 - batches
 - datasetio
 - eval
+- file_processor
 - files
 - inference
 - post_training
@@ -154,6 +155,9 @@ providers:
       metadata_store:
         table_name: files_metadata
         backend: sql_default
+  file_processor:
+  - provider_id: reference
+    provider_type: inline::reference
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard

src/llama_stack/distributions/starter-gpu/build.yaml

@@ -30,6 +30,8 @@ distribution_spec:
     - provider_type: remote::weaviate
     files:
     - provider_type: inline::localfs
+    file_processor:
+    - provider_type: inline::reference
     safety:
     - provider_type: inline::llama-guard
     - provider_type: inline::code-scanner

src/llama_stack/distributions/starter-gpu/run-with-postgres-store.yaml

@@ -5,6 +5,7 @@ apis:
 - batches
 - datasetio
 - eval
+- file_processor
 - files
 - inference
 - post_training
@@ -154,6 +155,9 @@ providers:
       metadata_store:
         table_name: files_metadata
         backend: sql_default
+  file_processor:
+  - provider_id: reference
+    provider_type: inline::reference
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard