Added draft implementation of built-in preprocessing for RAG.

llama_stack/providers/inline/preprocessing/simple_chunking/__init__.py

@@ -0,0 +1,18 @@
+# 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 .config import InclineSimpleChunkingConfig
+
+
+async def get_provider_impl(
+    config: InclineSimpleChunkingConfig,
+    _deps,
+):
+    from .simple_chunking import InclineSimpleChunkingImpl
+
+    impl = InclineSimpleChunkingImpl(config)
+    await impl.initialize()
+    return impl

llama_stack/providers/inline/preprocessing/simple_chunking/config.py

@@ -0,0 +1,11 @@
+# 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 pydantic import BaseModel
+
+
+class InclineSimpleChunkingConfig(BaseModel):
+    chunk_size_in_tokens: int = 512
+    chunk_overlap_ratio: int = 4

llama_stack/providers/inline/preprocessing/simple_chunking/simple_chunking.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.
+import logging
+from enum import Enum
+from typing import List, Tuple
+
+from llama_models.llama3.api import Tokenizer
+
+from llama_stack.apis.preprocessing import (
+    Preprocessing,
+    PreprocessingDataType,
+    PreprocessingInput,
+    PreprocessingResponse,
+    Preprocessor,
+    PreprocessorOptions,
+)
+from llama_stack.apis.vector_io import Chunk
+from llama_stack.providers.datatypes import PreprocessorsProtocolPrivate
+from llama_stack.providers.inline.preprocessing.simple_chunking import InclineSimpleChunkingConfig
+
+log = logging.getLogger(__name__)
+
+
+class SimpleChunkingOptions(Enum):
+    chunk_size_in_tokens = "chunk_size_in_tokens"
+    chunk_overlap_ratio = "chunk_overlap_ratio"
+
+
+class InclineSimpleChunkingImpl(Preprocessing, PreprocessorsProtocolPrivate):
+    # this preprocessor receives plain text and returns chunks
+    INPUT_TYPES = [PreprocessingDataType.raw_text_document]
+    OUTPUT_TYPES = [PreprocessingDataType.chunks]
+
+    def __init__(self, config: InclineSimpleChunkingConfig) -> None:
+        self.config = config
+
+    async def initialize(self) -> None: ...
+
+    async def shutdown(self) -> None: ...
+
+    async def register_preprocessor(self, preprocessor: Preprocessor) -> None: ...
+
+    async def unregister_preprocessor(self, preprocessor_id: str) -> None: ...
+
+    async def preprocess(
+        self,
+        preprocessor_id: str,
+        preprocessor_inputs: List[PreprocessingInput],
+        options: PreprocessorOptions,
+    ) -> PreprocessingResponse:
+        chunks = []
+
+        window_len, overlap_len = self._resolve_chunk_size_params(options)
+
+        for inp in preprocessor_inputs:
+            new_chunks = self.make_overlapped_chunks(
+                inp.preprocessor_input_id, inp.path_or_content, window_len, overlap_len
+            )
+            chunks.extend(new_chunks)
+
+        return PreprocessingResponse(status=True, results=chunks)
+
+    def _resolve_chunk_size_params(self, options: PreprocessorOptions) -> Tuple[int, int]:
+        window_len = options.get(str(SimpleChunkingOptions.chunk_size_in_tokens), self.config.chunk_size_in_tokens)
+
+        chunk_overlap_ratio = options.get(
+            str(SimpleChunkingOptions.chunk_overlap_ratio), self.config.chunk_overlap_ratio
+        )
+        overlap_len = window_len // chunk_overlap_ratio
+
+        return window_len, overlap_len
+
+    @staticmethod
+    def make_overlapped_chunks(document_id: str, text: str, window_len: int, overlap_len: int) -> List[Chunk]:
+        tokenizer = Tokenizer.get_instance()
+        tokens = tokenizer.encode(text, bos=False, eos=False)
+
+        chunks = []
+        for i in range(0, len(tokens), window_len - overlap_len):
+            toks = tokens[i : i + window_len]
+            chunk = tokenizer.decode(toks)
+            # chunk is a string
+            chunks.append(
+                Chunk(
+                    content=chunk,
+                    metadata={
+                        "token_count": len(toks),
+                        "document_id": document_id,
+                    },
+                )
+            )
+
+        return chunks